This patch adds a platform device driver that supports the OpenCores 10/100 Mbps Ethernet MAC. The driver expects three resources: one IORESOURCE_MEM resource defines the memory region for the core's memory-mapped registers while a second IORESOURCE_MEM resource defines the network packet buffer space. The third resource, of type IORESOURCE_IRQ, associates an interrupt with the driver. Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de> Acked-by: Florian Fainelli ...

The line: if ((bd == ugeth->txBd[txQ]) && (netif_queue_stopped(dev) == 0)) break; in ucc_geth_tx() didn not make sense to me. Rework & cleanup this logic to something understandable. --- Reworked the patch according to Antons comments. drivers/net/ucc_geth.c | 66 +++++++++++++++++++++++++++-------------------- 1 files changed, 38 insertions(+), 28 deletions(-) diff --git a/drivers/net/ucc_geth.c b/drivers/net/ucc_geth.c index 7fc91aa..465de3a 100644 --- ...

Also set NAPI weight to 64 as this is a common value. This will make the system alot more responsive while ping flooding the ucc_geth ethernet interaface. Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se> --- drivers/net/ucc_geth.c | 32 ++++++++++++-------------------- drivers/net/ucc_geth.h | 1 - 2 files changed, 12 insertions(+), 21 deletions(-) diff --git a/drivers/net/ucc_geth.c b/drivers/net/ucc_geth.c index 097aed8..7fc91aa 100644 --- ...

Yes, I know. I did it this way because I something broke under stress when ugeth->confBd[txQ] instead. The ucc_geth_tx() and ucc_geth_start_xmit() Sure, done. --

Your wish is my command. DMTF SMBIOS 2.6 specification http://www.dmtf.org/standards/smbios/ contains changes which provide this for PCI devices. Specifically, Type 9 ("System Slots") was extended to include the PCI domain/bus/device/function for each slot. Type 10 ("On Board Devices Information") could not be extended, thus it was deprecated, and new Type 41 ("Onboard Devices Extended Information") was created to be extensible and now includes PCI domain/bus/device/function information. ...

In addition to seeing the problem with CONFIG_FORCEDETH_NAPI disabled, I have now reproduced the problem with that configuration option enabled. So, NAPI might not be the problem at all. Adam Richter --

OK. I tried your patch. It seems fine. I was able to do "ping -f" for five minutes with no problems, and surf the web during that time. I preserving the rest of your message below, just to be clear about which patch I tested. --

Its past 2.2 ... Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk?> --- net/appletalk/ddp.c | 6 +----- 1 files changed, 1 insertions(+), 5 deletions(-) diff --git a/net/appletalk/ddp.c b/net/appletalk/ddp.c index 5abce07..3104be8 100644 --- a/net/appletalk/ddp.c +++ b/net/appletalk/ddp.c @@ -1571,14 +1571,10 @@ static int atalk_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr usat->sat_family != AF_APPLETALK) return -EINVAL; - /* netatalk doesn't ...

That isn't what I mean. The code he is referring to exists nowhere in the igbvf driver. I suppose I can edit the igbvf commit comments so that they don't mention the sysfs entry, but the code is in the igb driver. Also I just want to clarify. It isn't fair to compare igbvf to macvlan. A better comparison would be virtio since it is meant to run on a guest, not on the hypervisor/DOM0 OS. I have also been looking all over for an example that is even close to what we are looking for and ...

Spaces around "+"... Also, now it's obvious that we're just reading status or buf. So instead of these inlines we could use struct qe_bd as described in asm/qe.h. I.e. struct qe_bd *bd = ...; in_be32(&bd->buf); in_be16(&bd->status); Oh, wait. We can't, ucc_geth assumes status is u32, which includes the length field, i.e. ucc_fast.h defines: The cleanup work surely desires a separate patch, so bd2buf and bd2status are OK, for now. I'll test the patch as soon as I'll get ...

On Thu, Mar 26, 2009 at 05:43:25PM +0100, Joakim Tjernlund wrote: Would be great if you could investigate it more. Maybe there is a serious bug somewhere, or maybe you're introducing another (yet hidden) bug... -- Anton Vorontsov email: cbouatmailru@gmail.com irc://irc.freenode.net/bd2 --

xfrm: spin_lock() should be spin_unlock() in xfrm_state.c spin_lock() should be spin_unlock() in xfrm_state_walk_done(). caused by: commit 12a169e7d8f4b1c95252d8b04ed0f1033ed7cfe2 "ipsec: Put dumpers on the dump list" Reported-by: Marc Milgram <mmilgram@redhat.com> Signed-off-by: Chuck Ebbert <cebbert@redhat.com> --- --- linux-2.6.29.noarch.orig/net/xfrm/xfrm_state.c +++ linux-2.6.29.noarch/net/xfrm/xfrm_state.c @@ -1615,7 +1615,7 @@ void xfrm_state_walk_done(struct xfrm_st ...

