- 24 10月, 2014 8 次提交
-
-
由 Emmanuel Grumbach 提交于
This reverts commit aa11bbf3. This commit was causing connection issues and is not needed if IWL_MVM_RS_RSSI_BASED_INIT_RATE is set to false by default. Regardless of the issues mentioned above, this patch added the following WARNING: WARNING: CPU: 0 PID: 3946 at drivers/net/wireless/iwlwifi/mvm/tx.c:190 iwl_mvm_set_tx_params+0x60a/0x6f0 [iwlmvm]() Got an HT rate for a non data frame 0x8 CPU: 0 PID: 3946 Comm: wpa_supplicant Tainted: G O 3.17.0+ #6 Hardware name: LENOVO 20ANCTO1WW/20ANCTO1WW, BIOS GLET71WW (2.25 ) 07/02/2014 0000000000000009 ffffffff814fa911 ffff8804288db8f8 ffffffff81064f52 0000000000001808 ffff8804288db948 ffff88040add8660 ffff8804291b5600 0000000000000000 ffffffff81064fb7 ffffffffa07b73d0 0000000000000020 Call Trace: [<ffffffff814fa911>] ? dump_stack+0x41/0x51 [<ffffffff81064f52>] ? warn_slowpath_common+0x72/0x90 [<ffffffff81064fb7>] ? warn_slowpath_fmt+0x47/0x50 [<ffffffffa07a39ea>] ? iwl_mvm_set_tx_params+0x60a/0x6f0 [iwlmvm] [<ffffffffa07a3cf8>] ? iwl_mvm_tx_skb+0x48/0x3c0 [iwlmvm] [<ffffffffa079cb9b>] ? iwl_mvm_mac_tx+0x7b/0x180 [iwlmvm] [<ffffffffa0746ce9>] ? __ieee80211_tx+0x2b9/0x3c0 [mac80211] [<ffffffffa07492f3>] ? ieee80211_tx+0xb3/0x100 [mac80211] [<ffffffffa0749c49>] ? ieee80211_subif_start_xmit+0x459/0xca0 [mac80211] [<ffffffff814116e7>] ? dev_hard_start_xmit+0x337/0x5f0 [<ffffffff81430d46>] ? sch_direct_xmit+0x96/0x1f0 [<ffffffff81411ba3>] ? __dev_queue_xmit+0x203/0x4f0 [<ffffffff8142f670>] ? ether_setup+0x70/0x70 [<ffffffff814e96a1>] ? packet_sendmsg+0xf81/0x1110 [<ffffffff8140625c>] ? skb_free_datagram+0xc/0x40 [<ffffffff813f7538>] ? sock_sendmsg+0x88/0xc0 [<ffffffff813f7274>] ? move_addr_to_kernel.part.20+0x14/0x60 [<ffffffff811c47c2>] ? __inode_wait_for_writeback+0x62/0xb0 [<ffffffff813f7a91>] ? SYSC_sendto+0xf1/0x180 [<ffffffff813f88f9>] ? __sys_recvmsg+0x39/0x70 [<ffffffff8150066d>] ? system_call_fastpath+0x1a/0x1f ---[ end trace cc19a150d311fc63 ]--- which was reported here: https://bugzilla.kernel.org/show_bug.cgi?id=85691 CC: <stable@vger.kernel.org> [3.13+] Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
由 Emmanuel Grumbach 提交于
When mac80211 wants to ensure that a frame is sent, it calls the flush() callback. Until now, iwldvm implemented this by waiting that all the frames are sent (ACKed or timeout). In case of weak signal, this can take a significant amount of time, delaying the next connection (in case of roaming). Many users have reported that the flush would take too long leading to the following error messages to be printed: iwlwifi 0000:03:00.0: fail to flush all tx fifo queues Q 2 iwlwifi 0000:03:00.0: Current SW read_ptr 161 write_ptr 201 iwl data: 00000000: 00 00 00 00 00 00 00 00 fe ff 01 00 00 00 00 00 [snip] iwlwifi 0000:03:00.0: FH TRBs(0) = 0x00000000 [snip] iwlwifi 0000:03:00.0: Q 0 is active and mapped to fifo 3 ra_tid 0x0000 [9,9] [snip] Instead of waiting for these packets, simply drop them. This significantly improves the responsiveness of the network. Note that all the queues are flushed, but the VO one. This is not typically used by the applications and it likely contains management frames that are useful for connection or roaming. This bug is tracked here: https://bugzilla.kernel.org/show_bug.cgi?id=56581 But it is duplicated in distributions' trackers. A simple search in Ubuntu's database led to these bugs: https://bugs.launchpad.net/ubuntu/+source/linux-firmware/+bug/1270808 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1305406 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1356236 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1360597 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1361809 Cc: <stable@vger.kernel.org> Depends-on: 77be2c54 ("mac80211: add vif to flush call") Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
由 Matti Gottlieb 提交于
Don't add the time event to the list. We added it several times the same time event, which leads to an infinite loop when walking the list. Since we (currently) don't support more than one ROC for STA vif at a time, enforce this and don't add the time event to any list. We were also missing the locking of the mutex which led to a lockdep splat - fix that. Signed-off-by: NMatti Gottlieb <matti.gottlieb@intel.com> Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
由 Haim Dreyfuss 提交于
