- 04 Apr, 2009 2 commits
-
-
Jesse Brandeburg authored
e1000 (and e1000e, igb, ixgbe, ixgb) all do a series of operations each time a multicast address is added. The flow goes something like 1) stack adds one multicast address 2) stack passes whole current list of unicast and multicast addresses to driver 3) driver clears entire list in hardware 4) driver programs each multicast address using iomem in a loop This was causing multicast packets to be lost during the reprogramming process. reference with test program: http://kerneltrap.org/mailarchive/linux-netdev/2009/3/14/5160514/thread Thanks to Dave Boutcher for his report and test program. This driver fix prepares an array all at once in memory and programs it in one shot to the hardware, not requiring an "erase" cycle. Signed-off-by:
Jesse Brandeburg <jesse.brandeburg@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Alexander Beregalov authored
Fix this build error: drivers/net/vxge/vxge-main.c: In function 'vxge_get_vpath_no': drivers/net/vxge/vxge-main.c:695: error: dereferencing pointer to incomplete type ... Signed-off-by:
Alexander Beregalov <a.beregalov@gmail.com> Signed-off-by:
-
- 02 Apr, 2009 38 commits
-
-
Huang Weiyi authored
Remove unused #include <version.h> in drivers/net/qlge/qlge_ethtool. Signed-off-by:
Huang Weiyi <weiyi.huang@gmail.com> Signed-off-by:
-
Huang Weiyi authored
Remove unused #include <version.h> in drivers/net/dnet.c. Signed-off-by:
-
Ilpo Järvinen authored
It seems that trivial reset of pcount to one was not sufficient in tcp_retransmit_skb. Multiple counters experience a positive miscount when skb's pcount gets lowered without the necessary adjustments (depending on skb's sacked bits which exactly), at worst a packets_out miscount can crash at RTO if the write queue is empty! Triggering this requires mss change, so bidir tcp or mtu probe or like. Signed-off-by:
Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Reported-by:
Markus Trippelsdorf <markus@trippelsdorf.de> Tested-by:
Uwe Bugla <uwe.bugla@gmx.de> Signed-off-by:
-
Ilpo Järvinen authored
We need full-scale adjustment to fix a TCP miscount in the next patch, so just move it into a helper and call for that from the other places. Signed-off-by:
-
Jan Dumon authored
Some devices cannot send very short usb transfers. To get around this the firmware adds a known pattern and flags the driver that it should check for this pattern on short transfers. This flag was not taken into account by the driver. Signed-off-by:
Jan Dumon <j.dumon@option.com> Signed-off-by:
-
Jan Dumon authored
Changed the order in which things are freed. This fixes an oops when unplugging the device while network traffic is ongoing. Signed-off-by:
-
Segher Boessenkool authored
Add EXPORT_SYMBOL_GPL(fsl_pq_mdio_bus_name) for module builds Signed-off-by:
Segher Boessenkool <segher@kernel.crashing.org> Signed-off-by:
Kumar Gala <galak@kernel.crashing.org> Signed-off-by:
-
Anton Vorontsov authored
commit 1577ecef ("netdev: Merge UCC and gianfar MDIO bus drivers") introduced a regression so that UCC MDIO buses no longer work. This is because fsl_pq_mdio driver wrongly masks all non-TBI PHYs for !fsl,gianfar-mdio buses, while it should do that only for fsl,gianfar-tbi buses. Signed-off-by:
Anton Vorontsov <avorontsov@ru.mvista.com> Signed-off-by:
-
Anton Vorontsov authored
The driver should pass a device that actually specifies internal DMA ops, but currently it passes netdev's device, which is wrong and that causes following oops: Kernel BUG at c01c4df8 [verbose debug info unavailable] Oops: Exception in kernel mode, sig: 5 [#1] [...] NIP [c01c4df8] get_new_skb+0x7c/0xf8 LR [c01c4da4] get_new_skb+0x28/0xf8 Call Trace: [ef82be00] [c01c4da4] get_new_skb+0x28/0xf8 (unreliable) [ef82be20] [c01c4eb8] rx_bd_buffer_set+0x44/0x98 [ef82be40] [c01c62bc] ucc_geth_startup+0x11b0/0x147c [ef82be80] [c01c6674] ucc_geth_open+0xec/0x2a4 [ef82bea0] [c02288a4] dev_open+0xc0/0x11c [...] Fix this by passing of_device's device that specifies DMA ops in its archdata. Signed-off-by:
-
Alan Carvalho de Assis authored
This is a fix to get cs89x0 network driver working on i.MX31ADS Signed-off-by:
Alan Carvalho de Assis <acassis@gmail.com> Signed-off-by:
-
Atsushi Nemoto authored
This driver contains experimental NAPI code disabled by default. The commit bea3348e ("[NET]: Make NAPI polling independent of struct net_device objects.") converted the NAPI path of this driver but that conversion was not complete. This patch fixes a build error introduced by the commit. Signed-off-by:
Atsushi Nemoto <anemo@mba.ocn.ne.jp> Signed-off-by:
-
Jan Dumon authored
Add Vendor/Product ID's for new devices. Removed duplicate product ID 0x7361. Signed-off-by:
-
Kumar Gala authored
Now that the driver is exclusively an of_platform driver we no longer use the structs and #defines in fsl_devices.h Signed-off-by:
-
Kumar Gala authored
Now that the driver is exclusively an of_platform driver we no longer use the structs and #defines in fsl_devices.h Signed-off-by:
-
Larry Finger authored
On an SMP system, the following message is printed. The patch below gets fixes the problem. ================================= [ INFO: inconsistent lock state ] 2.6.29-Linus-05093-gc31f403 #57 --------------------------------- inconsistent {hardirq-on-W} -> {in-hardirq-W} usage. bash/4105 [HC1[1]:SC0[0]:HE0:SE1] takes: (&kaweth->device_lock){+...}, at: [<ffffffffa01aa286>] kaweth_usb_receive+0x77/0x1af [kaw eth] {hardirq-on-W} state was registered at: [<ffffffff80260503>] __lock_acquire+0x753/0x1685 [<ffffffff8026148a>] lock_acquire+0x55/0x71 [<ffffffff80461ba6>] _spin_lock+0x31/0x3d [<ffffffffa01aaa0c>] kaweth_start_xmit+0x2b/0x1e1 [kaweth] [<ffffffff803eccd3>] dev_hard_start_xmit+0x22e/0x2ad [<ffffffff803fe120>] __qdisc_run+0xf2/0x203 [<ffffffff803ed0cd>] dev_queue_xmit+0x263/0x39b [<ffffffffa03a47cb>] packet_sendmsg_spkt+0x1c4/0x20a [af_packet] [<ffffffff803de0c2>] sock_sendmsg+0xe4/0xfd [<ffffffff803dec8f>] sys_sendto+0xe4/0x10c [<ffffffff8020bccb>] system_call_fastpath+0x16/0x1b [<ffffffffffffffff>] 0xffffffffffffffff irq event stamp: 1280 hardirqs last enabled at (1279): [<ffffffff80461a71>] _spin_unlock_irqrestore+0x44/0x4c hardirqs last disabled at (1280): [<ffffffff8020bad7>] save_args+0x67/0x70 softirqs last enabled at (660): [<ffffffff8024192c>] __do_softirq+0x14d/0x15d softirqs last disabled at (651): [<ffffffff8020ce9c>] call_softirq+0x1c/0x28 Signed-off-by:Larry Finger <Larry.Finger@lwfinger.net> Signed-off-by:
-
Stephen Hemminger authored
GRO assumes that there is a one-to-one relationship between NAPI structure and network device. Some devices like sky2 share multiple devices on a single interrupt so only have one NAPI handler. Rather than split GRO from NAPI, just have GRO assume if device changes that it is a different flow. Signed-off-by:
Stephen Hemminger <shemminger@vyatta.com> Signed-off-by:
-
Karsten Wiese authored
Original comment (Karsten): On a MSI MS-6702E mainboard, when in rtl8169_init_one() for the first time after BIOS has run, IntrStatus reads 5 after chip has been reset. IntrStatus should equal 0 there, so patch changes IntrStatus reset to happen after chip reset instead of before. Remark (Francois): Assuming that the loglevel of the driver is increased above NETIF_MSG_INTR, the bug reveals itself with a typical "interrupt 0025 in poll" message at startup. In retrospect, the message should had been read as an hint of an unexpected hardware state several months ago :o( Fixes (at least part of) https://bugzilla.redhat.com/show_bug.cgi?id=460747Signed-off-by:
Karsten Wiese <fzu@wemgehoertderstaat.de> Signed-off-by:
Francois Romieu <romieu@fr.zoreil.com> Tested-by:
Josep <josep.puigdemont@gmail.com> Signed-off-by:
-
Mallikarjuna R Chilakala authored
While setting up the ring parameters using ethtool the driver can panic or leak memory as ixgbe_open tries to setup tx & rx resources. The updated logic will use ixgbe_down/up after successful allocation of tx & rx resources Signed-off-by:
Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com> Signed-off-by:
Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com> Signed-off-by:
-
Don Skidmore authored
We were basicly ignoring ethtool users request for FC autoneg and replying to queries with a "best guess". This patch enables the driver to store if we want to enable/disable autoneg FC and do the correct behavior. Signed-off-by:
Don Skidmore <donald.c.skidmore@intel.com> Signed-off-by:
-
Mallikarjuna R Chilakala authored
When network interface is made active we were not handling the error scenarios properly to clean up rx & tx resources which might result in a driver panic. Signed-off-by:
-
Alexander Duyck authored
This patch resolves an issue with map single being used to map a buffer and then unmap page being used to unmap it. In addition it handles any error conditions that may be detected using skb_dma_map. Signed-off-by:
Alexander Duyck <alexander.h.duyck@intel.com> Acked-by:
-
PJ Waskiewicz authored
MSI-X allocation broke after the 82599 merge on systems with more than 8 CPU cores. 82598 drops back into MSI mode, which isn't sufficient to run full, efficient 10G line rate. Signed-off-by:
-
Don Skidmore authored
In the past flow control wasn't enabled by default under the incorrect assumption that this opened up us to a denial of service attack. However since any switch that forwarded flow control would be extremely msiconfigured and/or buggy, this concern no longer out weighs the preformance gains from having FC enabled. Signed-off-by:
-
PJ Waskiewicz authored
The priority flow control settings from the netlink layer aren't taking effect in the base driver. The boolean pfc_mode_enable in the dcb_config struct isn't being set, so the hardware configuration code is never reached. Signed-off-by:
-
Don Skidmore authored
Ethtool tries to get advertised speed from phy.autoneg_advertised. However for copper media this wasn't happening until later do to an other fix which moved mac.ops.setup_link_speed placement in ixgbe_link_config(). This patch will display the default advertised speeds if it can't yet get this information from phy.autoneg_advertised. Signed-off-by:
-
Alexander Duyck authored
The ixgbe driver had issues when DEBUG was defined because the hw_dbg macro was incomplete. This patch completes the code based off of the code that already existed in the igb module. Signed-off-by:
-
Rafael J. Wysocki authored
Impact: Fix Yinghai Lu found one system with 82575EB where, in the kernel that is kexeced, probe igb failed with -2, the reason being that the adapter could not be brought back from D3 by the kexec kernel, most probably due to quirky hardware (it looks like the same behavior happened on forcedeth). Prevent igb from putting the adapter into D3 during shutdown except when we going to power off the system. For this purpose, seperate igb_shutdown() from igb_suspend() and use the appropriate PCI PM callbacks in both of them. Signed-off-by:
"Rafael J. Wysocki" <rjw@sisk.pl> Reported-by:
Yinghai Lu <yinghai@kernel.org> Signed-off-by:
-
Alexander Duyck authored