I dont have NIU hardware but it seems this driver could suffer from starvation of high numbered RX queues under flood. I suspect other multiqueue drivers might have same problem. With a standard budget of 64, we can consume all credit when handling first queue(s), and last queues might discard packets. Solve this by rotating the starting point, so that we garantee to scan at least one new queue at each niu_poll_core() invocation, even under stress. Also, change logic calling ...

From: Adam Richter <adam_richter2004@yahoo.com> Adam, first of all, you don't need to subscribe then unsubscribe from this mailing list just to post your report. I just saw you do that. We're not nazis and you can post to this mailing list without being a member. :-) Second of all, please look at my reply to your original report, there is a patch there for you to test. --

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Applied. --

From: Eric Leblond <eric@inl.fr> Applied. --

From: Jiri Pirko <jpirko@redhat.com> Looks good, applied, thanks! --

From: Florian Fainelli <florian@openwrt.org> Applied. --

From: Adam Richter <adam_richter2004@yahoo.com> We're pretty sure we know exactly what commit causes this. Can you try playing with a patch Jarek P. just posted in the thread where this bug is being discussed? (Subject: Revert "gro: Fix legacy path napi_complete crash"): diff --git a/net/core/dev.c b/net/core/dev.c index e3fe5c7..cf53c24 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -2589,7 +2589,11 @@ static int process_backlog(struct napi_struct *napi, int quota) skb = ...

From: Adam Richter <adam_richter2004@yahoo.com> Thanks for testing. --

From: Eric Leblond <eric@inl.fr> Applied. --

From: Florian Fainelli <florian@openwrt.org> Also applied, thanks. --

From: Roland Dreier <rdreier@cisco.com> Looks good, Linus please apply: --

From: Stephen Hemminger <shemminger@vyatta.com> Applied, thanks. --

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Applied. --

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Applied. --

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Applied. --

From: David Miller <davem@davemloft.net> Actually the problem is that you're using deprecated interfaces which have been removed in net-next-2.6, please use napi_*() instead of netif_rx_*().

From: Thierry Reding <thierry.reding@avionic-design.de> This doesn't build: drivers/net/ethoc.c: In function ‘ethoc_interrupt’: drivers/net/ethoc.c:510: error: implicit declaration of function ‘netif_rx_schedule_prep’ drivers/net/ethoc.c:511: error: implicit declaration of function ‘__netif_rx_schedule’ drivers/net/ethoc.c: In function ‘ethoc_poll’: drivers/net/ethoc.c:550: error: implicit declaration of function ‘netif_rx_complete’ You are missing some header includes.

From: Yu Zhao <yu.zhao@intel.com> We have a generic link creation mechanism in rtnl_link so I think this claim wrt. SR-IOV is false. Believe it or not, somebody took the time to make a generic mechanism for things many virtual network device drivers have to do, and igb/igbvf can use it too. --

nm-applet could support some sort of "named" adapters, though I'd rather have this done with udev rules (or something like that) so that the NIC's common name would be consistent in both the CLI and in the GUI. The only reason nm-applet does what it does now (pulling VID/PID and dropping stupid words like "Corporation") is so the user has *some* clue what NIC they are about to touch; using "eth0" and "eth1" and "eth2" isn't very helpful. But the distinction between "Intel Gigabit Ethernet" ...

This patch fixes the declaration of the logger structure in ebt_log and ebt_ulog: I forgot to remove the const option from their declaration in the commit ca735b3aaa945626ba65a3e51145bfe4ecd9e222 ("netfilter: use a linked list of loggers"). Pointed-out-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Eric Leblond <eric@inl.fr> --- net/bridge/netfilter/ebt_log.c | 2 +- net/bridge/netfilter/ebt_ulog.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git ...

