* Re: [net-next PATCH v3 00/12] Flow API
From: Thomas Graf @ 2015-01-24 13:34 UTC (permalink / raw)
To: Jamal Hadi Salim
Cc: John Fastabend, Jiri Pirko, Pablo Neira Ayuso, simon.horman,
sfeldma, netdev, davem, gerlitz.or, andy, ast
In-Reply-To: <54C39C8E.4@mojatatu.com>
On 01/24/15 at 08:22am, Jamal Hadi Salim wrote:
> It is up to user space to decide on what the policy should do.
> The kernel is not paid to think. You tell it what to do and it does it
> efficiently. So if you are going to tell it to have a mix and match
> of some things to execute in hardware and some in software then
> it may shoot someone's big toe.
OK. We seem agree on this part. In order to do so, user space needs
to know about hardware capabilities. If that should happen through
tc, so be it. John raised some open question around this and the
rtnl lock is currently a blocker on this architecture as well.
> IOW, user space should decide how a packet is going to flow.
> Agreed that we would need a good way to provide this knowledge
> to user space.
> BTW: Thomas, reading your other email quickly:
> the idea that metadata would be carried around on OF pipeline and
> some script at the end executes the actions is imo a hardware
> pipeline hack limitation. Why do i want to defer dropping a packet
> when some action is telling me to drop it? ;->
There is obviously no reason to defer a drop.
An example of deferred actions would be if only certain tables allow
certain actions but the matching to chose the action is done in a
previous tables. Or if you have multiple tables matching on the
original packet header and you need to defer the L2/L3 rewrite until
all matching and action construction is done.
> For some reason, brcm hardware in particulat requires that i
> complete the pipeline first.
> I dont know why we need such a limitation in s/ware (and tc will kill
> the pipeline when needed).
Not sure what "killing the pipeline" means ;-)
^ permalink raw reply
* Re: [net-next PATCH v3 00/12] Flow API
From: Jamal Hadi Salim @ 2015-01-24 13:22 UTC (permalink / raw)
To: Thomas Graf, John Fastabend, Jiri Pirko
Cc: Pablo Neira Ayuso, simon.horman, sfeldma, netdev, davem,
gerlitz.or, andy, ast
In-Reply-To: <20150123174609.GA23242@casper.infradead.org>
On 01/23/15 12:46, Thomas Graf wrote:
> I'm not sure I understand the offload qdisc yet. My interpretation
> so far is that it would contain childs which *must* be offloaded.
>
> How would one transparently offload tc in this model? e.g. let's
> assume we have a simple prio qdisc with u32 cls:
>
> eth0
> prio
> class
> class
> ...
> u32 ...
> u32 ...
>
My view is:
It is up to user space to decide on what the policy should do.
The kernel is not paid to think. You tell it what to do and it does it
efficiently. So if you are going to tell it to have a mix and match
of some things to execute in hardware and some in software then
it may shoot someone's big toe.
IOW, user space should decide how a packet is going to flow.
Agreed that we would need a good way to provide this knowledge
to user space.
BTW: Thomas, reading your other email quickly:
the idea that metadata would be carried around on OF pipeline and
some script at the end executes the actions is imo a hardware
pipeline hack limitation. Why do i want to defer dropping a packet
when some action is telling me to drop it? ;->
For some reason, brcm hardware in particulat requires that i
complete the pipeline first.
I dont know why we need such a limitation in s/ware (and tc will kill
the pipeline when needed).
Sorry, trying to post while doing other things so not paying close
attention to possibly other important details.
cheers,
jamal
^ permalink raw reply
* Re: [net-next PATCH v3 00/12] Flow API
From: Jamal Hadi Salim @ 2015-01-24 13:04 UTC (permalink / raw)
To: Jiri Pirko, Thomas Graf
Cc: John Fastabend, Pablo Neira Ayuso, simon.horman, sfeldma, netdev,
davem, gerlitz.or, andy, ast
In-Reply-To: <20150123161605.GN2065@nanopsycho.orion>
I am not big on high fives but here's my +1
Excellent diagram below.
cheers,
jamal
On 01/23/15 11:16, Jiri Pirko wrote:
>
> As I wrote earlier, the value is that userspace can easily use single
> xflows api to take care of all ways to handle flows (ovs kernel dp,
> rocker, other device, u32 tc filter + actions, you name it)
>
>
> my flow managing app
> |
> uspc |
> --------|----------------------------------------------------
> krnl |
> tc xflows api
> | | |
> | | ---------------------------------------------------
> | | |
> | ------------------ other xflows backend
> | |
> ovs xflows backend rocker driver xflows backend
> | |
> ovs dp |
> krnl |
> ----------------------------|--------------------------------
> hw |
> rocker switch
>
^ permalink raw reply
* Re: [net-next PATCH v3 00/12] Flow API
From: Jamal Hadi Salim @ 2015-01-24 13:01 UTC (permalink / raw)
To: John Fastabend, Thomas Graf
Cc: Jiri Pirko, Pablo Neira Ayuso, simon.horman, sfeldma, netdev,
davem, gerlitz.or, andy, ast
In-Reply-To: <54C26A1F.6060603@gmail.com>
On 01/23/15 10:34, John Fastabend wrote:
> First 'tc' infrastructure doesn't have any classifier that would map
> well to this today so you are talking about a new classifier looks like
> Jiri is calling it xflows. This is fine.
I know you know this (and apologies for the little Australian Bike
Shed tangent):
You can do _any_ classifier you want. xflows just happens to make the
OF people happy. Someone else who wants to classify on pcre like
strings can go ahead and write another one.
i.e there is no monopoly on what a classifier should be.
> Now 'xflows' needs to implement the same get operations that exist in
> this flow API otherwise writing meaningful policies as Thomas points out
> is crude at best.
It is crude only if you assume the kernel is doing your policies
and fixing any conflicts. Let the kernel do mechanisms and have user
space do the brainy part. No need to give total autonomy to the kernel.
> So this tc classifier supports 'get headers',
> 'get actions', and 'get tables' and then there associated graphs. All
> good so far. This is just an embedding of the existing API in the 'tc'
> netlink family. I've never had any issues with this. Finally you build
> up the 'get_flow' and 'set_flow' operations I still so no issue with
> this and its just an embedding of the existing API into a 'tc
> classifier'. My flow tool becomes one of the classifier tools.
>
You have very few generic verbs really within tc and i dont see
much more needed.
GET/SET(mods for create/append/replace)/DEL with the object
being a noun. Add a handful for capabilities exercising verbs and
you should be on your way.
BTW: I did have capabilities in actions for years but Cong sent a
patch about a year or so ago to kill them because they were not being
exercised from user space tc - I protested but Dave overruled me.
