* [PATCH 0/3 v5] Open vSwitch zerocopy upcall
From: Thomas Graf @ 2013-11-11 15:47 UTC (permalink / raw)
To: jesse, davem; +Cc: dev, netdev, eric.dumazet, dborkman, bhutchings
Respin of the zerocopy patches for the openvswitch upcall.
V5: - Removed padding requirement in user space
- Added OVS_DP_F_UNALIGNED flag let user space signal ability to
receive unaligned Netlink messages, fall back to linear copy
otherwise.
V4: - Daniel Borkmann pointed out that the style in skbuff.h has changed
in net-next, adapted to no longer using extern in headers.
V3: - Removed unneeded alignment of nlmsg_len after padding
V2: - Added skb_zerocopy_headlen() to calculate headroom of destination
buffer. This also takes care of the from->head_frag issue.
- Attribute alignment for frag_list case
- API documentation
- performance data for CHECKSUM_PARTIAL tx case
Thomas Graf (3):
net: Export skb_zerocopy() to zerocopy from one skb to another
openvswitch: Allow user space to announce ability to accept unaligned
Netlink messages
openvswitch: Use skb_zerocopy() for upcall
include/linux/skbuff.h | 3 ++
include/uapi/linux/openvswitch.h | 4 ++
net/core/skbuff.c | 85 ++++++++++++++++++++++++++++++++++++
net/netfilter/nfnetlink_queue_core.c | 59 ++-----------------------
net/openvswitch/datapath.c | 58 ++++++++++++++++--------
net/openvswitch/datapath.h | 2 +
6 files changed, 139 insertions(+), 72 deletions(-)
--
1.8.3.1
^ permalink raw reply
* [PATCH 3/3 net-next] openvswitch: Use skb_zerocopy() for upcall
From: Thomas Graf @ 2013-11-11 15:47 UTC (permalink / raw)
To: jesse, davem; +Cc: dev, netdev, eric.dumazet, dborkman, bhutchings
In-Reply-To: <cover.1384184208.git.tgraf@suug.ch>
Use of skb_zerocopy() avoids the expensive call to memcpy() when
copying the packet data into the Netlink skb. Completes checksum
through skb_checksum_help() if needed.
Cost of memcpy is significantly reduced from:
+ 7.48% vhost-8471 [k] memcpy
+ 5.57% ovs-vswitchd [k] memcpy
+ 2.81% vhost-8471 [k] csum_partial_copy_generic
to:
+ 5.72% ovs-vswitchd [k] memcpy
+ 3.32% vhost-5153 [k] memcpy
+ 0.68% vhost-5153 [k] skb_zerocopy
(megaflows disabled)
Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
net/openvswitch/datapath.c | 54 +++++++++++++++++++++++++++++++---------------
1 file changed, 37 insertions(+), 17 deletions(-)
diff --git a/net/openvswitch/datapath.c b/net/openvswitch/datapath.c
index 0a50574..8dc2496 100644
--- a/net/openvswitch/datapath.c
+++ b/net/openvswitch/datapath.c
@@ -108,10 +108,10 @@ int lockdep_ovsl_is_held(void)
#endif
static struct vport *new_vport(const struct vport_parms *);
-static int queue_gso_packets(struct net *, int dp_ifindex, struct sk_buff *,
- const struct dp_upcall_info *);
-static int queue_userspace_packet(struct net *, int dp_ifindex,
- struct sk_buff *,
+static int queue_gso_packets(struct datapath *, struct net *, int dp_ifindex,
+ struct sk_buff *, const struct dp_upcall_info *);
+static int queue_userspace_packet(struct datapath *, struct net *,
+ int dp_ifindex, struct sk_buff *,
const struct dp_upcall_info *);
/* Must be called with rcu_read_lock or ovs_mutex. */
@@ -292,9 +292,9 @@ int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb,
}
if (!skb_is_gso(skb))
- err = queue_userspace_packet(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
+ err = queue_userspace_packet(dp, ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
else
- err = queue_gso_packets(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
+ err = queue_gso_packets(dp, ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info);
if (err)
goto err;
@@ -310,7 +310,7 @@ err:
return err;
}
-static int queue_gso_packets(struct net *net, int dp_ifindex,
+static int queue_gso_packets(struct datapath *dp, struct net *net, int dp_ifindex,
struct sk_buff *skb,
const struct dp_upcall_info *upcall_info)
{
@@ -327,7 +327,7 @@ static int queue_gso_packets(struct net *net, int dp_ifindex,
/* Queue all of the segments. */
skb = segs;
do {
- err = queue_userspace_packet(net, dp_ifindex, skb, upcall_info);
+ err = queue_userspace_packet(dp, net, dp_ifindex, skb, upcall_info);
if (err)
break;
@@ -381,10 +381,11 @@ static size_t key_attr_size(void)
}
static size_t upcall_msg_size(const struct sk_buff *skb,
- const struct nlattr *userdata)
+ const struct nlattr *userdata,
+ unsigned int hdrlen)
{
size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
- + nla_total_size(skb->len) /* OVS_PACKET_ATTR_PACKET */
+ + nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
+ nla_total_size(key_attr_size()); /* OVS_PACKET_ATTR_KEY */
/* OVS_PACKET_ATTR_USERDATA */
@@ -394,14 +395,15 @@ static size_t upcall_msg_size(const struct sk_buff *skb,
return size;
}
-static int queue_userspace_packet(struct net *net, int dp_ifindex,
- struct sk_buff *skb,
+static int queue_userspace_packet(struct datapath *dp, struct net *net,
+ int dp_ifindex, struct sk_buff *skb,
const struct dp_upcall_info *upcall_info)
{
struct ovs_header *upcall;
struct sk_buff *nskb = NULL;
struct sk_buff *user_skb; /* to be queued to userspace */
struct nlattr *nla;
+ unsigned int hlen;
int err;
if (vlan_tx_tag_present(skb)) {
@@ -422,7 +424,21 @@ static int queue_userspace_packet(struct net *net, int dp_ifindex,
goto out;
}
- user_skb = genlmsg_new(upcall_msg_size(skb, upcall_info->userdata), GFP_ATOMIC);
+ /* Complete checksum if needed */
+ if (skb->ip_summed == CHECKSUM_PARTIAL &&
+ (err = skb_checksum_help(skb)))
+ goto out;
+
+ /* Older versions of OVS user space enforce alignment of the last
+ * Netlink attribute to NLA_ALIGNTO which would require extensive
+ * padding logic. Only perform zerocopy if padding is not required.
