On Mon, Nov 09, 2015 at 11:32:07AM +1100, Benjamin Herrenschmidt wrote:
>On Mon, 2015-11-09 at 11:10 +1100, Gavin Shan wrote:
>>
>> -if (likely(netif_running(netdev))) {
>> +/* When running in NCSI mode, the interface should be
>> + * ready to receive or transmit NCSI packet before it'
zero addr before calling copy_to_user()
Signed-off-by: Loganaden Velvindron
---
net/ipv4/ip_sockglue.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index c3c359a..d7a5a8b 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -
On Sun, Nov 8, 2015 at 7:31 PM, Dexuan Cui wrote:
>> From: David Miller [mailto:da...@davemloft.net]
>> Sent: Monday, November 9, 2015 11:24
>> ...
>> > Thanks, David!
>> > I understand 1 TX queue is the bottleneck (however in Simon's
>> > test, TX=1 => 36.7Gb/s, TX=8 => 37.7 Gb/s, so it looks the
Hi all,
I’ve been trying to create a bridge like driver for a while. The
project is running into a wall due to the inability of seeing either
netdevice_ops structure or hard_start_xmit function from my device
driver in both Intel ixgbe and e1000e drivers.
Basically, I have a structure preserving
> From: David Miller [mailto:da...@davemloft.net]
> Sent: Monday, November 9, 2015 11:24
> ...
> > Thanks, David!
> > I understand 1 TX queue is the bottleneck (however in Simon's
> > test, TX=1 => 36.7Gb/s, TX=8 => 37.7 Gb/s, so it looks the TX=1 bottleneck
> > is not so obvious).
> > I'm just won
On 9 November 2015 at 13:23, David Miller wrote:
> From: Dexuan Cui
> Date: Mon, 9 Nov 2015 03:11:35 +
>
>>> -Original Message-
>>> From: David Miller [mailto:da...@davemloft.net]
>>> Sent: Monday, November 9, 2015 10:53
>>> To: Dexuan Cui
>>> Cc: eric.duma...@gmail.com; d...@cumulus
> -Original Message-
> From: David Miller [mailto:da...@davemloft.net]
> Sent: Monday, November 9, 2015 10:53
> To: Dexuan Cui
> Cc: eric.duma...@gmail.com; d...@cumulusnetworks.com; Simon Xiao
> ; netdev@vger.kernel.org; Haiyang Zhang
> ; linux-ker...@vger.kernel.org;
> de...@linuxdriverp
From: Dexuan Cui
Date: Mon, 9 Nov 2015 03:11:35 +
>> -Original Message-
>> From: David Miller [mailto:da...@davemloft.net]
>> Sent: Monday, November 9, 2015 10:53
>> To: Dexuan Cui
>> Cc: eric.duma...@gmail.com; d...@cumulusnetworks.com; Simon Xiao
>> ; netdev@vger.kernel.org; Haiyan
> Once it figures out which gso_inner_segment to use, it calls
> __skb_udp_tunnel_segment with it, which then does some curious header
> calculations on various lengths (that I need to read carefully), and
> then proceeds to split the segments using our gso_inner_segment
> function of choice, and t
From: Daniel Borkmann
Date: Sun, 08 Nov 2015 01:33:56 +0100
> Hmm, yeah, on a (only quick) look, it seems this is mostly needed for
> the
> virtio_net related code in packet_snd() / packet_recvmsg(), not
> handled in
> RX/TX ring paths actually.
>
> $ git grep -n gso_size net/packet/
> net/packe
> From: devel [mailto:driverdev-devel-boun...@linuxdriverproject.org] On Behalf
> Of Eric Dumazet
> Sent: Sunday, November 8, 2015 3:36
> To: David Ahern
> Cc: netdev@vger.kernel.org; Haiyang Zhang ; linux-
> ker...@vger.kernel.org; de...@linuxdriverproject.org; David Miller
>
> Subject: Re: linu
From: Dexuan Cui
Date: Mon, 9 Nov 2015 02:39:24 +
>> Throughput on a single TCP flow for a 40G NIC can be tricky to tune.
> Why is a single TCP flow trickier than multiple TCP flows?
> IMO it should be easier to analyze the issue of a single TCP flow?
Because a single TCP flow can only use o
On Mon, Nov 9, 2015 at 2:40 AM, Herbert Xu wrote:
> You're right. I don't think the ordering matters.