There are still remnants - look at struct tcf_common field
tcfc_capab - the original intent was to have that look like netdev
features bitmask. In any case i never got to proper implementation
and have gained a lot of experience since those early days
and my thinking has changed.
> Now what should I attach my filter to? Typically we attach it to qdiscs
> today. But what does that mean for a switch device? I guess I need an
> _offloaded qdisc_? I don't want to run the same qdisc in my dataplane
> of the switch as I run on the ports going into/out of the sw dataplane.
I dont know if you need a qdisc necessarily that sits in hardware.
But you need to anchor your policy somewhere. The ingress qdisc is
really a dummy for this purpose. It is the beggining of the pipeline.
Most of the hardware i have looked at has some anchor point for the
hardware ACLs. Typically around a queue or a port. Sometime i find it
hard to use this model because of vendor SDKs and APIs they offer.
> Similarly I don't want to run the same set of filters. So at this point
> I have a set of qdiscs per port to represent the switch dataplane and
> a set of qdiscs attached to the software dataplane. If people think this
> is worth doing lets do it. It may get you a nice way to manage QOS while
> your @ it.
>
Lets discuss at the meeting. I am just skimming these emails (the
conference is chewing a lot of my time so i will mostly be absent).
Sorry if i am not responding to some things.
cheers,
jamal
>
>
^ permalink raw reply
* Re: [net-next PATCH v3 00/12] Flow API
From: Jamal Hadi Salim @ 2015-01-24 12:36 UTC (permalink / raw)
To: Jiri Pirko, Thomas Graf
Cc: Pablo Neira Ayuso, John Fastabend, simon.horman, sfeldma, netdev,
davem, gerlitz.or, andy, ast
In-Reply-To: <20150123113934.GD2065@nanopsycho.orion>
On 01/23/15 06:39, Jiri Pirko wrote:
> Maybe I did not express myself correctly. I do not care if this is
> exposed by rtnl or a separate genetlink. The issue still stands. And the
> issue is that the user have to use "the way A" to setup sw datapath and
> "the way B" to setup hw datapath. The preferable would be to have
> "the way X" which can be used to setup both sw and hw.
>
> And I believe that could be achieved. Consider something like this:
>
> - have cls_xflows tc classifier and act_xflows tc action as a wrapper
> (or api) for John's work. With possibility for multiple backends. The
> backend iface would looke very similar to what John has now.
> - other tc clses and acts will implement xflows backend
> - openvswitch datapath will implement xflows backend
> - rocker switch will implement xflows backend
> - other drivers will implement xflows backend
>
> Now if user wants to manipulate with any flow setting, he can just use
> cls_xflows and act_xflows to to that.
>
> This is very rough, but I just wanted to draw the picture. This would
> provide single entry to flow world manipulation in kernel, no matter if
> sw or hw.
>
> Thoughts?
>
Exactly my thinking as well.
I guess skipping a few emails helps.
cheers,
jamal
^ permalink raw reply
* Re: [net-next PATCH v3 00/12] Flow API
From: Jamal Hadi Salim @ 2015-01-24 12:34 UTC (permalink / raw)
To: Thomas Graf, Pablo Neira Ayuso
Cc: John Fastabend, simon.horman, sfeldma, netdev, davem, gerlitz.or,
andy, ast, Jiri Pirko
In-Reply-To: <20150122174445.GD25797@casper.infradead.org>
On 01/22/15 12:44, Thomas Graf wrote:
> On 01/22/15 at 05:49pm, Pablo Neira Ayuso wrote:
>
> You can achieve the exact same thing with an out of tree tc action,
> classifier or even a new link type. Nothing prevents an out of tree
> driver to register a new rtnetlink link type and do vendor specific
> crap.
>
They are not the same. The API lowers the barrier immensely; refer to
my response to John. And we are actually handing it to them.
From an equivalance perspective:
Youd have to convince Dave to allow all those TOE vendors to get in
with their direct hardware APIs (if those guys are still around).
cheers,
jamal
^ permalink raw reply
* Re: [net-next PATCH v3 00/12] Flow API
From: Jamal Hadi Salim @ 2015-01-24 12:29 UTC (permalink / raw)
To: John Fastabend, Thomas Graf
Cc: Pablo Neira Ayuso, simon.horman, sfeldma, netdev, davem,
gerlitz.or, andy, ast, Jiri Pirko
In-Reply-To: <54C12C1F.706@gmail.com>
Sorry I have been running around like a lunatic chicken so havent
had time to join the fun discussion. I hope we can make progress at
the meeting.
I am going to skim and jump through the emails and comment.
On 01/22/15 11:58, John Fastabend wrote:
> On 01/22/2015 07:13 AM, Thomas Graf wrote:
>> On 01/22/15 at 10:00am, Jamal Hadi Salim wrote:
>
> Correct this is fully exposed to user space, but it is also self
> contained inside the API meaning I can learn when to use it and what it
> does by looking at the other operations tables the table graph and
> supported headers. The assumption I am making that is not in the API
> explicitly yet. Is that actions named "set_field_name" perform the
> set operation on that field. We can and plan to extend the API to make
> this assumption explicit in the API.
>
From what you describe, you are running into a danger of going too low
level such that the interface will end up weighing too much into
flexibility/perfomance and less into usability. If there is one lesson
i learnt from netfilter is usability counts for something. You dont
want another u32 api (otherwise Jiri wouldnt have to write that new
classifier - there is nothing he is doing that cant be done with
u32).
> In this case I can "learn" that I can match on group_id in some tables
> and then use the above action to set the group_id in others.
>
And this discoverability was part of my concern especially when there is
no "stickiness" to the kernel or Linux tooling for that matter by
going direct to hardware. It is a tactical issue more than anything
else. With the approach, you need a little bit of clue of course, you
really dont even care about compiling the kernel. Essentially the
barrier to entry for SDKs is immensely lowered.