+ */
+ if (dp->user_features & OVS_DP_F_UNALIGNED)
+ hlen = skb_zerocopy_headlen(skb);
+ else
+ hlen = skb->len;
+
+ user_skb = genlmsg_new(upcall_msg_size(skb, upcall_info->userdata, hlen), GFP_ATOMIC);
if (!user_skb) {
err = -ENOMEM;
goto out;
@@ -441,13 +457,17 @@ static int queue_userspace_packet(struct net *net, int dp_ifindex,
nla_len(upcall_info->userdata),
nla_data(upcall_info->userdata));
- nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len);
+ /* Only reserve room for attribute header, packet data is added
+ * in skb_zerocopy() */
+ if (!(nla = nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, 0)))
+ goto out;
+ nla->nla_len = nla_attr_size(skb->len);
+
+ skb_zerocopy(user_skb, skb, skb->len, hlen);
- skb_copy_and_csum_dev(skb, nla_data(nla));
+ ((struct nlmsghdr *) user_skb->data)->nlmsg_len = user_skb->len;
- genlmsg_end(user_skb, upcall);
err = genlmsg_unicast(net, user_skb, upcall_info->portid);
-
out:
kfree_skb(nskb);
return err;
--
1.8.3.1
^ permalink raw reply related
* RE: TCP performance regression
From: Eric Dumazet @ 2013-11-11 16:17 UTC (permalink / raw)
To: David Laight; +Cc: Sujith Manoharan, netdev, Dave Taht
In-Reply-To: <AE90C24D6B3A694183C094C60CF0A2F6026B73FB@saturn3.aculab.com>
On Mon, 2013-11-11 at 15:43 +0000, David Laight wrote:
> > > Or, maybe:
> > > 5) call skb_orphan() (I think that is the correct function) when transmit
> > > packets are given to the hardware.
> >
> > This is the worth possible solution, as it basically re-enables
> ^^^^^ worst ?
> > bufferbloat again.
>
> It should be ok if the mac driver only gives the hardware a small
> number of bytes/packets - or one appropriate for the link speed.
There is some confusion here.
mvneta has a TX ring buffer, which can hold up to 532 TX descriptors.
If this driver used skb_orphan(), a single TCP flow could use the whole
TX ring.
TCP Small Queue would only limit the number of skbs on Qdisc.
Try then to send a ping message, it will have to wait a lot.
^ permalink raw reply
* Re: TCP performance regression
From: Sujith Manoharan @ 2013-11-11 16:13 UTC (permalink / raw)
To: Eric Dumazet; +Cc: Felix Fietkau, netdev, Dave Taht
In-Reply-To: <1384180069.16391.32.camel@edumazet-glaptop2.roam.corp.google.com>
Eric Dumazet wrote:
> We have many choices.
>
> 1) Add back a minimum of ~128 K of outstanding bytes per TCP session,
> so that buggy drivers can sustain 'line rate'.
>
> Note that with 100 concurrent TCP streams, total amount of bytes
> queued on the NIC is 12 MB.
> And pfifo_fast qdisc will drop packets anyway.
>
> Thats what we call 'BufferBloat'
>
> 2) Try lower values like 64K. Still bufferbloat.
>
> 3) Fix buggy drivers, using a proper logic, or shorter timers (mvneta
> case for example)
>
> 4) Add a new netdev attribute, so that well behaving NIC drivers do not
> have to artificially force TCP stack to queue too many bytes in
> Qdisc/NIC queues.
I think the quirks of 802.11 aggregation should be taken into account.
I am adding Felix to this thread, who would have more to say on latency/bufferbloat
with wireless drivers.
Sujith
^ permalink raw reply
* Re: [PATCH v2] mac80211: add assoc beacon timeout logic
From: Felipe Contreras @ 2013-11-11 16:23 UTC (permalink / raw)
To: Johannes Berg
Cc: linux-wireless Mailing List, netdev, John W. Linville,
David S. Miller
In-Reply-To: <1384184624.14334.31.camel-8Nb76shvtaUJvtFkdXX2HixXY32XiHfO@public.gmane.org>
On Mon, Nov 11, 2013 at 9:43 AM, Johannes Berg
<johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> wrote:
> On Mon, 2013-11-11 at 04:59 -0600, Felipe Contreras wrote:
>
>> Well the AP is sending beacons, but they seem to be corrupted,
>> although the corruption often seems to happen in a place that is not
>> so important.
>
> Indeed - the beacon you sent to me in private is damaged somewhere
> towards the end of the frame. Are we actually receiving it but ignoring
> it because it doesn't have the data we need?
The driver is not receiving it at all. I already debugged this:
http://article.gmane.org/gmane.linux.kernel.wireless.general/115429
However, I noticed that once in a very long time, sometimes it does
receive the corrupted frame and the association continues, and the
driver code detects it's a corrupted beacon frame.
> The firmware still
> shouldn't be filtering anything since it doesn't really look at the
> beacon information (or maybe it filters based on the DS IE? I'm not
> entirely sure)
That's what I thought, but I don't see it at all (only in monitor
mode, and in ad-hoc).
>> However, if I apply this patch, I don't notice any issue, it
>> associates and works fine. Maybe there's some subtle issues with
>> features I don't personally use, or perhaps there's the occasional
>> disconnection (although that could be due to something else), but
>> that's light years away from not associating at all.
>>
>> I'd say between a) some features not working and b) nothing working at
>> all, a) is preferred.
>>
>> If you think it's better that nothing works at all, then wouldn't it
>> make sense to time out and return an error? Currently we just keep
>> trying to associate *forever*.
>
> That's wpa_supplicant/userspace behaviour. The kernel will just drop the
> connection.
Nope, it keeps trying forever.
Oct 13 14:33:15 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
Oct 13 14:33:15 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
Oct 13 14:33:15 nysa kernel: wlan0: authenticated
Oct 13 14:33:15 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
Oct 13 14:33:18 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
Oct 13 14:33:18 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
Oct 13 14:33:18 nysa kernel: wlan0: authenticated
Oct 13 14:33:18 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
Oct 13 14:33:22 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
Oct 13 14:33:22 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
Oct 13 14:33:22 nysa kernel: wlan0: authenticated
Oct 13 14:33:22 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
...
>> > If the AP is sending beacons but the device isn't receiving them, then
>> > it's a driver bug and mac80211 shouldn't work around it.
>>
>> I agree, but I can't seem to convince Intel guys of that. The firmware
>> is dropping the corrupt beacon frames (although not always), so
>> there's nothing the driver can do afterwards.
>
> You realize I work for the same team in Intel as well? :)
Now I do.
>> But even if there were not beacons at all (corrupt or otherwise), I
>> still think waiting *forever* in a loop is not ideal, a) is preferred;
>> not having all the features, but still somehow work (from my point of
>> view it's more than somewhat).
>
> This isn't really true like I said above - the kernel can only drop the
> association, if userspace *insists* then it will try again and again.
But it's not doing this:
ieee80211_destroy_assoc_data(sdata, false);
cfg80211_assoc_timeout(sdata->dev, bss);
Which is what causes the association to stop for me.
So where exactly in the code is the association being "dropped"?
> I'd much rather try to get to the bottom of this. Maybe the firmware is
> dropping the beacon because the DS IE is broken? Or are we receiving it
> but ignoring it because it's broken?
It's not the latter.
I would rather fix the problem at the two levels, so even if the
firmware passes the corrupt frames correctly, the driver would still
somewhat work when there's no beacon frames at all.