Cool, so we're on the same page then.
In that case, any ideas about constructing UDP super-packets for GSO?
As Maciej pointed out, UFO is actually just IP fragmentation and UDP
checksums, but d
From: Iyappan Subramanian
Date: Sat, 7 Nov 2015 11:50:40 -0800
> This patch fixes the RGMII 10/100M mode by reprogramming the clock.
>
> Signed-off-by: Iyappan Subramanian
> Tested-by: Fushen Chen
Applied, thanks.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the bo
From: Eric Dumazet
Date: Sun, 8 Nov 2015 10:54:06 -0800
> Many contexts need to reach listener socket from skb attached
> to a request socket. This patch series add skb_to_full_sk() to
> clearly express this need and use it where appropriate.
Series applied, thanks Eric.
--
To unsubscribe from
From: Samuel Ortiz
Date: Mon, 9 Nov 2015 02:02:31 +0100
> This is the 1st NFC fixes pull request for 4.4.
>
> It includes bug fixes and one fix for a build failure, all of them
> introduced with the first NFC pull request for 4.4.
>
> We have:
>
> - Fix nfcmrvl SPI driver potential build error
On Sun, Nov 08, 2015 at 03:57:24PM +0100, Jason A. Donenfeld wrote:
> On Sun, Nov 8, 2015 at 11:57 AM, Herbert Xu
> wrote:
> > UDP carries no ordering information so this doesn't work.
>
> But if there's no ordering information, what's the problem? Isn't it
> good enough to send the packets in t
Hi David,
This is the 1st NFC fixes pull request for 4.4.
It includes bug fixes and one fix for a build failure, all of them
introduced with the first NFC pull request for 4.4.
We have:
- Fix nfcmrvl SPI driver potential build error due to a broken Kconfig
dependency.
- A few fixes for the fi
On Mon, Nov 09, 2015 at 11:36:10AM +1100, Benjamin Herrenschmidt wrote:
>On Mon, 2015-11-09 at 11:10 +1100, Gavin Shan wrote:
>> This enables IP/UDP/TCP offload checksum according to information
>> passed on from bootloader through device-tree. The offload doesn't
>> working properly when the inter
On Mon, Nov 09, 2015 at 11:30:35AM +1100, Benjamin Herrenschmidt wrote:
>On Mon, 2015-11-09 at 11:10 +1100, Gavin Shan wrote:
>> This replaces variable name "use_nc_si" with "use_ncsi". No logical
>> changes introduced.
>
>This driver patch is against an internal tree, not upstream. Please
>ignore.
On Mon, 2015-11-09 at 11:10 +1100, Gavin Shan wrote:
> +
> + /* Send NCSI packet */
> + skb_get(nr->nr_cmd);
> + ret = dev_queue_xmit_sk(NULL, nr->nr_cmd);
> + if (ret)
> + goto out;
> +
Probably just dev_queue_xmit() in recent kernels
This means the queue ne
On Mon, 2015-11-09 at 11:10 +1100, Gavin Shan wrote:
> This enables IP/UDP/TCP offload checksum according to information
> passed on from bootloader through device-tree. The offload doesn't
> working properly when the interface works in NCSI mode.
>
> Signed-off-by: Gavin Shan
> ---
> drivers/ne
On Mon, 2015-11-09 at 11:10 +1100, Gavin Shan wrote:
>
> - if (likely(netif_running(netdev))) {
> + /* When running in NCSI mode, the interface should be
> + * ready to receive or transmit NCSI packet before it's
> + * opened.
> + */
No, that's not right. open/close is when
On Mon, 2015-11-09 at 11:10 +1100, Gavin Shan wrote:
> This replaces variable name "use_nc_si" with "use_ncsi". No logical
> changes introduced.
This driver patch is against an internal tree, not upstream. Please
ignore. The upstream driver doesn't have any concept of "nc_si".
Gavin, next time yo
On Tue, Oct 06, 2015 at 02:09:42PM +1100, Gavin Shan wrote:
Please ignore this series. A new RFC patchset is going to be sent with more
improved
as I was suggested:
* Explaning what's NCSI and how the code is designed.