SDK joy. I know you were intending to replace ethtool - but you are
replacing it with a turbo engine and we need to look at a much bigger
scope.
cheers,
jamal
^ permalink raw reply
* Re: [PATCH net-next v3 2/5] swdevice: add new api to set and del bridge port attributes
From: Jiri Pirko @ 2015-01-24 11:26 UTC (permalink / raw)
To: roopa
Cc: sfeldma, jhs, bcrl, tgraf, john.fastabend, stephen, vyasevic,
ronen.arad, netdev, davem, shm, gospo
In-Reply-To: <54C2D4C9.5000400@cumulusnetworks.com>
Sat, Jan 24, 2015 at 12:10:01AM CET, roopa@cumulusnetworks.com wrote:
>On 1/23/15, 2:45 PM, roopa wrote:
>>On 1/23/15, 8:06 AM, Jiri Pirko wrote:
>>>Fri, Jan 23, 2015 at 04:58:57PM CET, roopa@cumulusnetworks.com wrote:
>>>>On 1/23/15, 2:41 AM, Jiri Pirko wrote:
>>>>
>>>><snip..>
>>>>>+
>>>>>+/**
>>>>>+ * netdev_switch_port_bridge_dellink - Notify switch device port
>>>>>of bridge
>>>>>+ * attribute delete
>>>>>+ *
>>>>>+ * @dev: port device
>>>>>+ * @nlh: netlink msg with bridge port attributes
>>>>>+ *
>>>>>+ * Notify switch device port of bridge port attribute delete
>>>>>+ */
>>>>>+int netdev_switch_port_bridge_dellink(struct net_device *dev,
>>>>>+ struct nlmsghdr *nlh, u16 flags)
>>>>>+{
>>>>>+ const struct net_device_ops *ops = dev->netdev_ops;
>>>>>+ struct net_device *lower_dev;
>>>>>+ struct list_head *iter;
>>>>>+ int ret = 0, err = 0;
>>>>>+
>>>>>+ if (!(dev->features & NETIF_F_HW_NETFUNC_OFFLOAD))
>>>>>+ return err;
>>>>>+
>>>>>+ if (ops->ndo_bridge_dellink) {
>>>>>+ WARN_ON(!ops->ndo_switch_parent_id_get);
>>>>>+ return ops->ndo_bridge_dellink(dev, nlh, flags);
>>>>>+ }
>>>>>+
>>>>>+ netdev_for_each_lower_dev(dev, lower_dev, iter) {
>>>>>+ err = netdev_switch_port_bridge_dellink(lower_dev, nlh,
>>>>>flags);
>>>>>+ if (err)
>>>>>+ ret = err;
>>>>>+ }
>>>>>+
>>>>>+ return ret;
>>>>>+}
>>>>>+EXPORT_SYMBOL(netdev_switch_port_bridge_dellink);
>>>>>--
>>>>>1.7.10.4
>>>>>
>>>>>Is there any other place, other than bridge code, this functions are
>>>>>suppored to be called from?
>>>>No other place today. Its usually the master that implements
>>>>ndo_bridge_setlink/dellink.
>>>>
>>>>>If not, which I consider likely, it would
>>>>>make more sense to me to:
>>>>>
>>>>>- move netdev_for_each_lower_dev iterations directly to bridge code
>>>>>- let the masters (bond, team, ..) implement ndo_bridge_*link and do
>>>>> the traversing there (can be in a form of pre-prepared default
>>>>> ndo callback (ndo_dflt_netdev_switch_port_bridge_*link)
>>>>But, i am still not understanding why i would modify bond, team and
>>>>other
>>>>slaves
>>>Well, that is the usual way to propagate ndo calls. People are used to
>>>this. It is visible right away in bonding/other code that is propagated
>>>some ndo call to slaves. With your code, that is somehow hidden and only
>>>dependent on NETIF_F_HW_NETFUNC_OFFLOAD flag.
>>>
>>>Note that there are only couple of "master drivers" (for this, most
>>>likely
>>>only bond and team modifications are needed).
>>ndo_bridge_setlink today is only implemented by drivers that implement
>>bridging function.
>>So, having the bond and team driver implement it...seems odd.
Well, it is not odd. It is ok I believe. Same as for example:
.ndo_set_mac_address
.ndo_change_mtu
.ndo_vlan_rx_add_vid
.ndo_vlan_rx_kill_vid
.ndo_poll_controller
.ndo_netpoll_setup
.ndo_netpoll_cleanup
All take care of propagating the ndo to slaves.
>>
>>But if you insist, i am going to do just that.
>A side note, I dont see any reason for ndo_bridge_setlink to be renamed to
>ndo_setlink. Because it seems to take the whole netlink msg anyways. It can
>be used to offload other link attributes besides bridging (vxlan and so on).
>Any thoughts on that ?.
Well, it is PF_BRIDGE so I think that the name is accurate.
>
>
>
>
^ permalink raw reply
* Re: [PATCH v2 1/4] net: stmmac: dwmac-rk: Don't set the regulator voltage for phy from the driver
From: Heiko Stübner @ 2015-01-24 10:17 UTC (permalink / raw)
To: Romain Perier
Cc: peppe.cavallaro-qxv4g6HH51o, netdev-u79uwXL29TY76Z2rM5mHXA,
linux-kernel-u79uwXL29TY76Z2rM5mHXA,
linux-rockchip-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
devicetree-u79uwXL29TY76Z2rM5mHXA,
roger.chen-TNX95d0MmH7DzftRWevZcw
In-Reply-To: <1421737780-1533-2-git-send-email-romain.perier-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Am Dienstag, 20. Januar 2015, 07:09:37 schrieb Romain Perier:
> As these settings can be directly expressed from devicetree for both fixed
> regulators and pmic-integrated regulators, it is more standard to set them
> from dts and let the regulator framework use the right voltage informations
> when it is used in the driver.
>
> Signed-off-by: Romain Perier <romain.perier-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Romain suggested to resend my tags against a real patch of the series, so it
turns up in patchwork, so
This series
Tested-by: Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>
Reviewed-by: Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>
@Dave: as said in the v1 mail, this would be nice to have for 3.20
Heiko
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* NEED AN URGENT NEW YEAR LOAN FOR ANY PURPOSE ? REPLY US FOR MORE INFO?
From: Macdonald Erick @ 2015-01-24 1:18 UTC (permalink / raw)
To: Recipients
---
This email has been checked for viruses by Avast antivirus software.
http://www.avast.com
^ permalink raw reply
* Re: [net-next] net: netcp: remove unused kconfig option and code
From: David Miller @ 2015-01-24 6:21 UTC (permalink / raw)
To: m-karicheri2; +Cc: netdev, linux-kernel
In-Reply-To: <1421781756-11151-1-git-send-email-m-karicheri2@ti.com>
From: Murali Karicheri <m-karicheri2@ti.com>
Date: Tue, 20 Jan 2015 14:22:36 -0500
> Currently CPTS is built into the netcp driver even though there is no
> call out to the CPTS driver. This patch removes the dependency in Kconfig
> and remove cpts.o from the Makefile for NetCP.