> Unfortunately, there's only so much we can do to work around broken APs.
Indeed, but 'so much' for this AP is really nothing, while with my
patch it's quite a lot.
--
Felipe Contreras
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" 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
* Re: [patch iproute2] ipaddress: add a black line for each device to make the output more readable
From: Jiri Benc @ 2013-11-11 16:31 UTC (permalink / raw)
To: Hangbin Liu; +Cc: network dev
In-Reply-To: <1384182210-8901-1-git-send-email-liuhangbin@gmail.com>
On Mon, 11 Nov 2013 23:03:30 +0800, Hangbin Liu wrote:
> When we have multi links, the output huddled together and make it hard to read.
> Let's use the old ifconfig output style.
No. You can't break others' scripts like that.
bash$ function ip { /sbin/ip "$@" | sed '2,$s/^[0-9]*:/\n&/' ; }
^^^ Here, fixed that for you.
Jiri
--
Jiri Benc
^ permalink raw reply
* RE: TCP performance regression
From: David Laight @ 2013-11-11 16:35 UTC (permalink / raw)
To: Eric Dumazet; +Cc: Sujith Manoharan, netdev, Dave Taht
In-Reply-To: <1384186622.16391.44.camel@edumazet-glaptop2.roam.corp.google.com>
> > It should be ok if the mac driver only gives the hardware a small
> > number of bytes/packets - or one appropriate for the link speed.
>
> There is some confusion here.
>
> mvneta has a TX ring buffer, which can hold up to 532 TX descriptors.
>
> If this driver used skb_orphan(), a single TCP flow could use the whole
> TX ring.
>
> TCP Small Queue would only limit the number of skbs on Qdisc.
>
> Try then to send a ping message, it will have to wait a lot.
532 is a ridiculously large number especially for a slow interface.
At a guess you don't want more than 10-20ms of data in the tx ring.
You might need extra descriptors for badly fragmented packets.
David
^ permalink raw reply
* Re: TCP performance regression
From: Felix Fietkau @ 2013-11-11 16:38 UTC (permalink / raw)
To: Sujith Manoharan, Eric Dumazet; +Cc: netdev, Dave Taht
In-Reply-To: <21121.575.539384.948990@gargle.gargle.HOWL>
On 2013-11-11 17:13, Sujith Manoharan wrote:
> Eric Dumazet wrote:
>> We have many choices.
>>
>> 1) Add back a minimum of ~128 K of outstanding bytes per TCP session,
>> so that buggy drivers can sustain 'line rate'.
>>
>> Note that with 100 concurrent TCP streams, total amount of bytes
>> queued on the NIC is 12 MB.
>> And pfifo_fast qdisc will drop packets anyway.
>>
>> Thats what we call 'BufferBloat'
>>
>> 2) Try lower values like 64K. Still bufferbloat.
>>
>> 3) Fix buggy drivers, using a proper logic, or shorter timers (mvneta
>> case for example)
>>
>> 4) Add a new netdev attribute, so that well behaving NIC drivers do not
>> have to artificially force TCP stack to queue too many bytes in
>> Qdisc/NIC queues.
>
> I think the quirks of 802.11 aggregation should be taken into account.
> I am adding Felix to this thread, who would have more to say on latency/bufferbloat
> with wireless drivers.
I don't think this issue is about something as simple as timer handling
for tx completion (or even broken/buggy drivers).
There's simply no way to make 802.11 aggregation work well and have
similar tx completion latency characteristics as Ethernet devices.
802.11 aggregation reduces the per-packet airtime overhead by combining
multiple packets into one transmission (saving a lot of time getting a
tx opportunity, transmitting the PHY header, etc.), which makes the
'line rate' heavily depend on the amount of buffering.
Aggregating multiple packets into one transmission also causes extra
packet loss, which is compensated by retransmission and reordering, thus
introducing additional latency.
I don't think that TSQ can do a decent job of mitigating bufferbloat on
802.11n devices without a significant performance hit, so adding a new
netdev attribute might be a good idea.
- Felix
^ permalink raw reply
* Re: [PATCH v2] mac80211: add assoc beacon timeout logic
From: Johannes Berg @ 2013-11-11 16:41 UTC (permalink / raw)
To: Felipe Contreras
Cc: linux-wireless Mailing List, netdev, John W. Linville,
David S. Miller
In-Reply-To: <CAMP44s2cVVe4-yW214B9hJu7n6M1onZpZaPCj5N0Fi1JWDKKsw@mail.gmail.com>
On Mon, 2013-11-11 at 10:23 -0600, Felipe Contreras wrote:
> The driver is not receiving it at all. I already debugged this:
>
> http://article.gmane.org/gmane.linux.kernel.wireless.general/115429
Hmm, ok. I pretty much didn't read that thread since some others were
jumping in.
> However, I noticed that once in a very long time, sometimes it does
> receive the corrupted frame and the association continues, and the
> driver code detects it's a corrupted beacon frame.
So how does it treat the corruption?
> > The firmware still
> > shouldn't be filtering anything since it doesn't really look at the
> > beacon information (or maybe it filters based on the DS IE? I'm not
> > entirely sure)
>
> That's what I thought, but I don't see it at all (only in monitor
> mode, and in ad-hoc).
Yes, that part is odd - that's really the root cause.
I didn't quickly find in the threads what device and firmware you were
using, mind identifying it (again)?
> Nope, it keeps trying forever.
>
> Oct 13 14:33:15 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
> Oct 13 14:33:15 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
> Oct 13 14:33:15 nysa kernel: wlan0: authenticated
> Oct 13 14:33:15 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
> Oct 13 14:33:18 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
> Oct 13 14:33:18 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
> Oct 13 14:33:18 nysa kernel: wlan0: authenticated
> Oct 13 14:33:18 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
> Oct 13 14:33:22 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
> Oct 13 14:33:22 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
> Oct 13 14:33:22 nysa kernel: wlan0: authenticated
> Oct 13 14:33:22 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
> ...
I see the same behaviour - but it's the supplicant's doing, it is indeed
getting the event that the AP connection failed (timed out):
wlan0: Event ASSOC_TIMED_OUT (15) received
> > This isn't really true like I said above - the kernel can only drop the
> > association, if userspace *insists* then it will try again and again.
>
> But it's not doing this:
>
> ieee80211_destroy_assoc_data(sdata, false);
> cfg80211_assoc_timeout(sdata->dev, bss);
>
> Which is what causes the association to stop for me.
>
> So where exactly in the code is the association being "dropped"?
This does get called in my setup.
> I would rather fix the problem at the two levels, so even if the
> firmware passes the corrupt frames correctly, the driver would still
> somewhat work when there's no beacon frames at all.