The code path for setting up phy loopback testing was out of date and was setting bits it didn't need to. This change cleans up the code path and removes some code that has no effect on teh driver. Signed-off-by:
-
Alexander Duyck authored
Increase the delay for copper phy init from 15ms to 100ms. This is to address issues seen in which ethtool -t was failing in some cases on 82576 based adapters. Signed-off-by:
-
Alexander Duyck authored
An issue was seen on suspend in which the system reported a page fault. This was due to the new reg_idx code being called after the queues were freed. This update prevents any for loops from going through the queues by setting the number of queues to 0 when they are freed. Signed-off-by:
-
Alexander Duyck authored
When device is on PCIe link trained as x2 the driver is currently reporting link width as "unknown". The original patch provided by Myron adds the x2 link support and my changes are cosmetic to clean up the readability of the conditional operators. Based on work by: Myron Stowe <myron.stowe@hp.com> Signed-off-by:
-
Grant Likely authored
The FEC Ethernet device isn't always attached to a phy. Be careful not to dereference phy_device if it is NULL. Also eliminates an unnecessary extra function from the ioctl path. Reported-by:
Henk Stegeman <henk.stegeman@gmail.com> Signed-off-by:
Grant Likely <grant.likely@secretlab.ca> Signed-off-by:
-
Henk Stegeman authored
Since not using net_device_ops gets you shunned out the cool crowd, this patch modifies the fec_mpc52xx Ethernet driver to provide the management hooks via a struct net_device_ops. Reported-by:
-
Grant Likely authored
The driver triggers a BUG_ON() when allocating DMA buffers because the arch/powerpc dma_ops aren't in the net_device's struct device. This patch fixes the problem by using the parent of_device which does have the correct dma_ops set. Signed-off-by:
Becky Bruce <beckyb@kernel.crashing.org> Signed-off-by:
-
Eric Dumazet authored
Commit 78454473 (netfilter: iptables: lock free counters) forgot to disable BH in arpt_do_table(), ipt_do_table() and ip6t_do_table() Use rcu_read_lock_bh() instead of rcu_read_lock() cures the problem. Reported-and-bisected-by:
Roman Mindalev <r000n@r000n.net> Signed-off-by:
Eric Dumazet <dada1@cosmosbay.com> Acked-by:
Patrick McHardy <kaber@trash.net> Acked-by:
-
Andy Grover authored
We have a 64bit value that needs to be set atomically. This is easy and quick on all 64bit archs, and can also be done on x86/32 with set_64bit() (uses cmpxchg8b). However other 32b archs don't have this. I actually changed this to the current state in preparation for mainline because the old way (using a spinlock on 32b) resulted in unsightly #ifdefs in the code. But obviously, being correct takes precedence. Signed-off-by:
Andy Grover <andy.grover@oracle.com> Signed-off-by:
-
Andy Grover authored
This fixes a bug where a connection was unexpectedly not on *any* list while being destroyed. It also cleans up some code duplication and regularizes some function names. * Grab appropriate lock in conn_free() and explain in comment * Ensure via locking that a conn is never not on either a dev's list or the nodev list * Add rds_xx_remove_conn() to match rds_xx_add_conn() * Make rds_xx_add_conn() return void * Rename remove_{,nodev_}conns() to destroy_{,nodev_}conns() and unify their implementation in a helper function * Document lock ordering as nodev conn_lock before dev_conn_lock Reported-by:Yosef Etigin <yosefe@voltaire.com> Signed-off-by:
-
Andy Grover authored
rs_send_drop_to() is called during socket close. If it takes m_rs_lock without disabling interrupts, then rds_send_remove_from_sock() can run from the rx completion handler and thus deadlock. Signed-off-by:
-