>
> Signed-off-by: Murali Karicheri <m-karicheri2@ti.com>
> ---
> drivers/net/ethernet/ti/Kconfig | 2 +-
> drivers/net/ethernet/ti/Makefile | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/ti/Kconfig b/drivers/net/ethernet/ti/Kconfig
> index e11bcfa..824e376 100644
> --- a/drivers/net/ethernet/ti/Kconfig
> +++ b/drivers/net/ethernet/ti/Kconfig
> @@ -73,7 +73,7 @@ config TI_CPSW
> config TI_CPTS
> boolean "TI Common Platform Time Sync (CPTS) Support"
> depends on TI_CPSW
> - depends on TI_CPSW || TI_KEYSTONE_NET
> + depends on TI_CPSW
Just remove the second dependency line, it's redundant because you've
made it identical to the line before it.
^ permalink raw reply
* Re: [PATCH net] amd-xgbe: Use proper Rx flow control register
From: David Miller @ 2015-01-24 6:18 UTC (permalink / raw)
To: thomas.lendacky; +Cc: netdev
In-Reply-To: <20150120182031.20236.62843.stgit@tlendack-t1.amdoffice.net>
From: Tom Lendacky <thomas.lendacky@amd.com>
Date: Tue, 20 Jan 2015 12:20:31 -0600
> Updated hardware documention shows the Rx flow control settings were
> moved from the Rx queue operation mode register to a new Rx queue flow
> control register. The old flow control settings are now reserved areas
> of the Rx queue operation mode register. Update the code to use the new
> register.
>
> Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Applied, thanks TOm.
^ permalink raw reply
* Re: [PATCH net-next 1/5] bonding: keep bond interface carrier off until at least one active member
From: Jay Vosburgh @ 2015-01-24 3:15 UTC (permalink / raw)
To: Jonathan Toppins
Cc: netdev, Scott Feldman, Andy Gospodarek, Veaceslav Falico,
Nikolay Aleksandrov
In-Reply-To: <54C2D39B.7070104@cumulusnetworks.com>
Jonathan Toppins <jtoppins@cumulusnetworks.com> wrote:
>On 1/21/15 2:14 AM, Jay Vosburgh wrote:
>> Jonathan Toppins <jtoppins@cumulusnetworks.com> wrote:
>>
>>> On 1/19/15 4:16 PM, Jay Vosburgh wrote:
>>>> Jonathan Toppins <jtoppins@cumulusnetworks.com> wrote:
>>>>
>>>>> From: Scott Feldman <sfeldma@cumulusnetworks.com>
>>>>>
>>>>> Bonding driver parameter min_links is now used to signal upper-level
>>>>> protocols of bond status. The way it works is if the total number of
>>>>> active members in slaves drops below min_links, the bond link carrier
>>>>> will go down, signaling upper levels that bond is inactive. When active
>>>>> members returns to >= min_links, bond link carrier will go up (RUNNING),
>>>>> and protocols can resume. When bond is carrier down, member ports are
>>>>> in stp fwd state blocked (rather than normal disabled state), so
>>>>> low-level ctrl protocols (LACP) can still get in and be processed by
>>>>> bonding driver.
>>>>
>>>> Presuming that "stp" is Spanning Tree, is the last sentence
>>>> above actually describing the behavior of a bridge port when a bond is
>>>> the member of the bridge? I'm not sure I understand what "member ports"
>>>> refers to (bridge ports or bonding slaves).
>>>
>>> Ack, maybe replacing the last sentence with something like:
>>> When bond is carrier down, the slave ports are only forwarding
>>> low-level control protocols (e.g. LACP PDU) and discarding all other
>>> packets.
>>
>> Ah, are you actually referring to the fact that slaves that are
>> up will still deliver packets to listeners that bind directly to the
>> slave or hook in through a rx_handler? This is, in part, the
>> "RX_HANDLER_EXACT" business in bond_handle_frame and
>> __netif_receive_skb_core.
>>
>> The decision for that has nothing to do with the protocol; I
>> seem to recall that FCoE (or maybe it's iSCSI) does its regular traffic
>> reception this way (although via dev_add_pack, not an rx_handler) so it
>> can run traffic regardless of the bonding master's state.
>
>I see, it seems you are basically saying; the slaves are up but when the
>logical bond interface is carrier down there was no code changed in
>bond_handle_frame() to actually drop frames other than LACPDUs. So
>basically having this statement makes no sense until there is code to
>actually drop those additional frames.
What I'm saying is that the fact that bond_handle_frame() does
not outright drop anything is a feature, not an oversight. It's done
this way on purpose so that the slave device can be utilized separately
from its participation in the bond. Off the top of my head, as I recall
this is used by LLDP, FCoE, and probably other "converged" Ethernet
facilities. I believe some network monitoring tools use this property
as well, but I don't recall the details.
Frames received on inactive slaves (which these active slaves
under a carrier-off bond are not; more on that in a bit) are marked as
such by the RX_HANDLER_EXACT return, and do not propagate up the stack
in the usual way, but are delivered only to packet listeners that bind
directly to the slave (typically via a dev_add_pack handler, which is
how LACPDUs were received prior to the rx_handler logic being
implemented).
The bonding inactive slave receive logic used to work the other
way around; anything not explicitly needed by the bond itself would be
dropped in this particular case. The "deliver to exact match" logic was
added later.
Now, the possible hole here that I think you're alluding to is
that if the bond sets itself carrier down due to a min_links violation,
the slaves in the active aggregator are not inactive, and there's
nothing in the receive path to prevent incoming packets from being
processed normally if a receiving slave is still up.
A quick test suggests that this is indeed the case; if I set up
a 802.3ad bond with min_links=3 and two slaves, incoming ICMP ECHOs to
the bond's IP appear to be received by icmp_echo, which calls
icmp_reply, which apparently fails. I'm basing this conclusion on the
IcmpInErrors, IcmpInEchoReps, IcmpOutErrors, IcmpMsgInType8 and
IcmpMsgOutType0 stat counters all incrementing more or less in lock
step. I haven't traced the code to see where it fails.
I'm not sure exactly what ought to be done in that case; one
thought (which I have not tested) is bond_should_deliver_exact_match()
always returns true if the bonding master is carrier down.
Transmit appears to not function (at the bond level), so that
part doesn't appear to be an issue.
>>>>> @@ -2381,10 +2386,15 @@ int bond_3ad_set_carrier(struct bonding *bond)
>>>>> ret = 0;
>>>>> goto out;
>>>>> }
>>>>> +
>>>>> + bond_for_each_slave_rcu(bond, slave, iter)
>>>>> + if (SLAVE_AD_INFO(slave)->aggregator.is_active)
>>>>> + active_slaves++;
>>>>> +
>>>>> active = __get_active_agg(&(SLAVE_AD_INFO(first_slave)->aggregator));
>>>>> - if (active) {
>>>>> + if (active && __agg_has_partner(active)) {
>>>>
>>>> Why "__agg_has_partner"? Since the "else" of this clause is:
>>>>
>>>> } else if (netif_carrier_ok(bond->dev)) {
>>>> netif_carrier_off(bond->dev);
>>>> }
>>>>
>>>> I'm wondering if this will do the right thing for the case that
>>>> there are no LACP partners at all (e.g., the switch ports do not have
>>>> LACP enabled), in which case the active aggregator should be a single
>>>> "individual" port as a fallback, but will not have a partner.