Like I said before - trying to work with an AP without beacons at all is
really bad, we shouldn't be doing it. We might not properly react to
radar events, and other things, for example.
johannes
^ permalink raw reply
* Re: TCP performance regression
From: Eric Dumazet @ 2013-11-11 16:44 UTC (permalink / raw)
To: Sujith Manoharan; +Cc: Arnaud Ebalard, netdev, Dave Taht, Thomas Petazzoni
In-Reply-To: <1384180787.16391.35.camel@edumazet-glaptop2.roam.corp.google.com>
On Mon, 2013-11-11 at 06:39 -0800, Eric Dumazet wrote:
> How following patch helps mvneta performance on current net-next tree
> for a single TCP (sending) flow ?
>
v2 (more chance to even compile ;)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 7d99e695a110..e8211277f15d 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -172,12 +172,12 @@
/* Various constants */
/* Coalescing */
-#define MVNETA_TXDONE_COAL_PKTS 16
+#define MVNETA_TXDONE_COAL_PKTS 1
#define MVNETA_RX_COAL_PKTS 32
#define MVNETA_RX_COAL_USEC 100
/* Timer */
-#define MVNETA_TX_DONE_TIMER_PERIOD 10
+#define MVNETA_TX_DONE_TIMER_PERIOD 1
/* Napi polling weight */
#define MVNETA_RX_POLL_WEIGHT 64
^ permalink raw reply related
* Re: [PATCH v2] mac80211: add assoc beacon timeout logic
From: Felipe Contreras @ 2013-11-11 16:53 UTC (permalink / raw)
To: Johannes Berg
Cc: linux-wireless Mailing List, netdev, John W. Linville,
David S. Miller
In-Reply-To: <1384188067.14334.45.camel@jlt4.sipsolutions.net>
On Mon, Nov 11, 2013 at 10:41 AM, Johannes Berg
<johannes@sipsolutions.net> wrote:
> On Mon, 2013-11-11 at 10:23 -0600, Felipe Contreras wrote:
>> However, I noticed that once in a very long time, sometimes it does
>> receive the corrupted frame and the association continues, and the
>> driver code detects it's a corrupted beacon frame.
>
> So how does it treat the corruption?
wlan0: associating with AP with corrupt beacon
>> > The firmware still
>> > shouldn't be filtering anything since it doesn't really look at the
>> > beacon information (or maybe it filters based on the DS IE? I'm not
>> > entirely sure)
>>
>> That's what I thought, but I don't see it at all (only in monitor
>> mode, and in ad-hoc).
>
> Yes, that part is odd - that's really the root cause.
>
> I didn't quickly find in the threads what device and firmware you were
> using, mind identifying it (again)?
Intel Corporation Centrino Advanced-N 6235 (rev 24)
iwlwifi 0000:02:00.0: loaded firmware version 18.168.6.1 op_mode iwldvm
>> Nope, it keeps trying forever.
>>
>> Oct 13 14:33:15 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
>> Oct 13 14:33:15 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
>> Oct 13 14:33:15 nysa kernel: wlan0: authenticated
>> Oct 13 14:33:15 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
>> Oct 13 14:33:18 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
>> Oct 13 14:33:18 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
>> Oct 13 14:33:18 nysa kernel: wlan0: authenticated
>> Oct 13 14:33:18 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
>> Oct 13 14:33:22 nysa kernel: wlan0: authenticate with e0:1d:3b:46:82:a0
>> Oct 13 14:33:22 nysa kernel: wlan0: send auth to e0:1d:3b:46:82:a0 (try 1/3)
>> Oct 13 14:33:22 nysa kernel: wlan0: authenticated
>> Oct 13 14:33:22 nysa kernel: wlan0: waiting for beacon from e0:1d:3b:46:82:a0
>> ...
>
> I see the same behaviour - but it's the supplicant's doing, it is indeed
> getting the event that the AP connection failed (timed out):
>
> wlan0: Event ASSOC_TIMED_OUT (15) received
Not in my setup.
>> > This isn't really true like I said above - the kernel can only drop the
>> > association, if userspace *insists* then it will try again and again.
>>
>> But it's not doing this:
>>
>> ieee80211_destroy_assoc_data(sdata, false);
>> cfg80211_assoc_timeout(sdata->dev, bss);
>>
>> Which is what causes the association to stop for me.
>>
>> So where exactly in the code is the association being "dropped"?
>
> This does get called in my setup.
Yes, because your setup is receiving beacons.
Check the code:
if ((ifmgd->assoc_data->need_beacon && !ifmgd->have_beacon) ||
ieee80211_do_assoc(sdata)) {
struct cfg80211_bss *bss = ifmgd->assoc_data->bss;
ieee80211_destroy_assoc_data(sdata, false);
cfg80211_assoc_timeout(sdata->dev, bss);
}
If there's no beacon, cfg80211_assoc_timeout() is not called.
I'm sure if you don't call ieee80211_rx_mgmt_beacon() at all you will
see the same behavior I see.
>> I would rather fix the problem at the two levels, so even if the
>> firmware passes the corrupt frames correctly, the driver would still
>> somewhat work when there's no beacon frames at all.
>
> Like I said before - trying to work with an AP without beacons at all is
> really bad, we shouldn't be doing it.
Why not? For all intents and purposes my system is not receiving any
beacons, and I don't see any problems.
What would you prefer? That nothing works at all?
> We might not properly react to
> radar events, and other things, for example.
So? I don't know what that means, but it can't be worst than not being
able to connect to the Internet whatsoever at all.
Cheers.
--
Felipe Contreras
^ permalink raw reply
* Re: [PATCH v2] mac80211: add assoc beacon timeout logic
From: Felipe Contreras @ 2013-11-11 16:56 UTC (permalink / raw)
To: Johannes Berg
Cc: linux-wireless Mailing List, netdev, John W. Linville,
David S. Miller
In-Reply-To: <CAMP44s3nkj2wZYCPve7EhyeD7sEFmpUN=-h015dOP68PcNSTug@mail.gmail.com>
On Mon, Nov 11, 2013 at 10:53 AM, Felipe Contreras
<felipe.contreras@gmail.com> wrote:
> On Mon, Nov 11, 2013 at 10:41 AM, Johannes Berg
> <johannes@sipsolutions.net> wrote:
>> On Mon, 2013-11-11 at 10:23 -0600, Felipe Contreras wrote:
>>> > This isn't really true like I said above - the kernel can only drop the
>>> > association, if userspace *insists* then it will try again and again.
>>>
>>> But it's not doing this:
>>>
>>> ieee80211_destroy_assoc_data(sdata, false);
>>> cfg80211_assoc_timeout(sdata->dev, bss);
>>>
>>> Which is what causes the association to stop for me.
>>>
>>> So where exactly in the code is the association being "dropped"?
>>
>> This does get called in my setup.
>
> Yes, because your setup is receiving beacons.
>
> Check the code:
>
> if ((ifmgd->assoc_data->need_beacon && !ifmgd->have_beacon) ||
> ieee80211_do_assoc(sdata)) {
> struct cfg80211_bss *bss = ifmgd->assoc_data->bss;
>
> ieee80211_destroy_assoc_data(sdata, false);
> cfg80211_assoc_timeout(sdata->dev, bss);
> }
>
> If there's no beacon, cfg80211_assoc_timeout() is not called.