* Dropped netlink interface so that we can focus on NCSI stack at current
NCSI packet encapsulated in ethernet frame is recognized by the
protocol field (0x88F8) of the ethernet frame header. According
to NCSI spec, the destination MAC field should be 0xFF's, but the
source MAC field could be arbitrary one since channel field in NCSI
header indicates the originator of th
NCSI spec (DSP0222) defined several objects: package, channel,
mode, filter, version and statistics etc. This defines data
structrues to represent those objects and exports interfaces
(functions) to manage them:
* The user (e.g. netdev driver) dereference NCSI device by
"struct ncsi_dev",
This replaces variable name "use_nc_si" with "use_ncsi". No logical
changes introduced.
Signed-off-by: Gavin Shan
---
drivers/net/ethernet/faraday/ftgmac100.c | 14 +++---
1 file changed, 7 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/faraday/ftgmac100.c
b/drivers/ne
This series of patches are prototype requesting for comments. Please focus on
the big picture at current stage, but any comments to improve the implementaion
are welcomed.
The following figure gives an example about how NCSI is deployed: The NCSI is
specified by DSP0222, which can be downloaded fr
This enables IP/UDP/TCP offload checksum according to information
passed on from bootloader through device-tree. The offload doesn't
working properly when the interface works in NCSI mode.
Signed-off-by: Gavin Shan
---
drivers/net/ethernet/faraday/ftgmac100.c | 5 -
1 file changed, 4 inserti
To the network interface that can run in NCSI mode on BMC side,
there might have multiple NCSI packages/channels. The available
NCSI packages/channels (topology) needs to be probed and one
channel should be selected as active one from the available
channels before the network interface on BMC side
The NIC has the possibility to connect to NCSI package and channel.
This supports NCSI enabled interface. When the network device is
registered, the accompanying NCSI device is registered. When the
interface is to be brought up, the NCSI device is started to probe
NCSI topology and choose one activ
From: Neil Horman
Date: Sun, 8 Nov 2015 16:56:41 -0500
> All I'm doing is, in effect masking the promote_secondaries sysctl
> for an interface doing a flush operation.
Packets and other entities not controlled by RTNL can see an inconsistent
state: no primary address on the interface.
You can a
On Sat, Nov 7, 2015 at 6:27 PM, Alexei Starovoitov
wrote:
> On Fri, Nov 06, 2015 at 09:36:17PM -0800, Yang Shi wrote:
>> ARM64 JIT used FP (x29) as eBPF fp register, but FP is subjected to
>> change during function call so it may cause the BPF prog stack base address
>> change too. Whenever, it po
Hi Greg,
This series cleans up a messy and poorly documented mechanism required
at tty final close to prevent drivers from crashing after h/w shutdown.
Without special handling, N_TTY echoing will cause driver i/o requests
_after_ h/w shutdown, which typically crashes the driver. Currently,
the t
The tty_port_close_start() helper already flushes the tty and ldisc
buffers on final close; tty drivers which use this helper need not
repeat tty_ldisc_flush().
Signed-off-by: Peter Hurley
---
drivers/char/pcmcia/synclink_cs.c | 3 ---
drivers/tty/amiserial.c | 2 --
drivers/tty/rocket
This driver's private completion variable, close_wait, is no longer
used for wait since "tty: Remove ASYNC_CLOSING checks in open()/hangup";
remove.
Signed-off-by: Peter Hurley
---
drivers/tty/rocket.c | 2 --
drivers/tty/rocket_int.h | 1 -
2 files changed, 3 deletions(-)
diff --git a/driv
tty->closing is exclusively used to cause the N_TTY line discipline
to drop further input on final tty close (except XON/XOFF output flow
control changes). In turn, this prevents the line discipline from
generating new tty driver i/o requests (eg., when echoing) after the tty
driver has performed h
On final port close (and thus final tty close), only output flow
control requests in the input data should be processed. Ignore all
other input data, including parity errors, overruns and breaks.
Signed-off-by: Peter Hurley
---
drivers/tty/n_tty.c | 2 --
1 file changed, 2 deletions(-)
diff --g
On Sat, Nov 07, 2015 at 01:49:25AM +0200, Julian Anastasov wrote:
>
> Hello,
>
> On Fri, 6 Nov 2015, Neil Horman wrote:
>
> > The solution is to recognize that its pointless to promote an address to be
> > a
> > new primary, if there is a possibility that it will just be removed in the
>
Allow for a filter to ignore certain nlmsg_flags.