>>>>
>>>> -J
>>>>
>>>
>>> I see your point. The initial thinking was the logical bond carrier should
>>> not be brought up until the bond has a partner and is ready to pass
>>> traffic, otherwise we start blackholing frames. Looking over the code it
>>> seems the aggregator.is_individual flag is only set to true when a slave
>>> is in half-duplex, this seems odd?
>>
>> The agg.is_individual flag and an "individual" aggregator are
>> subtly different things.
>>
>> The is_individual flag is part of the implementation of the
>> standard requirement that half duplex ports are not allowed to enable
>> LACP (and thus cannot aggregate, and end up as "Individual" in
>> standard-ese). The standard capitalizes "Individual" when it describes
>> the cannot-aggregate property of a port (note that half duplex is only
>> one reason of many for a port being Individual).
>>
>> An "individual" aggregator (my usage of 802.1AX 5.3.6 (b)) is an
>> aggregator containing exactly one port that is Individual. A port can
>> end up as Individual (for purposes of this discussion) either through
>> the is_individual business, or because the bonding port does run LACP,
>> but the link partner does not, and thus no LACPDUs are ever received.
>>
>> For either of the above cases (is_individual or no-LACP-parter),
>> then the active aggregator will be an "individual" aggregator, but will
>> not have a parter (__agg_has_partner() will be false). The standard has
>> a bunch of verbiage about this in 802.1AX 5.3.5 - 5.3.9.
>>
>>> My initial thinking to alleviate the concern is something like the
>>> following:
>>>
>>> if (active && !SLAVE_AD_INFO(slave)->aggregator.is_individual &&
>>> __agg_has_partner(active)) {
>>> /* set carrier based on min_links */
>>> } else if (active && SLAVE_AD_INFO(slave)->aggregator.is_individual) {
>>> /* set bond carrier state according to carrier state of slave */
>>> } else if (netif_carrier_ok(bond->dev)) {
>>> netif_carrier_off(bond->dev);
>>> }
>>
>> I'm not sure you need to care about is_individual or
>> __agg_has_partner at all. If either of those conditions is true for the
>> active aggregator, it will contain exactly one port, and so if min_links
>> is 2, you'll have carrier off, and if min_links is 1 or less you'll have
>> carrier on.
>>
>> If I'm reading the patch right, the real point (which isn't
>> really described very well in the change log) is that you're changing
>> the carrier decision to count only active ports in the active
>> aggregator, not the total number of ports as is currently done.
>>
>> I'm not sure why this change is needed:
>>
>> @@ -2381,10 +2386,15 @@ int bond_3ad_set_carrier(struct bonding *bond)
>> ret = 0;
>> goto out;
>> }
>> +
>> + bond_for_each_slave_rcu(bond, slave, iter)
>> + if (SLAVE_AD_INFO(slave)->aggregator.is_active)
>> + active_slaves++;
>> +
>> active = __get_active_agg(&(SLAVE_AD_INFO(first_slave)->aggregator));
>> - if (active) {
>> + if (active && __agg_has_partner(active)) {
>> /* are enough slaves available to consider link up? */
>> - if (active->num_of_ports < bond->params.min_links) {
>> + if (active_slaves < bond->params.min_links) {
>> if (netif_carrier_ok(bond->dev)) {
>> netif_carrier_off(bond->dev);
>> goto out;
>>
>> because a port (slave) that loses carrier or whose link partner
>> becomes unresponsive to LACPDUs will be removed from the aggregator. As
>> I recall, there are no "inactive" ports in an aggregator; all of them
>> have to match in terms of capabilities.
>>
>> In other words, I'm unsure of when the count of is_active ports
>> will not match active->num_of_ports.
>>
>> Also, the other parts of the patch add some extra updates to the
>> carrier state when a port is enabled or disabled, e.g.,
>>
>> @@ -189,6 +189,7 @@ static inline int __agg_has_partner(struct aggregator *agg)
>> static inline void __disable_port(struct port *port)
>> {
>> bond_set_slave_inactive_flags(port->slave, BOND_SLAVE_NOTIFY_LATER);
>> + bond_3ad_set_carrier(port->slave->bond);
>> }
>>
>> Again, I'm not sure why this is necessary, as the cases that
>> disable or enable a port will eventually call bond_3ad_set_carrier. For
>> example, ad_agg_selection_logic will, when changing active aggregator,
>> individually disable all ports of the old active and then may
>> individually enable ports of the new active if necessary, and then
>> finally call bond_3ad_set_carrier.
>>
>> In what situations is the patch's behavior an improvement (i.e.,
>> is there a situation I'm missing that doesn't do it right)?
>
>I think the addition of bond_3ad_set_carrier() to both __enable_port() and
>__disable_port() were optimizations so the bond carrier transition would
>happen faster, though I am not certain.
I get the impression that these patches have been around for a
while internally; it would be good to validate that there is an actual
performance change in the current mainline, since there does not seem to
be any functionality change.
>>
>> The last portion of the patch:
>>
>> --- a/drivers/net/bonding/bond_options.c
>> +++ b/drivers/net/bonding/bond_options.c
>> @@ -1181,6 +1181,7 @@ static int bond_option_min_links_set(struct bonding *bond,
>> netdev_info(bond->dev, "Setting min links value to %llu\n",
>> newval->value);
>> bond->params.min_links = newval->value;
>> + bond_set_carrier(bond);
>>
>> return 0;
>> }
>>
>> does seem to fix a legitimate bug, in that when min_links is
>> changed, it does not take effect in real time.
>>
>>> Maybe I am missing something and there is a simpler option.
>>>
>>> Thinking about how to validate this, it seems having a bond with two
>>> slaves and both slaves in half-duplex will force an aggregator that is
>>> individual to be selected.
>>>
>>> Thoughts?
>>
>> That's one way, yes. You'll also get an "individual" aggregator
>> if none of the link partners enable LACP.
>>
>
>It seems it might be better to drop the changes to __enable/disable_port
>and bond_3ad_set_carrier from this patch until more testing can be done
>from me, especially if you agree the other changes in this series are of
>benefit.