>
> I'm sure if you don't call ieee80211_rx_mgmt_beacon() at all you will
> see the same behavior I see.
My bad, actually the code that is not being called is:
cfg80211_unlink_bss(local->hw.wiphy, assoc_data->bss);
In ieee80211_do_assoc().
--
Felipe Contreras
^ permalink raw reply
* Re: [PATCH v2] mac80211: add assoc beacon timeout logic
From: Johannes Berg @ 2013-11-11 17:00 UTC (permalink / raw)
To: Felipe Contreras
Cc: linux-wireless Mailing List, netdev, John W. Linville,
David S. Miller
In-Reply-To: <CAMP44s3nkj2wZYCPve7EhyeD7sEFmpUN=-h015dOP68PcNSTug@mail.gmail.com>
On Mon, 2013-11-11 at 10:53 -0600, Felipe Contreras wrote:
> > I see the same behaviour - but it's the supplicant's doing, it is indeed
> > getting the event that the AP connection failed (timed out):
> >
> > wlan0: Event ASSOC_TIMED_OUT (15) received
>
> Not in my setup.
Well, dunno then. Different kernel versions? This clearly happens for
me.
> >> > This isn't really true like I said above - the kernel can only drop the
> >> > association, if userspace *insists* then it will try again and again.
> >>
> >> But it's not doing this:
> >>
> >> ieee80211_destroy_assoc_data(sdata, false);
> >> cfg80211_assoc_timeout(sdata->dev, bss);
> >>
> >> Which is what causes the association to stop for me.
> >>
> >> So where exactly in the code is the association being "dropped"?
> >
> > This does get called in my setup.
>
> Yes, because your setup is receiving beacons.
No ... I tested on hwsim, making it ask for dtim-before-assoc, and
short-circuiting the beacon-TX routing. It can't have been seeing
beacons.
> Check the code:
>
> if ((ifmgd->assoc_data->need_beacon && !ifmgd->have_beacon) ||
> ieee80211_do_assoc(sdata)) {
> struct cfg80211_bss *bss = ifmgd->assoc_data->bss;
>
> ieee80211_destroy_assoc_data(sdata, false);
> cfg80211_assoc_timeout(sdata->dev, bss);
> }
>
> If there's no beacon, cfg80211_assoc_timeout() is not called.
Yes it is.
"need_beacon && !have_beacon:
means - I wanted the beacon but didn't get it at the timeout.
> I'm sure if you don't call ieee80211_rx_mgmt_beacon() at all you will
> see the same behavior I see.
I'm sure I won't :)
> > Like I said before - trying to work with an AP without beacons at all is
> > really bad, we shouldn't be doing it.
>
> Why not? For all intents and purposes my system is not receiving any
> beacons, and I don't see any problems.
The not receiving part is a bug. I think you're probably receiving
beacons once associated though?
> What would you prefer? That nothing works at all?
Yes, that'd be much safer.
> > We might not properly react to
> > radar events, and other things, for example.
>
> So? I don't know what that means, but it can't be worst than not being
> able to connect to the Internet whatsoever at all.
It can make you break the law.
johannes
^ permalink raw reply
* Re: [PATCH v2] mac80211: add assoc beacon timeout logic
From: Johannes Berg @ 2013-11-11 17:01 UTC (permalink / raw)
To: Felipe Contreras
Cc: linux-wireless Mailing List, netdev, John W. Linville,
David S. Miller
In-Reply-To: <CAMP44s0Lsd2qMf3a-Vx4=HTd_UxzLVC6q1bAGPr+t6anubNknA@mail.gmail.com>
On Mon, 2013-11-11 at 10:56 -0600, Felipe Contreras wrote:
> On Mon, Nov 11, 2013 at 10:53 AM, Felipe Contreras
> <felipe.contreras@gmail.com> wrote:
> > On Mon, Nov 11, 2013 at 10:41 AM, Johannes Berg
> > <johannes@sipsolutions.net> wrote:
> >> On Mon, 2013-11-11 at 10:23 -0600, Felipe Contreras wrote:
>
> >>> > This isn't really true like I said above - the kernel can only drop the
> >>> > association, if userspace *insists* then it will try again and again.
> >>>
> >>> But it's not doing this:
> >>>
> >>> ieee80211_destroy_assoc_data(sdata, false);
> >>> cfg80211_assoc_timeout(sdata->dev, bss);
> >>>
> >>> Which is what causes the association to stop for me.
> >>>
> >>> So where exactly in the code is the association being "dropped"?
> >>
> >> This does get called in my setup.
> >
> > Yes, because your setup is receiving beacons.
> >
> > Check the code:
> >
> > if ((ifmgd->assoc_data->need_beacon && !ifmgd->have_beacon) ||
> > ieee80211_do_assoc(sdata)) {
> > struct cfg80211_bss *bss = ifmgd->assoc_data->bss;
> >
> > ieee80211_destroy_assoc_data(sdata, false);
> > cfg80211_assoc_timeout(sdata->dev, bss);
> > }
> >
> > If there's no beacon, cfg80211_assoc_timeout() is not called.
> >
> > I'm sure if you don't call ieee80211_rx_mgmt_beacon() at all you will
> > see the same behavior I see.
>
> My bad, actually the code that is not being called is:
>
> cfg80211_unlink_bss(local->hw.wiphy, assoc_data->bss);
>
> In ieee80211_do_assoc().
That's not really interesting though, it just deletes the scan entry. If
it was deleted, then the supplicant would just scan again and probably
retry the connection.
johannes
^ permalink raw reply
* RE: [PATCH] usb: xhci: Link TRB must not occur with a USB payload burst.
From: David Laight @ 2013-11-11 17:12 UTC (permalink / raw)
To: Alan Stern
Cc: Sarah Sharp, netdev-u79uwXL29TY76Z2rM5mHXA,
linux-usb-u79uwXL29TY76Z2rM5mHXA
In-Reply-To: <Pine.LNX.4.44L0.1311081303460.1162-100000-IYeN2dnnYyZXsRXLowluHWD2FQJk+8+b@public.gmane.org>
> From: Alan Stern [mailto:stern-nwvwT67g6+6dFdvTe/nMLpVzexx5G7lz@public.gmane.org]
> On Fri, 8 Nov 2013, David Laight wrote:
>
...
> GET_MAX_PACKET always returns MaxPacketSize, and for USB-3 bulk
> endpoints, MaxPacketSize is always 1024. MaxBurstSize can be anything
> from 1 to 16.
I've just read something that explained about bursts.
> > > According to my version of the spec (Rev 1.0, dated 5/21/2010), if a TD
> > > is larger than the MBP and its length isn't a multiple of the MBP, then
> > > the last MBP boundary in the TD must also be a TRB boundary. This
> > > follows from two statements in section 4.11.7.1:
> > >
> > > If the TD Transfer Size is an even multiple of the MBP then all
> > > TD Fragments shall define exact multiples of MBP data bytes.
> > > If not, then only the last TD Fragment shall define less than
> > > MBP data (or the Residue) bytes.