Signed-off-by: Phil Sutter
---
Changed since v1:
- Fix typo in #define of rtnl_dump_filter().
include/libnetlink.h | 7 ++-
lib/libnetlink.c | 10 ++
2 files changed, 12 insertions(+), 5 deletions(-)
diff --git a/include/libnet
Since it's no longer relevant whether an IP address is primary or
secondary when flushing, ipaddr_flush() can be simplified a bit.
Signed-off-by: Phil Sutter
---
ip/ipaddress.c | 38 +-
1 file changed, 1 insertion(+), 37 deletions(-)
diff --git a/ip/ipaddress
Since it uses only a single filter, rtnl_dump_filter() can be used.
Signed-off-by: Phil Sutter
---
ip/iptoken.c | 6 +-
1 file changed, 1 insertion(+), 5 deletions(-)
diff --git a/ip/iptoken.c b/ip/iptoken.c
index a38194c..428f133 100644
--- a/ip/iptoken.c
+++ b/ip/iptoken.c
@@ -95,10 +95,6
The core issue here is that with promote_secondaries sysctl setting
being turned off, removing the primary address implicitly removes all
secondaries as well. Iproute is aware of this and therefore tries to
remove all secondary addresses first to circumvent errors due to
removing non-existent addre
Right after ipaddr_reset_filter(), filter.family is always AF_UNSPEC.
Signed-off-by: Phil Sutter
---
ip/ipaddress.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/ip/ipaddress.c b/ip/ipaddress.c
index 23e0308..ff7f2a8 100644
--- a/ip/ipaddress.c
+++ b/ip/ipaddress.c
@@ -
Linux version 3.1 introduced a consistency check for netlink dumps in
commit 670dc28 ("netlink: advertise incomplete dumps"). This bites
iproute2 when flushing more addresses than can fit into a single
RTM_GETADDR response. To silence the spurious error message "Dump was
interrupted and may be inco
This patch series aims to silence spurious error messages when flushing
an interface with many addresses assigned and provides a few cleanups
found along the way.
The original issue was a test-case adding 40k IPv4 addresses to an
interface and calling 'ip addr flush' afterwards. Initially, this
re
The core issue here is that with promote_secondaries sysctl setting
being turned off, removing the primary address implicitly removes all
secondaries as well. Iproute is aware of this and therefore tries to
remove all secondary addresses first to circumvent errors due to
removing non-existent addre
Since it's no longer relevant whether an IP address is primary or
secondary when flushing, ipaddr_flush() can be simplified a bit.
Signed-off-by: Phil Sutter
---
ip/ipaddress.c | 38 +-
1 file changed, 1 insertion(+), 37 deletions(-)
diff --git a/ip/ipaddress
This patch series aims to silence spurious error messages when flushing
an interface with many addresses assigned and provides a few cleanups
found along the way.
The original issue was a test-case adding 40k IPv4 addresses to an
interface and calling 'ip addr flush' afterwards. Initially, this
re
Linux version 3.1 introduced a consistency check for netlink dumps in
commit 670dc28 ("netlink: advertise incomplete dumps"). This bites
iproute2 when flushing more addresses than can fit into a single
RTM_GETADDR response. To silence the spurious error message "Dump was
interrupted and may be inco
Allow for a filter to ignore certain nlmsg_flags.
Signed-off-by: Phil Sutter
---
include/libnetlink.h | 7 ++-
lib/libnetlink.c | 10 ++
2 files changed, 12 insertions(+), 5 deletions(-)
diff --git a/include/libnetlink.h b/include/libnetlink.h
index 2280c39..38e54f4 100644
---
Since it uses only a single filter, rtnl_dump_filter() can be used.
Signed-off-by: Phil Sutter
---
ip/iptoken.c | 6 +-
1 file changed, 1 insertion(+), 5 deletions(-)
diff --git a/ip/iptoken.c b/ip/iptoken.c
index a38194c..428f133 100644
--- a/ip/iptoken.c
+++ b/ip/iptoken.c
@@ -95,10 +95,6
Right after ipaddr_reset_filter(), filter.family is always AF_UNSPEC.
Signed-off-by: Phil Sutter
---
ip/ipaddress.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/ip/ipaddress.c b/ip/ipaddress.c
index 23e0308..ff7f2a8 100644
--- a/ip/ipaddress.c
+++ b/ip/ipaddress.c
@@ -
SYNACK packets might be attached to request sockets.
Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of
listener")
Signed-off-by: Eric Dumazet
---
net/netfilter/nft_meta.c | 36
1 file changed, 20 insertions(+), 16 deletions(-)
SYNACK packets might be attached to a request socket,
xt_owner wants to gte the listener in this case.
Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of
listener")
Signed-off-by: Eric Dumazet
---
net/netfilter/xt_owner.c | 6 --
1 file changed, 4 insertions(+),
Generalize selinux_skb_sk() added in commit 212cd0895330
("selinux: fix random read in selinux_ip_postroute_compat()")
so that we can use it other contexts.
Use it right away in selinux_netlbl_skbuff_setsid()
Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of
listene
SYNACK packets might be attached to request sockets.
Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of
listener")
Signed-off-by: Eric Dumazet
---
net/sched/cls_flow.c | 15 +++
1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/net/sched/cls
This module wants to access sk->sk_security, which is not
available for request sockets.
Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of
listener")
Signed-off-by: Eric Dumazet
---
security/smack/smack_netfilter.c | 11 +++
1 file changed, 7 insertions(+),
SYNACK packets might be attached to request sockets.
Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of
listener")
Signed-off-by: Eric Dumazet
---
net/sched/em_meta.c | 138 ++--
1 file changed, 92 insertions(+), 46 de
Many contexts need to reach listener socket from skb attached
to a request socket. This patch series add skb_to_full_sk() to
clearly express this need and use it where appropriate.
Eric Dumazet (6):
net: add skb_to_full_sk() helper and use it in
selinux_netlbl_skbuff_setsid()
smack: use sk
From: Yang Shi
Date: Fri, 6 Nov 2015 21:38:58 -0800
> aarch64 and s390x support eBPF JIT too, correct document to reflect this and
> avoid any confusion.
>
> Signed-off-by: Yang Shi
Applied, thanks.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message
On Sun, Nov 08, 2015 at 02:51:01PM +0100, Gerhard Wiesinger wrote:
> On 25.10.2015 17:29, Greg KH wrote:
> >On Sun, Oct 25, 2015 at 11:48:54AM +0100, Gerhard Wiesinger wrote:
> >>On 25.10.2015 10:46, Willy Tarreau wrote:
> >>>ipset *triggered* the problem. The whole stack dump would tell more.
> >>
On Sun, 2015-11-08 at 11:15 +0100, Dmitry Vyukov wrote:
> Hello,
>
> I've got the following deadlock report on commit
> d1e41ff11941784f469f17795a4d9425c2eb4b7a (Nov 5).
>
>
> [ INFO: possible circular locking dependency detected ]
> 4.3.0+ #39 Not tainted
> -
On Sun, Nov 8, 2015 at 11:57 AM, Herbert Xu wrote:
> UDP carries no ordering information so this doesn't work.
But if there's no ordering information, what's the problem? Isn't it
good enough to send the packets in the order they were sendto()d? Or
in any order at all?
--
To unsubscribe from this
On 25.10.2015 17:29, Greg KH wrote:
On Sun, Oct 25, 2015 at 11:48:54AM +0100, Gerhard Wiesinger wrote:
On 25.10.2015 10:46, Willy Tarreau wrote:
ipset *triggered* the problem. The whole stack dump would tell more.
OK, find the stack traces in the bug report:
https://bugzilla.redhat.com/show_bu
On Sun, Nov 08, 2015 at 11:36:53AM +0100, Jason A. Donenfeld wrote:
>
> Wouldn't there be some significant savings from bundling together
> several UDP packets meant for the same destination, and sending those
> all as one super-packet, so they don't each have to traverse the whole
> networking and
E-mail your CV TO Lee Yang Travel & recruitment service
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Maciej,
On Sun, Nov 8, 2015 at 12:40 AM, Maciej Żenczykowski
wrote:
> This isn't particularly efficient. This is basically equivalent to doing
> GSO before the superpacket reaches your driver (you might get some
> savings by not bothering to look at the packet headers of the second
> and on p
Hello,
I've got the following deadlock report on commit
d1e41ff11941784f469f17795a4d9425c2eb4b7a (Nov 5).
[ INFO: possible circular locking dependency detected ]
4.3.0+ #39 Not tainted
---
syzkaller_execu/18311 is trying to acquire lock:
(rtnl
Hi all,
We have this very disturbing issue on a few of our production servers,
which disconnects VMs
from their network.
* The Vms are part of an oVirt host so each vm is attached to a l2
bridge with a tap device.
* The bridge has an IP on it and is connected via a bond
Issue:
when the ma
71 matches
Mail list logo