I haven't looked at patch 3 in detail yet; patches 2, 4 and 5
appear to be ok.
For this patch, the patch fragment immediately above (min_links
take effect immediately) looks good. I think the rest of it still needs
some evaluation.
-J
---
-Jay Vosburgh, jay.vosburgh@canonical.com
^ permalink raw reply
* Re: [PATCHv2 net-next] xen-netback: always fully coalesce guest Rx packets
From: David Miller @ 2015-01-24 1:57 UTC (permalink / raw)
To: david.vrabel; +Cc: netdev, xen-devel, ian.campbell, wei.liu2
In-Reply-To: <1421765392-30303-1-git-send-email-david.vrabel@citrix.com>
From: David Vrabel <david.vrabel@citrix.com>
Date: Tue, 20 Jan 2015 14:49:52 +0000
> Always fully coalesce guest Rx packets into the minimum number of ring
> slots. Reducing the number of slots per packet has significant
> performance benefits when receiving off-host traffic.
>
> Results from XenServer's performance benchmarks:
>
> Baseline Full coalesce
> Interhost VM receive 7.2 Gb/s 11 Gb/s
> Interhost aggregate 24 Gb/s 24 Gb/s
> Intrahost single stream 14 Gb/s 14 Gb/s
> Intrahost aggregate 34 Gb/s 34 Gb/s
>
> However, this can increase the number of grant ops per packet which
> decreases performance of backend (dom0) to VM traffic (by ~10%)
> /unless/ grant copy has been optimized for adjacent ops with the same
> source or destination (see "grant-table: defer releasing pages
> acquired in a grant copy"[1] expected in Xen 4.6).
>
> [1] http://lists.xen.org/archives/html/xen-devel/2015-01/msg01118.html
>
> Signed-off-by: David Vrabel <david.vrabel@citrix.com>
> Acked-by: Ian Campbell <ian.campbell@citrix.com>
> ---
> Changes in v2:
> - Updated commit message with better results.
Applied, thanks David.
^ permalink raw reply
* Re: [PATCH net-next 0/6] netns: advertise netns via netlink
From: David Miller @ 2015-01-24 1:51 UTC (permalink / raw)
To: nicolas.dichtel; +Cc: netdev
In-Reply-To: <1421763347-4354-1-git-send-email-nicolas.dichtel@6wind.com>
From: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Date: Tue, 20 Jan 2015 15:15:41 +0100
>
> The first patch of the serie fix a bug of the previous serie (present in
> net-next only).
> The rest of the serie adds an attribute to advertise the peer netns for
> rtnetlink messages where this information is needed by userland to be able to
> interpret fully the received message.
"series" has a final 's'. Well, as least you consistently misspell it
3 times :-)
"Series" applied, thanks!
^ permalink raw reply
* Re: [PATCH] net: dsa/mv88e6352: make mv88e6352_wait generic
From: Guenter Roeck @ 2015-01-24 1:47 UTC (permalink / raw)
To: Vivien Didelot, netdev; +Cc: linux-kernel, kernel
In-Reply-To: <1422056081-30088-1-git-send-email-vivien.didelot@savoirfairelinux.com>
On 01/23/2015 03:34 PM, Vivien Didelot wrote:
> Some busy bits are available in the global register 1, such as the ATU
> Busy bit. We may want to use this function to wait for them to change,
> so add a new parameter to mv88e6352_wait() instead of hard-coding
> REG_GLOBAL2.
>
> In the meantime, since the REG_READ() macro already checks for error,
> remove the redundant check for ret < 0.
>
> Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Guenter
^ permalink raw reply
* [PATCH] net: llc: use correct size for sysctl timeout entries
From: Sasha Levin @ 2015-01-24 1:47 UTC (permalink / raw)
To: linux-kernel
Cc: Sasha Levin, Arnaldo Carvalho de Melo, David S. Miller,
open list:NETWORKING [GENERAL]
The timeout entries are sizeof(int) rather than sizeof(long), which
means that when they were getting read we'd also leak kernel memory
to userspace along with the timeout values.
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
---
net/llc/sysctl_net_llc.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/net/llc/sysctl_net_llc.c b/net/llc/sysctl_net_llc.c
index 612a5dd..799bafc 100644
--- a/net/llc/sysctl_net_llc.c
+++ b/net/llc/sysctl_net_llc.c
@@ -18,28 +18,28 @@ static struct ctl_table llc2_timeout_table[] = {
{
.procname = "ack",
.data = &sysctl_llc2_ack_timeout,
- .maxlen = sizeof(long),
+ .maxlen = sizeof(sysctl_llc2_ack_timeout),
.mode = 0644,
.proc_handler = proc_dointvec_jiffies,
},
{
.procname = "busy",
.data = &sysctl_llc2_busy_timeout,
- .maxlen = sizeof(long),
+ .maxlen = sizeof(sysctl_llc2_busy_timeout),
.mode = 0644,
.proc_handler = proc_dointvec_jiffies,
},
{
.procname = "p",
.data = &sysctl_llc2_p_timeout,
- .maxlen = sizeof(long),
+ .maxlen = sizeof(sysctl_llc2_p_timeout),
.mode = 0644,
.proc_handler = proc_dointvec_jiffies,
},
{
.procname = "rej",
.data = &sysctl_llc2_rej_timeout,
- .maxlen = sizeof(long),
+ .maxlen = sizeof(sysctl_llc2_rej_timeout),
.mode = 0644,
.proc_handler = proc_dointvec_jiffies,
},
--
1.7.10.4
^ permalink raw reply related
* Re: [PATCH net-next] act_connmark: Add missing dependency on NF_CONNTRACK_MARK
From: David Miller @ 2015-01-24 1:46 UTC (permalink / raw)
To: tgraf; +Cc: netdev, nbd, jhs
In-Reply-To: <eb5e12e8fec691cb08e663f2f8ed183ac92e5a1f.1421757793.git.tgraf@suug.ch>
From: Thomas Graf <tgraf@suug.ch>
Date: Tue, 20 Jan 2015 13:44:25 +0100
> Depending on NETFILTER is not sufficient to ensure the presence of the
> 'mark' field in nf_conn, also needs to depend on NF_CONNTRACK_MARK.
>
> Fixes: 22a5dc ("net: sched: Introduce connmark action")
> Cc: Felix Fietkau <nbd@openwrt.org>
> Cc: Jamal Hadi Salim <jhs@mojatatu.com>
> Signed-off-by: Thomas Graf <tgraf@suug.ch>
Applied, thanks Thomas.