> >
> > No, that doesn't stop there being only 1 TD fragment.
> > (I think it means exact multiple of the MBP.)
>
> Suppose, for example, the MBP is 1024. If you have a TD with length
> 1500, and if it had only one fragment, the last (and only) fragment's
> length would not less than the MBP and it would not be an exact
> multiple of the MBP.
That doesn't matter - eg example 2 in figure 25
> I agree that the text is not as clear as it should be.
Reading it all again makes me think that a LINK trb is only
allowed on the burst boundary (which might be 16k bytes).
The only real way to implement that is to ensure that TD never
contain LINK TRB.
There are other things that can be done at a TD fragment boundary
(or only once per TD fragment) - but the code doesn't attempt any
of them.
The restriction might be there to simplify the hardware to
retransmit as TRB burst.
My USB3 ethernet card (ax88179_178a driver) is running a lot more
reliably with these changes.
David
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" 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
* Re: [net-next PATCH v4 0/2] l2 hardware accelerated macvlans
From: John Fastabend @ 2013-11-11 17:16 UTC (permalink / raw)
To: vyasevic
Cc: nhorman, davem, alexander.h.duyck, andy, netdev,
jeffrey.t.kirsher, vyasevich
In-Reply-To: <527C662B.4080300@redhat.com>
On 11/7/2013 8:18 PM, Vlad Yasevich wrote:
[...]
>> If folks find this series acceptable there are a few
>> items we can work on next. First broadcast and multicast
>> will use the hardware even for local traffic with this
>> series. It would be best (I think) to use the software
>> path for macvlan to macvlan traffic and save the PCIe
>> bus. This depends on how much you value CPU time vs
>> PCIE bandwidth. This will need another patch series
>> to flush out.
>>
>
> John
>
> So, I've been looking at these patches and the more I
> look the more I wonder how much benefit there is in
> sending unicast macvlan<->macvlan traffic through the hw.
> It looks like any bulk transfers would have to undergo
> tso segmentaion and GRO, where as this can be completely
> bypassed right now with software.
>
> Did you run any numbers?
>
I have run some numbers on previous versions of the code for
this. I'll take some new ones with the final code today/tomorrow
and post them.
Just wanted to let you know I saw this...
Thanks,
John
> Thanks
> -vlad
>
>
^ permalink raw reply
* Re: TCP performance regression
From: Eric Dumazet @ 2013-11-11 17:38 UTC (permalink / raw)
To: Felix Fietkau; +Cc: Sujith Manoharan, netdev, Dave Taht
In-Reply-To: <52810800.9020402@openwrt.org>
On Mon, 2013-11-11 at 17:38 +0100, Felix Fietkau wrote:
> On 2013-11-11 17:13, Sujith Manoharan wrote:
> > Eric Dumazet wrote:
> >> We have many choices.
> >>
> >> 1) Add back a minimum of ~128 K of outstanding bytes per TCP session,
> >> so that buggy drivers can sustain 'line rate'.
> >>
> >> Note that with 100 concurrent TCP streams, total amount of bytes
> >> queued on the NIC is 12 MB.
> >> And pfifo_fast qdisc will drop packets anyway.
> >>
> >> Thats what we call 'BufferBloat'
> >>
> >> 2) Try lower values like 64K. Still bufferbloat.
> >>
> >> 3) Fix buggy drivers, using a proper logic, or shorter timers (mvneta
> >> case for example)
> >>
> >> 4) Add a new netdev attribute, so that well behaving NIC drivers do not
> >> have to artificially force TCP stack to queue too many bytes in
> >> Qdisc/NIC queues.
> >
> > I think the quirks of 802.11 aggregation should be taken into account.
> > I am adding Felix to this thread, who would have more to say on latency/bufferbloat
> > with wireless drivers.
> I don't think this issue is about something as simple as timer handling
> for tx completion (or even broken/buggy drivers).
>
> There's simply no way to make 802.11 aggregation work well and have
> similar tx completion latency characteristics as Ethernet devices.
>
> 802.11 aggregation reduces the per-packet airtime overhead by combining
> multiple packets into one transmission (saving a lot of time getting a
> tx opportunity, transmitting the PHY header, etc.), which makes the
> 'line rate' heavily depend on the amount of buffering.
How long a TX packet is put on hold hoping a following packet will
come ?
> Aggregating multiple packets into one transmission also causes extra
> packet loss, which is compensated by retransmission and reordering, thus
> introducing additional latency.
>
> I don't think that TSQ can do a decent job of mitigating bufferbloat on
> 802.11n devices without a significant performance hit, so adding a new
> netdev attribute might be a good idea.
The netdev attribute would work, but might not work well if using a
tunnel...
^ permalink raw reply
* RE: TCP performance regression
From: Eric Dumazet @ 2013-11-11 17:41 UTC (permalink / raw)
To: David Laight; +Cc: Sujith Manoharan, netdev, Dave Taht
In-Reply-To: <AE90C24D6B3A694183C094C60CF0A2F6026B73FD@saturn3.aculab.com>
On Mon, 2013-11-11 at 16:35 +0000, David Laight wrote:
> > > It should be ok if the mac driver only gives the hardware a small
> > > number of bytes/packets - or one appropriate for the link speed.
> >
> > There is some confusion here.
> >
> > mvneta has a TX ring buffer, which can hold up to 532 TX descriptors.
> >
> > If this driver used skb_orphan(), a single TCP flow could use the whole
> > TX ring.
> >
> > TCP Small Queue would only limit the number of skbs on Qdisc.
> >
> > Try then to send a ping message, it will have to wait a lot.
>
> 532 is a ridiculously large number especially for a slow interface.
> At a guess you don't want more than 10-20ms of data in the tx ring.
> You might need extra descriptors for badly fragmented packets.
Thats why we invented BQL.
Problem is most driver authors don't care of the problem.
They already have hard time to make bug free drivers.
BQL is adding pressure and expose long standing bugs.
Some drivers have large TX rings to lower race probabilities.
^ permalink raw reply
* Re: TCP performance regression
From: Felix Fietkau @ 2013-11-11 17:44 UTC (permalink / raw)
To: Eric Dumazet; +Cc: Sujith Manoharan, netdev, Dave Taht
In-Reply-To: <1384191515.16391.49.camel@edumazet-glaptop2.roam.corp.google.com>
On 2013-11-11 18:38, Eric Dumazet wrote:
> On Mon, 2013-11-11 at 17:38 +0100, Felix Fietkau wrote:
>> I don't think this issue is about something as simple as timer handling
>> for tx completion (or even broken/buggy drivers).
>>
>> There's simply no way to make 802.11 aggregation work well and have
>> similar tx completion latency characteristics as Ethernet devices.
>>
>> 802.11 aggregation reduces the per-packet airtime overhead by combining
>> multiple packets into one transmission (saving a lot of time getting a
>> tx opportunity, transmitting the PHY header, etc.), which makes the
>> 'line rate' heavily depend on the amount of buffering.