The tx power should be limited from many reasons. currently, setting the tx power is available by the mvm only for station interface. Adding the tx power condition to bss_info_changed_ap_ibss make it available also for AP. Signed-off-by: NHaim Dreyfuss <haim.dreyfuss@intel.com> Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
由 Emmanuel Grumbach 提交于
The probe requests sent during scan must get BT prio 3. Fix that. Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
由 Emmanuel Grumbach 提交于
Cc: <stable@vger.kernel.org> [3.16+] Fixes: 2adc8949 ("iwlwifi: mvm: BT Coex - fix boost register / LUT values") Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
由 Liad Kaufman 提交于
I changed the string but forgot to update the fix also to MODULE_FIRMWARE(). Signed-off-by: NLiad Kaufman <liad.kaufman@intel.com> Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
由 Emmanuel Grumbach 提交于
The LTR is the handshake between the device and the root complex about the latency allowed when the bus exits power save. This configuration was missing and this led to high latency in the link power up. The end user could experience high latency in the network because of this. Cc: <stable@vger.kernel.org> [3.10+] Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
-
- 09 10月, 2014 8 次提交
-
-
由 Eric Dumazet 提交于
Add two helpers so that drivers do not have to care of BQL being available or not. Signed-off-by: NEric Dumazet <edumazet@google.com> Reported-by: NJim Davis <jim.epost@gmail.com> Fixes: 29d40c90 ("net/mlx4_en: Use prefetch in tx path") Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Dan Carpenter 提交于
The tx_desc struct holds 8 __be64 values. The original code in ring_tx_db() took a tx_desc pointer then casted it to an int pointer and then casted it to a u64 pointer. It was confusing and triggered some static checker warnings. I have changed the cxgb_pio_copy() function to only take tx_desc pointers. This isn't really a loss of flexibility because anything else was buggy to begin with. I also removed the casting on the destination pointer since that was unnecessary and a bit messy. Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Dan Carpenter 提交于
"cntxt_id" is an unsigned int but "udb" is a u64 so there is a potential shift wrapping bug here. Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Support skb->xmit_more in i40e is straightforward : we need to move around i40e_maybe_stop_tx() call to correctly test netif_xmit_stopped() before taking the decision to not kick the NIC. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDaniel Borkmann <dborkman@redhat.com> Acked-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 LEROY Christophe 提交于
When using a MPC8xx as a router, 'perf' shows a significant time spent in fs_enet_interrupt() and fs_enet_start_xmit(). 'perf annotate' shows that the time spent in fs_enet_start_xmit is indeed spent between spin_unlock_irqrestore() and the following instruction, hence in interrupt handling. This is due to the TX complete interrupt that fires after each transmitted packet. This patch modifies the handling of TX complete to use NAPI. With this patch, my NAT router offers a throughput improved by 21% Original performance: [root@localhost tmp]# scp toto pgs:/tmp toto 100% 256MB 2.8MB/s 01:31 Performance with the patch: [root@localhost tmp]# scp toto pgs:/tmp toto 100% 256MB 3.4MB/s 01:16 Signed-off-by: NChristophe Leroy <christophe.leroy@c-s.fr> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 LEROY Christophe 提交于
In the probe function, use_napi is inconditionnaly set to 1. This patch removes all the code which is conditional to !use_napi, and removes use_napi which has then become useless. Signed-off-by: NChristophe Leroy <christophe.leroy@c-s.fr> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Chun-Hao Lin 提交于
RTL8168EP is Realtek PCIe Gigabit Ethernet controller with DASH support. It is a successor chip of RTL8168DP. For RTL8168EP, the read/write ocp register is via eri channel type 2, so I move ocp_xxx() related functions under rtl_eri_xxx. And use r8168dp_ocp_xxx() for RTL8168DP ocp read/write, r8168ep_ocp_xxx() for RTL8168EP ocp read/write. The way of checking dash enable is different with RTL8168DP. I use r8168dp_check_dash()for RTL8168DP and r8168ep_check_dash() for RTL8168EP, to check if dash is enabled. The driver_start() and driver_stop() of RTL8168EP is also different with RTL8168DP. I use rtl8168dp_driver_xxx() for RTL8168DP and rtl8168ep_driver_xxx for RTL8168EP. Right now, RTL8168EP phy mcu did not need firmware code patch, so I did not add firmware code for it. so I did not add firmware code for it. Signed-off-by: NChun-Hao Lin <hau@realtek.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Pawel Moll 提交于
Because CCN's cycle counter always runs, it will generate an interrupt on overflow even if the relevant perf event was not requested, causing a spurious warning message. Fixed now by warning on only normal counter unwanted overflows. Also cleaning the overflow mask at init now, not to warn on event previously requested by firmware. Signed-off-by: NPawel Moll <pawel.moll@arm.com> Signed-off-by: NOlof Johansson <olof@lixom.net> Signed-off-by: NArnd Bergmann <arnd@arndb.de>
-
- 08 10月, 2014 10 次提交
-
-
由 Linus Torvalds 提交于
Commit 7bced397 ("net_dma: simple removal") removed the long-dead net_dma code, but unintentionally broke the build for the qualcomm crypto engine (qce) that had accidentally depended on the inclusion of <linux/dmaengine.h> through the networking header files that used to have it. Add the required dmaengine.h include explicitly to fix the breakage. Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
-
由 Takao Indoh 提交于
This patch fixes a bug on hotmod removing. After ipmi interface is removed using hotmod, kernel panic occurs when rmmod impi_si. For example, try this: # echo "remove,"`cat /proc/ipmi/0/params` > \ /sys/module/ipmi_si/parameters/hotmod # rmmod ipmi_si Then, rmmod fails with the following messages. ------------[ cut here ]------------ WARNING: CPU: 12 PID: 10819 at /mnt/repos/linux/lib/list_debug.c:53 __list_del_entry+0x63/0xd0() CPU: 12 PID: 10819 Comm: rmmod Not tainted 3.17.0-rc1 #19 Hardware name: FUJITSU-SV PRIMERGY BX920 S2/D3030, BIOS 080015 Rev.3D81.3030 02/10/2012 Call Trace: dump_stack+0x45/0x56 warn_slowpath_common+0x7d/0xa0 warn_slowpath_fmt+0x4c/0x50 __list_del_entry+0x63/0xd0 list_del+0xd/0x30 cleanup_one_si+0x2a/0x230 [ipmi_si] ipmi_pnp_remove+0x15/0x20 [ipmi_si] pnp_device_remove+0x24/0x40 __device_release_driver+0x7f/0xf0 driver_detach+0xb0/0xc0 bus_remove_driver+0x55/0xd0 driver_unregister+0x2c/0x50 pnp_unregister_driver+0x12/0x20 cleanup_ipmi_si+0xbc/0xf0 [ipmi_si] SyS_delete_module+0x132/0x1c0 system_call_fastpath+0x16/0x1b ---[ end trace 70b4377268f85c23 ]--- list_del in cleanup_one_si() fails because the smi_info is already removed when hotmod removing. When ipmi interface is removed by hotmod, smi_info is removed by cleanup_one_si(), but is is still set in drvdata. Therefore when rmmod ipmi_si, ipmi_pnp_remove tries to remove it again and fails. By this patch, a pointer to smi_info in drvdata is cleared when hotmod removing so that it will be not accessed when rmmod. changelog: v2: - Clear drvdata in cleanup_one_si - Change subject v1: https://lkml.org/lkml/2014/9/8/741Signed-off-by: NTakao Indoh <indou.takao@jp.fujitsu.com> Signed-off-by: NCorey Minyard <cminyard@mvista.com> Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
-
由 Arnd Bergmann 提交于
Building ipmi on arm with gcc-4.9 results in this warning for an allmodconfig build: drivers/char/ipmi/ipmi_si_intf.c: In function 'ipmi_thread': include/linux/time.h:28:5: warning: 'busy_until.tv_sec' may be used uninitialized in this function [-Wmaybe-uninitialized] if (lhs->tv_sec > rhs->tv_sec) ^ drivers/char/ipmi/ipmi_si_intf.c:1007:18: note: 'busy_until.tv_sec' was declared here struct timespec busy_until; ^ The warning is bogus and this case can not occur. Apparently this is a false positive resulting from gcc getting a little smarter about tracking assignments but not smart enough. Marking the ipmi_thread_busy_wait function as inline gives the gcc optimization logic enough information to figure out for itself that the case cannot happen, which gets rid of the warning without adding any fake initialization. Signed-off-by: NArnd Bergmann <arnd@arndb.de> Signed-off-by: NCorey Minyard <cminyard@mvista.com> Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
-
由 Benjamin Herrenschmidt 提交于
If an IPMI controller is used by the firmware and as such marked with a reserved status, we shouldn't use it. Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: NCorey Minyard <cminyard@mvista.com> Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
-
由 Corey Minyard 提交于
The code to send the channel config errors was missing an error report in one place and needed some more information in another, and had an extraneous bit of code. Clean all that up. Signed-off-by: NCorey Minyard <cminyard@mvista.com> Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
-
由 Eric Dumazet 提交于
Testing xmit_more support with netperf and connected UDP sockets, I found strange dst refcount false sharing. Current handling of IFF_XMIT_DST_RELEASE is not optimal. Dropping dst in validate_xmit_skb() is certainly too late in case packet was queued by cpu X but dequeued by cpu Y The logical point to take care of drop/force is in __dev_queue_xmit() before even taking qdisc lock. As Julian Anastasov pointed out, need for skb_dst() might come from some packet schedulers or classifiers. This patch adds new helper to cleanly express needs of various drivers or qdiscs/classifiers. Drivers that need skb_dst() in their ndo_start_xmit() should call following helper in their setup instead of the prior : dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; -> netif_keep_dst(dev); Instead of using a single bit, we use two bits, one being eventually rebuilt in bonding/team drivers. The other one, is permanent and blocks IFF_XMIT_DST_RELEASE being rebuilt in bonding/team. Eventually, we could add something smarter later. Signed-off-by: NEric Dumazet <edumazet@google.com> Cc: Julian Anastasov <ja@ssi.bg> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Drivers should avoid NETDEV_TX_BUSY as much as possible. They should stop the tx queue before qdisc even tries to push another packet, to avoid requeues. For a driver supporting skb->xmit_more, this is likely to be a prereq anyway, otherwise we could have a tx deadlock : We need to force a doorbell if TX ring is full. Signed-off-by: NEric Dumazet <edumazet@google.com> Acked-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Sylvain "ythier" Hitier 提交于
In commit 6f2b6a30, # 3c59x: Add dma error checking and recovery the intent is to split out the mapping from the byte-swapping in order to insert a dma_mapping_error() check. Kinda this semantic patch: // See http://coccinelle.lip6.fr/ // // Beware, grouik-and-dirty! @@ expression DEV, X, Y, Z; @@ - cpu_to_le32(pci_map_single(DEV, X, Y, Z)) + dma_addr_t addr = pci_map_single(DEV, X, Y, Z); + if (dma_mapping_error(&DEV->dev, addr)) + /* snip */; + cpu_to_le32(addr) However, the #else part (of the #if DO_ZEROCOPY test) is changed this way: - cpu_to_le32(pci_map_single(DEV, X, Y, Z)) + dma_addr_t addr = cpu_to_le32(pci_map_single(DEV, X, Y, Z)); // ^^^^^^^^^^^ // That mismatches the 3 other changes! + if (dma_mapping_error(&DEV->dev, addr)) + /* snip */; + cpu_to_le32(addr) Let's remove the leftover cpu_to_le32() for coherency. v2: Better changelog. v3: Add Acked-by Fixes: 6f2b6a30 # 3c59x: Add dma error checking and recovery Acked-by: NNeil Horman <nhorman@tuxdriver.com> Signed-off-by: NSylvain "ythier" Hitier <sylvain.hitier@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Petri Gynther 提交于
GENET MAC has three Tx ring priority registers: - GENET_x_TDMA_PRIORITY0 for queues 0-5 - GENET_x_TDMA_PRIORITY1 for queues 6-11 - GENET_x_TDMA_PRIORITY2 for queues 12-16 Fix bcmgenet_init_multiq() to program them correctly. Signed-off-by: NPetri Gynther <pgynther@google.com> Acked-by: NFlorian Fainelli <f.fainelli@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Petri Gynther 提交于
Adjust fixed_phy_register() to return struct phy_device *, so that it becomes easy to use fixed PHYs without device tree support: phydev = fixed_phy_register(PHY_POLL, &fixed_phy_status, NULL); fixed_phy_set_link_update(phydev, fixed_phy_link_update); phy_connect_direct(netdev, phydev, handler_fn, phy_interface); This change is a prerequisite for modifying bcmgenet driver to work without a device tree on Broadcom's MIPS-based 7xxx platforms. Signed-off-by: NPetri Gynther <pgynther@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
- 07 10月, 2014 5 次提交
-
-
由 Andy Zhou 提交于
Fix a openvswitch compilation error when CONFIG_INET is not set: ===================================================== In file included from include/net/geneve.h:4:0, from net/openvswitch/flow_netlink.c:45: include/net/udp_tunnel.h: In function 'udp_tunnel_handle_offloads': >> include/net/udp_tunnel.h:100:2: error: implicit declaration of function 'iptunnel_handle_offloads' [-Werror=implicit-function-declaration] >> return iptunnel_handle_offloads(skb, udp_csum, type); >> ^ >> >> include/net/udp_tunnel.h:100:2: warning: return makes pointer from integer without a cast >> >> cc1: some warnings being treated as errors ===================================================== Reported-by: Nkbuild test robot <fengguang.wu@intel.com> Signed-off-by: NAndy Zhou <azhou@nicira.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Ivan Vecera 提交于
When Tx VLAN offloading is disabled frames with size ~ MTU are not transmitted as the driver does not account 4 bytes of VLAN header added by stack. It should use VLAN_ETH_HLEN instead of ETH_HLEN. The second problem is with newer BNA chips (BNA 1860). These chips filter out any VLAN tagged frames in Tx path. This is a problem when Tx VLAN offloading is disabled and frames are tagged by stack. Older chips like 1010/1020 are not affected as they probably don't do such filtering. Cc: Rasesh Mody <rasesh.mody@qlogic.com> Signed-off-by: NIvan Vecera <ivecera@redhat.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 hayeswang 提交于
Restart autonegotiation is necessary after setting EEE. Signed-off-by: NHayes Wang <hayeswang@realtek.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Mahesh Bandewar 提交于
Earlier change to use usable slave array for TLB mode had an additional performance advantage. So extending the same logic to all other modes that use xmit-hash for slave selection (viz 802.3AD, and XOR modes). Also consolidating this with the earlier TLB change. The main idea is to build the usable slaves array in the control path and use that array for slave selection during xmit operation. Measured performance in a setup with a bond of 4x1G NICs with 200 instances of netperf for the modes involved (3ad, xor, tlb) cmd: netperf -t TCP_RR -H <TargetHost> -l 60 -s 5 Mode TPS-Before TPS-After 802.3ad : 468,694 493,101 TLB (lb=0): 392,583 392,965 XOR : 475,696 484,517 Signed-off-by: NMahesh Bandewar <maheshb@google.com> Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Mahesh Bandewar 提交于
It's a trivial fix to display xmit_hash_policy for this new TLB mode since it uses transmit-hash-poilicy as part of bonding-master info (/proc/net/bonding/<bonding-interface). Signed-off-by: NMahesh Bandewar <maheshb@google.com> Reviewed-by: NNikolay Aleksandrov <nikolay@redhat.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
- 06 10月, 2014 9 次提交
-
-
由 Eric Dumazet 提交于
Instead of setting inline threshold using module parameter only on driver load, use set_tunable() to set it dynamically. No need to store the threshold per ring, using instead the netdev global priv->prof->inline_thold Initial value still is set using the module parameter, therefore backward compatability is kept. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Reorganize code to call is_inline() once, so compiler can inline it Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Properly clear tx_info->ts_requested Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Access skb_headlen() once in tx flow Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Acces skb_shinfo(skb) once in tx flow. Also, rename @i variable to @i_frag to avoid confusion, as the "goto tx_drop_unmap;" relied on this @i variable. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
mlx4_en_process_tx_cq() carefully fetches and writes ring->last_nr_txbb and ring->cons only one time to avoid false sharing Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
mlx4_en_free_tx_desc() uses a prefetchw(&skb->users) to speed up consume_skb() prefetchw(&ring->tx_queue->dql) to speed up BQL update Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Add frag0_dma/frag0_byte_count into mlx4_en_tx_info to avoid a cache line miss in TX completion for frames having one dma element. (We avoid reading back the tx descriptor) Note this could be extended to 2/3 dma elements later, as we have free room in mlx4_en_tx_info Also, mlx4_en_free_tx_desc() no longer accesses skb_shinfo(). We use a new nr_maps fields in mlx4_en_tx_info to avoid 2 or 3 cache misses. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NAmir Vadai <amirv@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-