^ permalink raw reply
* Re: [PATCH] ethernet: fm10k: Actually drop 4 bits
From: Rasmus Villemoes @ 2015-01-24 0:50 UTC (permalink / raw)
To: Vick, Matthew
Cc: Alexander Duyck, Kirsher, Jeffrey T,
e1000-devel@lists.sourceforge.net, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org
In-Reply-To: <D0E81E0C.682EB%matthew.vick@intel.com>
On Sat, Jan 24 2015, "Vick, Matthew" <matthew.vick@intel.com> wrote:
> Good catch! I noticed this too and was getting a patch together to address
> this.
>
> The difference is that I was planning on not silently accepting an invalid
> VLAN ID to begin with and returning FM10K_ERR_PARAM if the VLAN was
> invalid, which I think is a better approach for this situation. If it's
> alright with you, I'll generate the patch shortly and credit you via your
> Reported-by.
Sure, do what you think is best.
Rasmus
^ permalink raw reply
* Re: [PATCH] ethernet: fm10k: Actually drop 4 bits
From: Vick, Matthew @ 2015-01-24 0:18 UTC (permalink / raw)
To: Rasmus Villemoes, Alexander Duyck, Kirsher, Jeffrey T
Cc: e1000-devel@lists.sourceforge.net, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org
In-Reply-To: <1421967198-16667-1-git-send-email-linux@rasmusvillemoes.dk>
On 1/22/15, 2:53 PM, "Rasmus Villemoes" <linux@rasmusvillemoes.dk> wrote:
>The comment explains the intention, but vid has type u16. Before the
>inner shift, it is promoted to int, which has plenty of space for all
>vid's bits, so nothing is dropped. Use a simple mask instead.
>
>Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
>---
> drivers/net/ethernet/intel/fm10k/fm10k_pf.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
>diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_pf.c
>b/drivers/net/ethernet/intel/fm10k/fm10k_pf.c
>index 275423d4f777..b1c57d0166a9 100644
>--- a/drivers/net/ethernet/intel/fm10k/fm10k_pf.c
>+++ b/drivers/net/ethernet/intel/fm10k/fm10k_pf.c
>@@ -335,7 +335,7 @@ static s32 fm10k_update_xc_addr_pf(struct fm10k_hw
>*hw, u16 glort,
> return FM10K_ERR_PARAM;
>
> /* drop upper 4 bits of VLAN ID */
>- vid = (vid << 4) >> 4;
>+ vid &= 0x0fff;
>
> /* record fields */
> mac_update.mac_lower = cpu_to_le32(((u32)mac[2] << 24) |
Good catch! I noticed this too and was getting a patch together to address
this.
The difference is that I was planning on not silently accepting an invalid
VLAN ID to begin with and returning FM10K_ERR_PARAM if the VLAN was
invalid, which I think is a better approach for this situation. If it's
alright with you, I'll generate the patch shortly and credit you via your
Reported-by.
Cheers,
Matthew
^ permalink raw reply
* [PATCH] net: dsa/mv88e6352: make mv88e6352_wait generic
From: Vivien Didelot @ 2015-01-23 23:34 UTC (permalink / raw)
To: netdev; +Cc: Vivien Didelot, Guenter Roeck, linux-kernel, kernel
Some busy bits are available in the global register 1, such as the ATU
Busy bit. We may want to use this function to wait for them to change,
so add a new parameter to mv88e6352_wait() instead of hard-coding
REG_GLOBAL2.
In the meantime, since the REG_READ() macro already checks for error,
remove the redundant check for ret < 0.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
---
drivers/net/dsa/mv88e6352.c | 13 +++++--------
1 file changed, 5 insertions(+), 8 deletions(-)
diff --git a/drivers/net/dsa/mv88e6352.c b/drivers/net/dsa/mv88e6352.c
index 258d9ef..e13adc7 100644
--- a/drivers/net/dsa/mv88e6352.c
+++ b/drivers/net/dsa/mv88e6352.c
@@ -22,17 +22,14 @@
#include <net/dsa.h>
#include "mv88e6xxx.h"
-static int mv88e6352_wait(struct dsa_switch *ds, int reg, u16 mask)
+static int mv88e6352_wait(struct dsa_switch *ds, int reg, int offset, u16 mask)
{
unsigned long timeout = jiffies + HZ / 10;
while (time_before(jiffies, timeout)) {
int ret;
- ret = REG_READ(REG_GLOBAL2, reg);
- if (ret < 0)
- return ret;
-
+ ret = REG_READ(reg, offset);
if (!(ret & mask))
return 0;
@@ -43,17 +40,17 @@ static int mv88e6352_wait(struct dsa_switch *ds, int reg, u16 mask)
static inline int mv88e6352_phy_wait(struct dsa_switch *ds)
{
- return mv88e6352_wait(ds, 0x18, 0x8000);
+ return mv88e6352_wait(ds, REG_GLOBAL2, 0x18, 0x8000);
}
static inline int mv88e6352_eeprom_load_wait(struct dsa_switch *ds)
{
- return mv88e6352_wait(ds, 0x14, 0x0800);
+ return mv88e6352_wait(ds, REG_GLOBAL2, 0x14, 0x0800);
}
static inline int mv88e6352_eeprom_busy_wait(struct dsa_switch *ds)
{
- return mv88e6352_wait(ds, 0x14, 0x8000);
+ return mv88e6352_wait(ds, REG_GLOBAL2, 0x14, 0x8000);
}
static int __mv88e6352_phy_read(struct dsa_switch *ds, int addr, int regnum)
--
2.2.2
^ permalink raw reply related
* Re: [net-next PATCH v3 00/12] Flow API
From: Thomas Graf @ 2015-01-23 23:16 UTC (permalink / raw)
To: John Fastabend
Cc: Jiri Pirko, Jamal Hadi Salim, Pablo Neira Ayuso, simon.horman,
sfeldma, netdev, davem, gerlitz.or, andy, ast
In-Reply-To: <54C2A805.7030401@gmail.com>
[Skipping the tc model part for now. Need some time to digest
all of that]
On 01/23/15 at 11:59am, John Fastabend wrote:
> On 01/23/2015 09:46 AM, Thomas Graf wrote:
> >.... if we can get rid of the rtnl lock in the flow mod path ;-)
>
> Well isn't it the qdisc lock here? And its not needed anymore for
> filters/actions only qdisc's use it because they are not lock-safe
> yet. Its been on my backlog to start replacing the skb lists with
> lock-free rings but I haven't got anywhere on this yet.
>
> Although a hardware doesn't really need a queuing discipline its
> done in hardware so you could drop the qdisc lock in this case.
I'm not even in the data path yet with that comment. I'm worried
with the locking in the control path as talking rtnetlink implies
taking rtnl for each flow modification.