>
> How long a TX packet is put on hold hoping a following packet will
> come ?
TX packets in the aggregation queue are held as long as the hardware
queue holds two A-MPDUs (each of which can contain up to 32 packets).
If the aggregation queues are empty and the hardware queue is not full,
the next tx packet from the network stack is pushed to the hardware
queue immediately.
- Felix
^ permalink raw reply
* Re: TCP performance regression
From: Dave Taht @ 2013-11-11 18:03 UTC (permalink / raw)
To: Eric Dumazet; +Cc: Felix Fietkau, Sujith Manoharan, netdev@vger.kernel.org
In-Reply-To: <1384191515.16391.49.camel@edumazet-glaptop2.roam.corp.google.com>
On Mon, Nov 11, 2013 at 9:38 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> On Mon, 2013-11-11 at 17:38 +0100, Felix Fietkau wrote:
>> On 2013-11-11 17:13, Sujith Manoharan wrote:
>> > Eric Dumazet wrote:
>> >> We have many choices.
>> >>
>> >> 1) Add back a minimum of ~128 K of outstanding bytes per TCP session,
>> >> so that buggy drivers can sustain 'line rate'.
>> >>
>> >> Note that with 100 concurrent TCP streams, total amount of bytes
>> >> queued on the NIC is 12 MB.
>> >> And pfifo_fast qdisc will drop packets anyway.
>> >>
>> >> Thats what we call 'BufferBloat'
>> >>
>> >> 2) Try lower values like 64K. Still bufferbloat.
>> >>
>> >> 3) Fix buggy drivers, using a proper logic, or shorter timers (mvneta
>> >> case for example)
>> >>
>> >> 4) Add a new netdev attribute, so that well behaving NIC drivers do not
>> >> have to artificially force TCP stack to queue too many bytes in
>> >> Qdisc/NIC queues.
>> >
>> > I think the quirks of 802.11 aggregation should be taken into account.
>> > I am adding Felix to this thread, who would have more to say on latency/bufferbloat
>> > with wireless drivers.
As I just got dropped in the middle of this convo, I tend to think
that the mac80211 questions is should be handled in it's own thread as
this conversation seemed to be about a certain ethernet driver's
flaws.
>> I don't think this issue is about something as simple as timer handling
>> for tx completion (or even broken/buggy drivers).
>>
>> There's simply no way to make 802.11 aggregation work well and have
>> similar tx completion latency characteristics as Ethernet devices.
I don't quite share all of felix's pessimism. It will tend to be
burstier, yes, but I felt that would not look that much different than
napi -> BQL.
>> 802.11 aggregation reduces the per-packet airtime overhead by combining
>> multiple packets into one transmission (saving a lot of time getting a
>> tx opportunity, transmitting the PHY header, etc.), which makes the
>> 'line rate' heavily depend on the amount of buffering.
making aggregation work well is key to fixing wifi worldwide.
Presently aggregation performance is pretty universally terrible under
real loads and tcp.
(looking further ahead, getting multi-user mimo to work in 802.11ac
would also be helpful but I'm not even sure the IEEE figured that out
yet. Ath10k hw2 do it?)
> How long a TX packet is put on hold hoping a following packet will
> come ?
>
>
>
>> Aggregating multiple packets into one transmission also causes extra
>> packet loss, which is compensated by retransmission and reordering, thus
>> introducing additional latency.
I was extremely encouraged by Yucheng's presentation at ietf on some
vast improvements on managing re-ordering problems. I daydreamed that
it would become possible to eliminate the reorder buffer in lower
levels of the wireless stack(s?).
See slides and fantasize:
http://www.ietf.org/proceedings/88/slides/slides-88-iccrg-6.pdf
The rest of the preso was good, too.
I also thought the new pacing stuff would cause trouble in wifi and aggregation.
>> I don't think that TSQ can do a decent job of mitigating bufferbloat on
>> 802.11n devices without a significant performance hit, so adding a new
>> netdev attribute might be a good idea.
I am not sure which part of what subsystem(s) is really under debate
here. TSQ limits the number of packets that can be outstanding in a
stream. The characteristics of a wifi connection (EDCA scheduling and
aggregated batching) play merry hell with TCP assumptions. The recent
work on fixing TSO offloads shows what can happen if that underlying
set of assumptions is fixed.
My overall take on this, tho, is to take the latest bits of TSQ and
"fq" code, and go measure the effect on wifi stations rather than
discuss what layer is busted or what options need to be added to
netdev. Has anyone done that? I've been busy with 3.10.x
Personally I don't have much of a problem if TSQ hurts single stream
TCP throughput on wifi. I would vastly prefer aggregation to work
better for multiple streams with vastly smaller buffers than it does.
That would be a bigger win, overall.
> The netdev attribute would work, but might not work well if using a
> tunnel...
I am going to make some coffee and catch up. Please excuse whatever
noise I just introduced.
>
>
>
--
Dave Täht
Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html
^ permalink raw reply
* Re: [PATCH v2] mac80211: add assoc beacon timeout logic
From: Felipe Contreras @ 2013-11-11 18:06 UTC (permalink / raw)
To: Johannes Berg
Cc: linux-wireless Mailing List, netdev, John W. Linville,
David S. Miller
In-Reply-To: <1384189227.14334.48.camel-8Nb76shvtaUJvtFkdXX2HixXY32XiHfO@public.gmane.org>
On Mon, Nov 11, 2013 at 11:00 AM, Johannes Berg
<johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> wrote:
> On Mon, 2013-11-11 at 10:53 -0600, Felipe Contreras wrote:
>> > Like I said before - trying to work with an AP without beacons at all is
>> > really bad, we shouldn't be doing it.
>>
>> Why not? For all intents and purposes my system is not receiving any
>> beacons, and I don't see any problems.
>
> The not receiving part is a bug. I think you're probably receiving
> beacons once associated though?
Nope. Never.
>> What would you prefer? That nothing works at all?
>
> Yes, that'd be much safer.
How exactly?
>> > We might not properly react to
>> > radar events, and other things, for example.
>>
>> So? I don't know what that means, but it can't be worst than not being
>> able to connect to the Internet whatsoever at all.
>
> It can make you break the law.
How?
I'm reading this document[1], and if that's what you are referring to,
then for starters it only applies to the master mode, my patch changes
the behavior only on station mode.
Moreover, if continuing the association without beacons has a legal a
problem, that problem would exist for drivers that don't have the
IEEE80211_HW_NEED_DTIM_BEFORE_ASSOC flag, wouldn't it? How exactly
would trying to associate with need_beacon break the law, but not if
!need_beacon?
[1] http://wireless.kernel.org/en/developers/DFS
--
Felipe Contreras
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" 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
* Re: TCP performance regression
From: Dave Taht @ 2013-11-11 18:31 UTC (permalink / raw)
To: Eric Dumazet
Cc: Felix Fietkau, Sujith Manoharan, netdev@vger.kernel.org,
Avery Pennarun
In-Reply-To: <CAA93jw56T32c7NN6Ttnbda6v5mciQhdhD50nC_SoXW8nFti+8g@mail.gmail.com>
Ah, this thread started with a huge regression in ath10k performance
with the new TSQ stuff, and isn't actually about a two line fix to the
mv ethernet driver.
http://comments.gmane.org/gmane.linux.network/290269
I suddenly care a lot more. And I'll care a lot, lot, lot more, if
someone can post a rrul test for before and after the new fq scheduler
and tsq change on this driver on this hardware... What, if anything,
in terms of improvements or regressions, happened to multi-stream
throughput and latency?
https://github.com/tohojo/netperf-wrapper
^ permalink raw reply
* Re: TCP performance regression
From: Sujith Manoharan @ 2013-11-11 18:29 UTC (permalink / raw)
To: Dave Taht; +Cc: Eric Dumazet, Felix Fietkau, netdev@vger.kernel.org
In-Reply-To: <CAA93jw56T32c7NN6Ttnbda6v5mciQhdhD50nC_SoXW8nFti+8g@mail.gmail.com>
Dave Taht wrote:
> Personally I don't have much of a problem if TSQ hurts single stream
> TCP throughput on wifi. I would vastly prefer aggregation to work
> better for multiple streams with vastly smaller buffers than it does.
> That would be a bigger win, overall.
ath9k doesn't hold very deep queues for aggregated traffic. A maximum
of 128 packets can be buffered for each Access Class queue and still
good throughput is obtained, even for 3x3 scenarios.
A loss of almost 50% throughput is seen in 1x1 setups and the penalty
becomes higher with more streams. I don't think such a big loss
in performance is acceptable to achieve low latency.
Sujith
^ permalink raw reply
* Re: [patch net-next RFC 0/2] ipv6: allow temporary address management for user-created addresses
From: David Miller @ 2013-11-11 18:42 UTC (permalink / raw)
To: jiri
Cc: netdev, kuznet, jmorris, yoshfuji, kaber, thaller, stephen,
hannes, vyasevich, dcbw
In-Reply-To: <1384181359-23199-1-git-send-email-jiri@resnulli.us>
I'm not really going to read patches not properly at least CC:'d to
netdev, sorry Jiri.
^ permalink raw reply
* gso: Handle new frag_list of frags GRO packets
From: Herbert Xu @ 2013-11-11 18:52 UTC (permalink / raw)
To: Eric Dumazet
Cc: Ben Hutchings, David Miller, christoph.paasch, netdev, hkchu,
mwdalton
In-Reply-To: <20131107070647.GB31638@gondor.apana.org.au>
Recently GRO started generating packets with frag_lists of frags.
This was not handled by GSO, thus leading to a crash.
Thankfully these packets are of a regular form and are easy to
handle. This patch handles them in two ways. For completely
non-linear frag_list entries, we simply continue to iterate over
the frag_list frags once we exhaust the normal frags. For frag_list
entries with linear parts, we call pskb_trim on the first part
of the frag_list skb, and then process the rest of the frags in
the usual way.
This patch also kills a chunk of dead frag_list code that has
obviously never ever been run since it ends up generating a bogus
GSO-segmented packet with a frag_list entry.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 3735fad..557e1a5 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -2776,6 +2776,7 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
struct sk_buff *segs = NULL;
struct sk_buff *tail = NULL;
struct sk_buff *fskb = skb_shinfo(skb)->frag_list;
+ skb_frag_t *skb_frag = skb_shinfo(skb)->frags;
unsigned int mss = skb_shinfo(skb)->gso_size;
unsigned int doffset = skb->data - skb_mac_header(skb);
unsigned int offset = doffset;
@@ -2815,16 +2816,38 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
if (hsize > len || !sg)
hsize = len;
- if (!hsize && i >= nfrags) {
- BUG_ON(fskb->len != len);
+ if (!hsize && i >= nfrags && skb_headlen(fskb) &&
+ (skb_headlen(fskb) == len || sg)) {
+ BUG_ON(skb_headlen(fskb) > len);
+
+ i = 0;
+ nfrags = skb_shinfo(fskb)->nr_frags;
+ skb_frag = skb_shinfo(fskb)->frags;
+ pos += skb_headlen(fskb);
+
+ while (pos < offset + len) {
+ BUG_ON(i >= nfrags);
+
+ size = skb_frag_size(skb_frag);
+ if (pos + size > offset + len)
+ break;
+
+ i++;
+ pos += size;
+ skb_frag++;
+ }
- pos += len;
nskb = skb_clone(fskb, GFP_ATOMIC);
fskb = fskb->next;
if (unlikely(!nskb))
goto err;
+ if (unlikely(pskb_trim(nskb, len))) {
+ kfree_skb(nskb);
+ goto err;
+ }
+
hsize = skb_end_offset(nskb);
if (skb_cow_head(nskb, doffset + headroom)) {
kfree_skb(nskb);
@@ -2861,7 +2884,7 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
nskb->data - tnl_hlen,
doffset + tnl_hlen);
- if (fskb != skb_shinfo(skb)->frag_list)
+ if (nskb->len == len + doffset)
goto perform_csum_check;
if (!sg) {
@@ -2879,8 +2902,28 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
skb_shinfo(nskb)->tx_flags = skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
- while (pos < offset + len && i < nfrags) {
- *frag = skb_shinfo(skb)->frags[i];
+ while (pos < offset + len) {
+ if (i >= nfrags) {
+ BUG_ON(skb_headlen(fskb));
+
+ i = 0;
+ nfrags = skb_shinfo(fskb)->nr_frags;
+ skb_frag = skb_shinfo(fskb)->frags;
+
+ BUG_ON(!nfrags);
+
+ fskb = fskb->next;
+ }
+
+ if (unlikely(skb_shinfo(nskb)->nr_frags >=
+ MAX_SKB_FRAGS)) {
+ net_warn_ratelimited(
+ "skb_segment: too many frags: %u %u\n",
+ pos, mss);
+ goto err;
+ }
+
+ *frag = *skb_frag;
__skb_frag_ref(frag);
size = skb_frag_size(frag);
@@ -2893,6 +2936,7 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
if (pos + size <= offset + len) {
i++;
+ skb_frag++;
pos += size;
} else {
skb_frag_size_sub(frag, pos + size - (offset + len));
@@ -2902,25 +2946,6 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
frag++;
}
- if (pos < offset + len) {
- struct sk_buff *fskb2 = fskb;
-
- BUG_ON(pos + fskb->len != offset + len);
-
- pos += fskb->len;
- fskb = fskb->next;
-
- if (fskb2->next) {
- fskb2 = skb_clone(fskb2, GFP_ATOMIC);
- if (!fskb2)
- goto err;
- } else
- skb_get(fskb2);
-
- SKB_FRAG_ASSERT(nskb);
- skb_shinfo(nskb)->frag_list = fskb2;
- }
-
skip_fraglist:
nskb->data_len = len - hsize;
nskb->len += nskb->data_len;
Thanks,
--
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
^ permalink raw reply related
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