Hi Stephen, og_unregister' discards qualifiers from pointer target type Thanks a lot for pointing this. I stupidely forgot to build this module during my testing. I've made the necessary modifications and a patch fixing this will follow this mail. Doing some testing of the ebt_ulog module, I've found some problems. One of them was the following messages: sys_init_module: 'ebt_ulog'->init suspiciously returned 1, it should follow 0/-E convention sys_init_module: loading module ...

ok, I'm almost sending them (last compile test on every patch, please give me a few minutes) Cyrill --

That could be awkward. What if the second or subsequent instance I was thinking of the sender keeping hold of the sent skb and Could you give the main usage scenarios for timestamps, including what the user-space app does with the info? Thanks, -- Visit Openswan at http://www.openswan.org/ Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt --

Hi Adam: Any chance you can test this patch instead of the previous one? net: Fix netpoll lockup in legacy receive path When I fixed the GRO crash in the legacy receive path I used napi_complete to replace __napi_complete. Unfortunately they're not the same when NETPOLL is enabled, which may result in us not calling __napi_complete at all. What's more, we really do need to keep the __napi_complete call within the IRQ-off section since in theory an IRQ can occur in between and fill up ...

From: Jesse Brandeburg <jesse.brandeburg@intel.com> As noticed by Alan Cox, it is possible for e1000e to exit its interrupt handler or NAPI with interrupts enabled even when the driver is unloading or being configured administratively down. fix related to fix for: http://bugzilla.kernel.org/show_bug.cgi?id=12876 Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com> CC: Alan Cox <alan@lxorguk.ukuu.org.uk> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> --- ...

From: Jesse Brandeburg <jesse.brandeburg@intel.com> e1000e (and e1000, 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 ...

From: Alexander Duyck <alexander.h.duyck@intel.com> The tx cleanup routine was stopping after 64 packets and this was causing issues resulting in the ring not being completely cleaned. This change updates the driver to clean the entire ring and if it doesn't it then will retry on the next pass. Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> --- drivers/net/e1000/e1000_main.c | 21 +++++++++------------ 1 files ...

From: Jesse Brandeburg <jesse.brandeburg@intel.com> Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> --- drivers/net/e1000e/netdev.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c index b4ae046..bfb2d6c 100644 --- a/drivers/net/e1000e/netdev.c +++ b/drivers/net/e1000e/netdev.c @@ -48,7 +48,7 @@ #include "e1000.h" -#define ...

Ok. I was just so happy .. (See "Status update on Sun Neptune 10Gbit driver earlier). But then it "blew up" again: Mar 26 13:25:49 hest kernel: [25335.505049] ------------[ cut here ]------------ Mar 26 13:25:49 hest kernel: [25335.505055] WARNING: at net/sched/sch_generic.c:226 dev_watchdog+0x1fd/0x210() Mar 26 13:25:49 hest kernel: [25335.505057] Hardware name: Sun Fire X4600 M2 Mar 26 13:25:49 hest kernel: [25335.505059] NETDEV WATCHDOG: eth4 (niu): transmit timed out Mar 26 ...

Don't worry :) I did not forget about this - just needed a bit time to investigate... Yeah, this look's like a problem. In __vlan_hwaccel_rx there is following line: skb->dev = vlan_group_get_device(grp, vlan_tci & VLAN_VID_MASK); This rewrites the dev so latter on when netif_receive_skb is called the hook will be not called (because dev->master will not be set). Ok I will put the hook inside the skb_bond_should_drop() - it seems like a correct solution... --

commit e487eb99cf9381a4f8254fa01747a85818da612b Author: Holger Eitzenberger <holger@eitzenberger.org> Date: Wed Mar 25 18:26:30 2009 +0100 netlink: add nla_policy_len() It calculates the max. length of a Netlink policy, which is usefull for allocating Netlink buffers roughly the size of the actual message. Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org> Signed-off-by: Patrick McHardy <kaber@trash.net> diff --git a/include/net/netlink.h ...

commit 5c0de29d06318ec8f6e3ba0d17d62529dbbdc1e8 Author: Holger Eitzenberger <holger@eitzenberger.org> Date: Wed Mar 25 21:52:17 2009 +0100 netfilter: nf_conntrack: add generic function to get len of generic policy Usefull for all protocols which do not add additional data, such as GRE or UDPlite. Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org> Signed-off-by: Patrick McHardy <kaber@trash.net> diff --git ...

commit d271e8bd8c60ce059ee36d836ba063cfc61c3e21 Author: Holger Eitzenberger <holger@eitzenberger.org> Date: Thu Mar 26 13:37:14 2009 +0100 ctnetlink: compute generic part of event more acurately On a box with most of the optional Netfilter switches turned off some of the NLAs are never send, e. g. secmark, mark or the conntrack byte/packet counters. As a worst case scenario this may possibly still lead to ctnetlink skbs being reallocated in netlink_trim() ...

commit a400c30edb1958ceb53c4b8ce78989189b36df47 Author: Holger Eitzenberger <holger@eitzenberger.org> Date: Wed Mar 25 21:53:39 2009 +0100 netfilter: nf_conntrack: calculate per-protocol nlattr size Signed-off-by: Holger Eitzenberger <holger@eitzenberger.org> Signed-off-by: Patrick McHardy <kaber@trash.net> diff --git a/net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c b/net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c index 8b681f2..7d2ead7 100644 --- ...

Hi Dave, following are a few late netfilter patches and fixes for 2.6.30, containing: - Eric's patch to use SLAB_DESTROY_BY_RCU in conntrack, which reduces the conntrack size and avoids temporarily exceeding the configured maximum amount of entries before the RCU threshold kicks in. - another patch from Eric to factorize the optimized ifname comparisons - a fix from Eric to use hlist_add_head_rcu in nf_conntrack_set_hashsize() to avoid a race condition - a number of patches from ...

commit d0dba7255b541f1651a88e75ebdb20dd45509c2f Author: Holger Eitzenberger <holger@eitzenberger.org> Date: Wed Mar 25 18:24:48 2009 +0100 netfilter: ctnetlink: add callbacks to the per-proto nlattrs There is added a single callback for the l3 proto helper. The two callbacks for the l4 protos are necessary because of the general structure of a ctnetlink event, which is in short: CTA_TUPLE_ORIG <l3/l4-proto-attributes> CTA_TUPLE_REPLY ...

commit b8dfe498775de912116f275680ddb57c8799d9ef Author: Eric Dumazet <dada1@cosmosbay.com> Date: Wed Mar 25 17:31:52 2009 +0100 netfilter: factorize ifname_compare() We use same not trivial helper function in four places. We can factorize it. Signed-off-by: Eric Dumazet <dada1@cosmosbay.com> Signed-off-by: Patrick McHardy <kaber@trash.net> diff --git a/include/linux/netfilter/x_tables.h b/include/linux/netfilter/x_tables.h index e8e08d0..72918b7 100644 --- ...

commit 1f9352ae2253a97b07b34dcf16ffa3b4ca12c558 Author: Patrick McHardy <kaber@trash.net> Date: Wed Mar 25 19:26:35 2009 +0100 netfilter: {ip,ip6,arp}_tables: fix incorrect loop detection Commit e1b4b9f ([NETFILTER]: {ip,ip6,arp}_tables: fix exponential worst-case search for loops) introduced a regression in the loop detection algorithm, causing sporadic incorrectly detected loops. When a chain has already been visited during the check, it is treated as ...

I guess this isn't going to get any prettier :) Could you please send me the latest version of the missing patches against the current nf-next tree? --

So what are these missing features in the existing interfaces? --

Hi David, I know that I should ask to the ath5k developers but.. when we'll have a fully working ath5k driver for AP mode? I think this would a very nice feature to have and quite urgent to definitely migrate from the old madwifi drivers. --

The sender must be prepared for packet loss anyway (for example, during stress tests I have seen that packets were dropped between socket layer and device driver). Getting a TX time stamp only for some, but not all expected interfaces, or getting some results much later isn't that different. After a certain timeout the sender must assume that the packet was lost and resend. An unexpected response for a packet that was supposed to be The sender is in user space, at least in the case of ...

I understand your point of view. In any case, at first glance, I understand that the urb->transfer_buffer directly points to the preallocated skb data. These are filled by the HWs. I mean, the buffers are treated by the HW. So I guess, the meaning of the rx_fixup functions is just to solve this kind of situations. In fact, each usb net driver has an own fixup code according to their HW specifications. Peppe --

ok, I have reverted the GRO and I have not got the problem. Thanks. --

on i386 build: drivers/built-in.o: In function `be_worker': be_main.c:(.text+0x827c7): undefined reference to `__udivdi3' be_main.c:(.text+0x8286c): undefined reference to `__udivdi3' Probably one of the called functions (that contains division) is inlined. -- ~Randy --

> > Since the issue isn't the igbvf driver there is no reason for it to > > be held up. > > I disagree, I think both cases should be fixed. > > Just because we do something already never means that it's > ok to proliferate the mistake further. The igbvf driver doesn't proliferate any mistake. The num_vfs sysfs attribute that everyone finds so objectionable is in the igb driver, and if you follow the patchwork URL that was provided, you would see that you applied the patch adding it ...

> Btw, as far as I can tell, that function had never been tested with any > polling. It had a line like > > netif_napi_add(netdev, &nesvnic->napi, nes_netdev_poll, 128); > > but that line was _before_ 'nesvnic' was actually initialized (well, it > had been initialized to NULL, so it passed in some NULL-pointer-offset to > that function). Yes, looks like the bug was introduced by d0929553 ("infiniband: convert nes driver to net_device_ops"). And unfortunately the ...

Hi Eric, Thanks. --=20 Cheers, Stephen Rothwell sfr@canb.auug.org.au http://www.canb.auug.org.au/~sfr/

The code that was added to increase headroom was wrong. It doesn't handle the case where gfar_add_fcb() changes the skb. Better to do check at start of transmit (outside of lock), where error handling is better anyway. Signed-off-by: Stephen Hemminger <shemminger@vyatta.com> --- a/drivers/net/gianfar.c 2009-03-26 09:14:39.273669929 -0700 +++ b/drivers/net/gianfar.c 2009-03-26 09:22:46.477545004 -0700 @@ -1239,19 +1239,9 @@ static int gfar_enet_open(struct net_dev return err; } ...

The benet driver is doing a 64 bit divide, which is not supported in Linux kernel on 32 bit architectures. The correct way to do this is to use do_div(). Compile tested on i386 only. Signed-off-by: Stephen Hemminger <shemminger@vyatta.com> --- a/drivers/net/benet/be_main.c 2009-03-25 15:31:35.000000000 -0700 +++ b/drivers/net/benet/be_main.c 2009-03-26 12:28:21.645388379 -0700 @@ -16,6 +16,7 @@ */ #include "be.h" +#include <asm/div64.h> MODULE_VERSION(DRV_VER); ...

Oh, okay. I've only been building against the vanilla tree. I'll grab a copy of net-next-2.6 and fix the driver up against that. Thanks, Thierry --

Btw, as far as I can tell, that function had never been tested with any polling. It had a line like netif_napi_add(netdev, &nesvnic->napi, nes_netdev_poll, 128); but that line was _before_ 'nesvnic' was actually initialized (well, it had been initialized to NULL, so it passed in some NULL-pointer-offset to that function). I moved that initialization up in the merge, but I obviously didn't test anything. So I'd suggest somebody actually test this case, because it clearly must ...

Can I get a sign-off from Roland too? Linus --

The default QP number in RSS is no longer part of the RSS table. We want to separate TCP streams from other streams because we want to handle them differently. Changed the RSS mask to perform the hashing on TCP packets only and to send all the others to the default ring. Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il> --- drivers/net/mlx4/en_main.c | 2 +- drivers/net/mlx4/en_params.c | 2 +- drivers/net/mlx4/en_rx.c | 4 ++-- drivers/net/mlx4/mlx4_en.h | 4 ...

Allowing using linear skbs instead skb frags for certain rings. Working with linear skbs shows better performance when working with non TCP traffic (LRO shows better performance with skb frags). Rss allows us to separate TCP from non TCP traffic and send the packets to different rings. Then, every ring that receives TCP traffic is assigned to work with skb frags, while the ring that receives non TCP traffic works with linear skbs. This change improves RX Bandwidth for small UDP messages (under ...

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il> --- drivers/net/mlx4/en_resources.c | 6 ++++++ drivers/net/mlx4/en_rx.c | 6 ------ drivers/net/mlx4/en_tx.c | 1 + drivers/net/mlx4/mlx4_en.h | 1 + 4 files changed, 8 insertions(+), 6 deletions(-) diff --git a/drivers/net/mlx4/en_resources.c b/drivers/net/mlx4/en_resources.c index a054520..65ca706 100644 --- a/drivers/net/mlx4/en_resources.c +++ b/drivers/net/mlx4/en_resources.c @@ -94,3 +94,9 ...

We do have some powerful tools for network device configuration, but they still can't cover the multi queue and SR-IOV NICs (also David Woodhouse's Solos DSL, etc.) configuration requirement. Unless someone improves the existing tools to support configuring these new hardware features, people would keep trying to use the sysfs which is the easiest way so far. Thanks, Yu --

Hi. Passive OS fingerprinting netfilter module allows to passively detect remote OS and perform various netfilter actions based on that knowledge. This module compares some data (WS, MSS, options and it's order, ttl, df and others) from packets with SYN bit set with dynamically loaded OS fingerprints. Fingerprint matching rules can be downloaded from OpenBSD source tree and loaded via netfilter netlink subsystem into the kernel via special util found in archive. Archive contains library ...