Agreed that we wouldn't depend on the qdisc lock for offloaded flows.
^ permalink raw reply
* Re: [PATCH net-next v3 2/5] swdevice: add new api to set and del bridge port attributes
From: roopa @ 2015-01-23 23:10 UTC (permalink / raw)
To: Jiri Pirko
Cc: sfeldma, jhs, bcrl, tgraf, john.fastabend, stephen, vyasevic,
ronen.arad, netdev, davem, shm, gospo
In-Reply-To: <54C2CF1A.1080905@cumulusnetworks.com>
On 1/23/15, 2:45 PM, roopa wrote:
> On 1/23/15, 8:06 AM, Jiri Pirko wrote:
>> Fri, Jan 23, 2015 at 04:58:57PM CET, roopa@cumulusnetworks.com wrote:
>>> On 1/23/15, 2:41 AM, Jiri Pirko wrote:
>>>
>>> <snip..>
>>>> +
>>>> +/**
>>>> + * netdev_switch_port_bridge_dellink - Notify switch device
>>>> port of bridge
>>>> + * attribute delete
>>>> + *
>>>> + * @dev: port device
>>>> + * @nlh: netlink msg with bridge port attributes
>>>> + *
>>>> + * Notify switch device port of bridge port attribute delete
>>>> + */
>>>> +int netdev_switch_port_bridge_dellink(struct net_device *dev,
>>>> + struct nlmsghdr *nlh, u16 flags)
>>>> +{
>>>> + const struct net_device_ops *ops = dev->netdev_ops;
>>>> + struct net_device *lower_dev;
>>>> + struct list_head *iter;
>>>> + int ret = 0, err = 0;
>>>> +
>>>> + if (!(dev->features & NETIF_F_HW_NETFUNC_OFFLOAD))
>>>> + return err;
>>>> +
>>>> + if (ops->ndo_bridge_dellink) {
>>>> + WARN_ON(!ops->ndo_switch_parent_id_get);
>>>> + return ops->ndo_bridge_dellink(dev, nlh, flags);
>>>> + }
>>>> +
>>>> + netdev_for_each_lower_dev(dev, lower_dev, iter) {
>>>> + err = netdev_switch_port_bridge_dellink(lower_dev, nlh,
>>>> flags);
>>>> + if (err)
>>>> + ret = err;
>>>> + }
>>>> +
>>>> + return ret;
>>>> +}
>>>> +EXPORT_SYMBOL(netdev_switch_port_bridge_dellink);
>>>> --
>>>> 1.7.10.4
>>>>
>>>> Is there any other place, other than bridge code, this functions are
>>>> suppored to be called from?
>>> No other place today. Its usually the master that implements
>>> ndo_bridge_setlink/dellink.
>>>
>>>> If not, which I consider likely, it would
>>>> make more sense to me to:
>>>>
>>>> - move netdev_for_each_lower_dev iterations directly to bridge code
>>>> - let the masters (bond, team, ..) implement ndo_bridge_*link and do
>>>> the traversing there (can be in a form of pre-prepared default
>>>> ndo callback (ndo_dflt_netdev_switch_port_bridge_*link)
>>> But, i am still not understanding why i would modify bond, team and
>>> other
>>> slaves
>> Well, that is the usual way to propagate ndo calls. People are used to
>> this. It is visible right away in bonding/other code that is propagated
>> some ndo call to slaves. With your code, that is somehow hidden and only
>> dependent on NETIF_F_HW_NETFUNC_OFFLOAD flag.
>>
>> Note that there are only couple of "master drivers" (for this, most
>> likely
>> only bond and team modifications are needed).
> ndo_bridge_setlink today is only implemented by drivers that implement
> bridging function.
> So, having the bond and team driver implement it...seems odd.
>
> But if you insist, i am going to do just that.
A side note, I dont see any reason for ndo_bridge_setlink to be renamed
to ndo_setlink. Because it seems to take the whole netlink msg anyways.
It can be used to offload other link attributes besides bridging (vxlan
and so on).
Any thoughts on that ?.
^ permalink raw reply
* Re: [PATCH net-next v3 1/5] netdev: introduce new NETIF_F_HW_NETFUNC_OFFLOAD feature flag for switch device offloads
From: roopa @ 2015-01-23 23:09 UTC (permalink / raw)
To: Jiri Pirko
Cc: Thomas Graf, sfeldma, jhs, bcrl, john.fastabend, stephen,
vyasevic, ronen.arad, netdev, davem, shm, gospo
In-Reply-To: <20150123190518.GP2065@nanopsycho.orion>
On 1/23/15, 11:05 AM, Jiri Pirko wrote:
> Fri, Jan 23, 2015 at 06:56:54PM CET, tgraf@suug.ch wrote:
>> On 01/23/15 at 07:48am, roopa wrote:
>>> On 1/23/15, 1:44 AM, Jiri Pirko wrote:
>>>> How about rather "HW_DATAPATH_OFFLOAD"? Feels more accurate. By the
>>>> name, I still cannot understand what NETFUNC should mean.
>>> It was supposed to mean 'network function offload'. sure, will consider
>>> HW_DATAPATH_OFFLOAD
>>> or if anybody has other suggestions.
>> I'm still fascinated that the description says 'for all switch asic
>> offloads', 'switch drivers', 'switch ports' but we try very hard to
>> come up with a different name ;-)
> +1
> But we'll see in future. I bet that "switch" will stick.
because....am sure anything else will get rejected ;).
...The API/patches is currently targeting switch ASIC devices but the
idea behind having the flag
not be called 'switch' was to cover all the devices being discussed
today (previous threads on this indicate that).
Its fair to say that I only care about switch asics today. If that's the
popular vote i am going with that.
renaming it in the next series...
^ permalink raw reply
* Re: [PATCH] ping: Fix race in free in receive path
From: Eric Dumazet @ 2015-01-23 23:07 UTC (permalink / raw)
To: subashab; +Cc: netdev, edumazet
In-Reply-To: <45fab13fce077924f957cba84ba20ba4.squirrel@www.codeaurora.org>
On Fri, 2015-01-23 at 22:26 +0000, subashab@codeaurora.org wrote:
> An exception is seen in ICMP ping receive path where the skb
> destructor sock_rfree() tries to access a freed socket. This happens
> because ping_rcv() releases socket reference with sock_put() and this
> internally frees up the socket. Later icmp_rcv() will try to free the
> skb and as part of this, skb destructor is called and which leads
> to a kernel panic as the socket is freed already in ping_rcv().
...
> Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
> Cc: Eric Dumazet <edumazet@google.com>
> ---
Thanks !
Signed-off-by: Eric Dumazet <edumazet@google.com>
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox