On 06-03-2007 20:19, David Miller wrote:
Wei, I have to revert your change, it is incorrect as pointed
out by other people here on netdev.
BSD sockets basically define the 'backlog' parameter to listen()
to mean allow backlog + 1 connections to be queued to the socket.
This allows a backlog
When system under heavy stress and must allocate new work
instead of reusing old one, new work must use correct
completion callback.
Patch is based on Philipp's and Lars' work.
I only cleaned small stuff (and removed spaces instead of tabs).
Signed-off-by: Philipp Reisner [EMAIL PROTECTED]
Document that equal-cost multipath routing with caching does not work
for forwarded packets.
Signed-Off-By: Tore Anderson [EMAIL PROTECTED]
---
* Jarek Poplawski
It is probably one of the most often asked questions
on the LARTC, so I'd suggest to look at its archives.
[...]
Multipath with
On Wed, Mar 07, 2007 at 01:37:44PM +0100, Tore Anderson wrote:
Document that equal-cost multipath routing with caching does not work
for forwarded packets.
Signed-Off-By: Tore Anderson [EMAIL PROTECTED]
---
* Jarek Poplawski
It is probably one of the most often asked questions
on the
* Jarek Poplawski
I don't feel bothered at all. Your letter could
make an impression there is a lack of help and
google not working, but it's not always true.
So, sometimes people probably wait to see, they
will only help and not do all your work.
Of course. I tried hard to figure it out
Hi,
In the current code SID 0 indicates that the socket is to be un-bound.
That are the semantics used by the kernel code, yes - but even pppd uses
different semantics (which can't quite work, of course ...).
Supporting unbinding of the socket was intended to permit the PPPoE
session to be
On Tue, Mar 06, 2007 at 07:39:21PM -0800, Michael K. Edwards wrote:
On 3/6/07, Ralf Baechle [EMAIL PROTECTED] wrote:
This small change btw. delivers about ~ 3% extra performance on a very
slow test system.
Has this change been tested / benchmarked under VMWare? pcnet32 is
the (default?)
This change can be made; the unbinding behavior can be removed and SID 0
can be made valid. I hope I was clear in my previous e-mail that I
didn't object to this.
PPPoE connections are unstable. Ethernet frames get dropped. Things
die randomly. And yes, you typically want to have a cron job or
Hi,
This change can be made; the unbinding behavior can be removed and SID 0
can be made valid. I hope I was clear in my previous e-mail that I
didn't object to this.
Not quite. But now I think I got it ;-)
PPPoE connections are unstable. Ethernet frames get dropped. Things
die randomly.
On 3/6/07, Ralf Baechle [EMAIL PROTECTED] wrote:
Price question: why would this patch make a difference under VMware? :-)
Moving the struct pcnet32_private from the GFP_DMA32 init_block to the
GFP_KERNEL netdev allocation may be a win even on systems where
GFP_DMA32 is normally cached, because
Adrian Bunk wrote:
Subject: forcedeth: skb_over_panic
References : http://bugzilla.kernel.org/show_bug.cgi?id=8058
Submitter : Albert Hopkins [EMAIL PROTECTED]
Status : unknown
I think this is fixed by the recent forcedeth 3-patch patchset? Can
Albert or others confirm?
On Tue, Mar 06, 2007 at 08:20:52PM -0800, David Miller wrote:
From: Stephen Hemminger [EMAIL PROTECTED]
Date: Tue, 6 Mar 2007 14:47:06 -0800
The Newton-Raphson method is quadratically convergent so
only a small fixed number of steps are necessary.
Therefore it is faster to unroll the
pktgen currently only works on network devices with type ARPHRD_ETHER. Add
support for the loopback device, type ARPHRD_LOOPBACK.
I've tested this on my system, using a modified pktgen.conf-1-1 with
s/eth1/lo/g, and it works fine; the network device statistics confirm packet
transmission and
Sami Farin wrote:
On Tue, Mar 06, 2007 at 23:53:49 +0200, Sami Farin wrote:
...
And I found bug in gcc-4.1.2, it gave 0 for ncubic results
when doing 1000 loops test... gcc-4.0.3 works.
Found it.
--- cbrt-test.c~ 2007-03-07 00:20:54.735248105 +0200
+++ cbrt-test.c 2007-03-07
On Tue, 6 Mar 2007, Baruch Even wrote:
[snip]
+ newtp-highest_sack = treq-snt_isn + 1;
That's the only initialization that you have for highest_sack, I think
that you should initialize it when a loss is detected to the start_seq
of the first packet that wasn't acked.
I agree
It is maintained only in during slowpath of tcp_ack(). In most
cases this seqno is available in the last ACK but there is no
guarantee for that. The new fast recovery loss marking algorithm
needs this as entry point.
Signed-off-by: Ilpo Järvinen [EMAIL PROTECTED]
---
include/linux/tcp.h |
Complete rewrite for update_scoreboard and mark_head_lost. Couple
of hints became unnecessary because of this change. Changes
!TCPCB_TAGBITS check from the original to !(S|L) but it shouldn't
make a difference, and if there ever is a R only skb TCP will
mark it as LOST too. The algorithm uses some
Hi,
This patch enables ip command to add IFA_F_HOMEADDRESS and
IFA_F_NODAD flags while adding the IPv6 address.
Signed-off-by: Noriaki TAKAMIYA[EMAIL PROTECTED]
Signed-off-by: Masahide NAKAMURA [EMAIL PROTECTED]
---
ip/ipaddress.c | 23 ++-
1 files changed, 22
Hi Jeff,
4 small updates of myri10ge for 2.6.21 will follow this mail:
1) fix error checking and return value in myri10ge_allocate_rings
2) use pci_map_page to prepare the dmatest buffer
3) prevent 4k rdma on SGI TIOCE chipset
4) add a wc_enabled flag to myri10ge_priv
Thanks,
Brice
-
To
Fix a missing error check in myri10ge_allocate_rings() and set status
to -ENOMEM before all actual allocations so that the error path returns
what it should.
Signed-off-by: Brice Goglin [EMAIL PROTECTED]
---
drivers/net/myri10ge/myri10ge.c |4
1 file changed, 4 insertions(+)
Index:
Allocate a specific page and use pci_map_page for dma test instead
of relying on another existing buffer.
Signed-off-by: Brice Goglin [EMAIL PROTECTED]
---
drivers/net/myri10ge/myri10ge.c | 22 --
1 file changed, 16 insertions(+), 6 deletions(-)
Index:
Do not use 4k rdma request on SGI TIOCE chipset since this
bridge does not support it.
Signed-off-by: Brice Goglin [EMAIL PROTECTED]
---
drivers/net/myri10ge/myri10ge.c |6 ++
1 file changed, 6 insertions(+)
Index: linux-rc/drivers/net/myri10ge/myri10ge.c
Add a wc_enabled flag in the myri10ge_priv instead of relying
on mtrr = 0.
Signed-off-by: Brice Goglin [EMAIL PROTECTED]
---
drivers/net/myri10ge/myri10ge.c | 10 +++---
1 file changed, 7 insertions(+), 3 deletions(-)
Index: linux-rc/drivers/net/myri10ge/myri10ge.c
On 3/7/07, Ralf Baechle [EMAIL PROTECTED] wrote:
GFP_* flags have no influence on caching or prefetching.
The zone modifier flags (like GFP_DMA) can in principle affect the
cache/prefetch policy, since they affect what physical address range
the memory is allocated from. I don't know whether
I checked the TCP write queue abstraction patch into the net-2.6.22
GIT tree at:
kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6.22.git
And below is the current version of the second patch which actually
does the RB tree stuff.
This infrastructure is necessary for the SKB hinting
From: Evgeniy Polyakov [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 15:10:30 +0300
When system under heavy stress and must allocate new work
instead of reusing old one, new work must use correct
completion callback.
Patch is based on Philipp's and Lars' work.
I only cleaned small stuff (and
Recent change broke build of UDP.
Signed-off-by: Stephen Hemminger [EMAIL PROTECTED]
--- net-2.6.22.orig/net/ipv4/udp.c 2007-03-07 13:43:59.0 -0800
+++ net-2.6.22/net/ipv4/udp.c 2007-03-07 13:45:43.0 -0800
@@ -1283,7 +1283,7 @@
* the network is concerned, anyway)
On Wed, Mar 07, 2007 at 01:47:43PM -0800, Stephen Hemminger wrote:
Recent change broke build of UDP.
Signed-off-by: Stephen Hemminger [EMAIL PROTECTED]
Thanks for that Stephen. Somehow this one got away. Perhaps we
can recode these macros so that they always get compiled-checked?
Cheers,
On Thu, 08 Mar 2007 03:15:43 +0900 (JST)
Noriaki TAKAMIYA [EMAIL PROTECTED] wrote:
Hi,
This patch enables ip command to add IFA_F_HOMEADDRESS and
IFA_F_NODAD flags while adding the IPv6 address.
Signed-off-by: Noriaki TAKAMIYA[EMAIL PROTECTED]
Signed-off-by: Masahide NAKAMURA [EMAIL
On a Ubuntu x86_64 machine the command:
tc qdisc ls
gets a RTNETLINK error. It looks like some data structure alignment
issue got introduced between 2.6.18 and 2.6.19-rc1 and continues on.
Since netlink data structure layout is part of the user ABI, we need to go back
and keep compatibility.
From: Herbert Xu [EMAIL PROTECTED]
Date: Thu, 8 Mar 2007 08:51:48 +1100
Perhaps we can recode these macros so that they always get
compiled-checked?
Yes, we really should.
-
To unsubscribe from this list: send the line unsubscribe netdev in
the body of a message to [EMAIL PROTECTED]
More
From: Stephen Hemminger [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 13:47:43 -0800
Recent change broke build of UDP.
Signed-off-by: Stephen Hemminger [EMAIL PROTECTED]
Applied, thanks Stephen.
-
To unsubscribe from this list: send the line unsubscribe netdev in
the body of a message to [EMAIL
From: [EMAIL PROTECTED]
Date: Tue, 06 Mar 2007 02:41:56 -0800
From: Atsushi Nemoto [EMAIL PROTECTED]
It seems netif_receive_skb() was designed not to call from irq context, but
NAPI + NETPOLL break this rule. If netif_receive_skb() was called from irq
context, redirect to netif_rx()
Mitsuru Chinen [EMAIL PROTECTED] wrote:
On linux-2.6.21-rc2 or later, IPv6 link-local address is not assigned to
some kind of interfaces during system start-up. (I found this issue
occures with e100, e1000 and tg3.)
Here is an alternative fix.
[IPV6]: Do not set IF_READY if device is down
From: Herbert Xu [EMAIL PROTECTED]
Date: Thu, 08 Mar 2007 10:47:56 +1100
Mitsuru Chinen [EMAIL PROTECTED] wrote:
On linux-2.6.21-rc2 or later, IPv6 link-local address is not assigned to
some kind of interfaces during system start-up. (I found this issue
occures with e100, e1000 and
--- Dale Farnsworth [EMAIL PROTECTED] wrote:
On Tue, Mar 06, 2007 at 02:42:02AM -0800, [EMAIL PROTECTED] wrote:
From: Giridhar Pemmasani [EMAIL PROTECTED]
During initialization, mv643xx driver registers IRQ before setting up
tx/rx
rings. This causes kernel oops because mv643xx_poll,
Here is the lastest version of the netem patch to use hrtimers.
It is against the current net tree, so it will need adjusting to fit
with new psched/ktime stuff.
---
include/net/pkt_sched.h |3 ++
net/sched/sch_api.c | 30 +
net/sched/sch_netem.c | 68
One change introduced by the workqueue removal patch is that adding an
interface that is up to a bridge which is also up does not ever call
br_stp_enable_port(), leaving the port in DISABLED state until we do
ifconfig down and up or link events occur.
The following patch to the br_add_if function
From: Stephen Hemminger [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 16:03:55 -0800
One change introduced by the workqueue removal patch is that adding an
interface that is up to a bridge which is also up does not ever call
br_stp_enable_port(), leaving the port in DISABLED state until we do
While debugging I found it useful to see what packet types are in use.
This creates /proc/net/ptype with the format:
Type Device Function
ALL eth0 [880b01b6] :af_packet:packet_rcv_spkt+0x0
0800 [802330f8] ip_rcv+0x0
0011 [881021ef]
On Wed, 07 Mar 2007 15:18:44 -0800 (PST), David Miller [EMAIL PROTECTED]
wrote:
netpoll_rx() should be invokable from hardware interrupt context.
What is the crash you are seeing?
The problem is not netpoll_rx(). It should be called from irq context.
The problem is, netif_receive_skb() is
On Wed, 07 Mar 2007 15:54:38 -0800 (PST)
David Miller [EMAIL PROTECTED] wrote:
From: Herbert Xu [EMAIL PROTECTED]
Date: Thu, 08 Mar 2007 10:47:56 +1100
Mitsuru Chinen [EMAIL PROTECTED] wrote:
On linux-2.6.21-rc2 or later, IPv6 link-local address is not assigned to
some kind of
Help
--
Luca Fornasari
FURNA.COM
-
To unsubscribe from this list: send the line unsubscribe netdev in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
From: Stephen Hemminger [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 16:42:59 -0800
While debugging I found it useful to see what packet types are in use.
This creates /proc/net/ptype with the format:
Type Device Function
ALL eth0 [880b01b6] :af_packet:packet_rcv_spkt+0x0
On Wed, 7 Mar 2007, David Miller wrote:
I checked the TCP write queue abstraction patch into the net-2.6.22
GIT tree at:
kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6.22.git
While tried to apply your patch below, I noticed it's already there. It
turns out that
From: Ilpo_Järvinen [EMAIL PROTECTED]
Date: Thu, 8 Mar 2007 06:21:03 +0200 (EET)
On Wed, 7 Mar 2007, David Miller wrote:
I checked the TCP write queue abstraction patch into the net-2.6.22
GIT tree at:
kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6.22.git
While tried to
Since Linus pulled in several networking bug fixes earlier
this evening, I've rebased the net-2.6.22 tree.
The TCP RB-tree stuff is in there too. There is some bug in
it still with TSO that causes an OOPS in tcp_ack() which I'll
try to fix tonight. If you want to play simply disable TSO
on
It is maintained only in during slowpath of tcp_ack(). In most
cases this seqno is available in the last ACK but there is no
guarantee for that. The new fast recovery loss marking algorithm
needs this as entry point.
Signed-off-by: Ilpo Järvinen [EMAIL PROTECTED]
---
Now rebased these two
Complete rewrite for update_scoreboard and mark_head_lost. Couple
of hints became unnecessary because of this change, NewReno code
was greatly simplified. I changed !TCPCB_TAGBITS check from the
original to a !(S|L) check but it shouldn't make a difference,
and if there ever is a R only skb TCP
From: Dale Farnsworth [EMAIL PROTECTED]
Signed-off-by: Dale Farnsworth [EMAIL PROTECTED]
Acked-by: Giridhar Pemmasani [EMAIL PROTECTED]
---
Giri, thank you very much for reporting the problem and confirming
the fix.
Jeff, please apply.
Andrew, this patch supersedes the -mm patch:
I just checked in a fix for the TSO oops with RB-tree to
net-2.6.22 as follows:
commit 7de509de7c166598b0873dc9ef4c98424307f6b5
Author: David S. Miller [EMAIL PROTECTED]
Date: Wed Mar 7 21:31:17 2007 -0800
[TCP]: Fix bug in tcp_rb_insert.
We should advance rb_parent inside of the
From: Andrew Morton [EMAIL PROTECTED]
Revert 7628b0a8c01a02966d2228bdf741ddedb128e8f8. Thomas Bachler
reports:
Commit 7628b0a8c01a02966d2228bdf741ddedb128e8f8 (drivers/net/tulip/dmfe:
support basic carrier detection) breaks networking on my Davicom DM9009.
ethtool always reports there
From: YOSHIFUJI Hideaki [EMAIL PROTECTED]
Tetsuo Handa [EMAIL PROTECTED] told me that connect(2) with TCPv6
socket almost always took a few minutes to return when we did not have any
ports available in the range of net.ipv4.ip_local_port_range.
The reason was that we used incorrect seed for
On Wed, 07 Mar 2007 09:12:12 -0800
Greg KH [EMAIL PROTECTED] wrote:
From: Andrew Morton [EMAIL PROTECTED]
Revert 7628b0a8c01a02966d2228bdf741ddedb128e8f8. Thomas Bachler
reports:
Commit 7628b0a8c01a02966d2228bdf741ddedb128e8f8 (drivers/net/tulip/dmfe:
support basic carrier
From: Andi Kleen [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 13:12:46 +0100
Well that still needs the ugly div64_64 function. At least my goal was to
eliminate that, not make it faster (I don't see any evidence this function
is particularly performance critical). You prefer to keep div64_64?
I
On Wed, 2007-03-07 at 06:06 -0500, Jeff Garzik wrote:
Adrian Bunk wrote:
Subject: forcedeth: skb_over_panic
References : http://bugzilla.kernel.org/show_bug.cgi?id=8058
Submitter : Albert Hopkins [EMAIL PROTECTED]
Status : unknown
I think this is fixed by the recent
Dave,
I did some research based on your feedback. Specifically, I
looked at removing -map_queue() and allowing -enqueue() to take care
of mapping and locking of the queues (and -dequeue()). I found that it
can be done either way, but I'm not sure I like taking the locking out
of
This patch removes the following not or no longer used exports:
- drivers/char/random.c: secure_tcp_sequence_number
- net/dccp/options.c: sysctl_dccp_feat_sequence_window
- net/netlink/af_netlink.c: netlink_set_err
Signed-off-by: Adrian Bunk [EMAIL PROTECTED]
---
This patch was already sent on:
On Tue, Mar 06, 2007 at 12:44:08AM -0800, Andrew Morton wrote:
...
Changes since 2.6.20-rc2-mm1:
...
git-netdev-all.patch
...
git trees
...
drivers/net/wireless/libertas/debugfs.c contains:
#define item_addr(n) ((u32) ((wlan_adapter *)0)-n)
This is wrong on 64bit architectures.
cu
Adrian
I didn't say to use skb-priority, I said to shrink skb-priority down
to a u16 and then make another u16 which will store your queue mapping
value.
-
To unsubscribe from this list: send the line unsubscribe netdev in
the body of a message to [EMAIL PROTECTED]
More majordomo info at
From: Adrian Bunk [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 23:40:59 +0100
This patch removes the following not or no longer used exports:
- drivers/char/random.c: secure_tcp_sequence_number
- net/dccp/options.c: sysctl_dccp_feat_sequence_window
- net/netlink/af_netlink.c: netlink_set_err
On Wed, Mar 07, 2007 at 11:40:59PM +0100, Adrian Bunk wrote:
This patch removes the following not or no longer used exports:
- drivers/char/random.c: secure_tcp_sequence_number
This part looks reasonable.
Acked-by: Matt Mackall [EMAIL PROTECTED]
--
Mathematics is the supreme nostalgia of our
On Wed, 7 Mar 2007 23:41:16 +0100
Adrian Bunk [EMAIL PROTECTED] wrote:
On Tue, Mar 06, 2007 at 12:44:08AM -0800, Andrew Morton wrote:
...
Changes since 2.6.20-rc2-mm1:
...
git-netdev-all.patch
...
git trees
...
drivers/net/wireless/libertas/debugfs.c contains:
#define
From: Eric Paris [EMAIL PROTECTED]
Date: Fri, 02 Mar 2007 13:29:50 -0500
The security hooks to check permissions to remove an xfrm_policy were
actually done after the policy was removed. Since the unlinking and
deletion are done in xfrm_policy_by* functions this moves the hooks
inside those
From: Eric Paris [EMAIL PROTECTED]
Date: Fri, 02 Mar 2007 13:51:24 -0500
pfkey_spdget neither had an LSM security hook nor auditing for the
removal of xfrm_policy structs. The security hook was added when it was
moved into xfrm_policy_byid instead of the callers to that function by
my
From: Eric Paris [EMAIL PROTECTED]
Date: Fri, 02 Mar 2007 15:15:36 -0500
Inside pfkey_delete and xfrm_del_sa the audit hooks were not called if
there was any permission/security failures in attempting to do the del
operation (such as permission denied from security_xfrm_state_delete).
This
From: David Miller [EMAIL PROTECTED]
Date: Wed, 07 Mar 2007 15:43:16 -0800 (PST)
From: Eric Paris [EMAIL PROTECTED]
Date: Fri, 02 Mar 2007 13:51:24 -0500
pfkey_spdget neither had an LSM security hook nor auditing for the
removal of xfrm_policy structs. The security hook was added when it
The basic calculation has to be done in 32 bits to avoid
doing 64 bit divide by 3. The value x is only 22bits max
so only need full 64 bits only for x^2.
Signed-off-by: Stephen Hemminger [EMAIL PROTECTED]
---
net/ipv4/tcp_cubic.c |8
1 file changed, 4 insertions(+), 4 deletions(-)
On Wed, 2007-03-07 at 10:14 -0800, Stephen Hemminger wrote:
On Wed, 07 Mar 2007 09:12:12 -0800
Greg KH [EMAIL PROTECTED] wrote:
From: Andrew Morton [EMAIL PROTECTED]
Revert 7628b0a8c01a02966d2228bdf741ddedb128e8f8. Thomas Bachler
reports:
Commit
On Wed, 7 Mar 2007, Dan Williams wrote:
Definitely right. If it doesn't work for your card, it needs to be
fixed for your card.
Well, regressions are regressions. And they are a *lot* more important
than any new features. If it doesn't work, it gets reverted.
Linus
-
To
David Miller wrote:
From: Stephen Hemminger [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 17:07:31 -0800
The basic calculation has to be done in 32 bits to avoid
doing 64 bit divide by 3. The value x is only 22bits max
so only need full 64 bits only for x^2.
Signed-off-by: Stephen Hemminger
From: Stephen Hemminger [EMAIL PROTECTED]
Date: Wed, 07 Mar 2007 19:10:47 -0800
David Miller wrote:
What about Willy Tarreau's supposedly even faster variant?
Or does this incorporate that set of improvements?
That's what this is:
x = (2 * x + (uint32_t)div64_64(a,
On Wed, Mar 07, 2007 at 07:10:47PM -0800, Stephen Hemminger wrote:
David Miller wrote:
From: Stephen Hemminger [EMAIL PROTECTED]
Date: Wed, 7 Mar 2007 17:07:31 -0800
The basic calculation has to be done in 32 bits to avoid
doing 64 bit divide by 3. The value x is only 22bits max
so
On Wed, 7 Mar 2007 08:42:39 +0200 (EET)
Pekka Savola [EMAIL PROTECTED] wrote:
On Tue, 6 Mar 2007, Chris Friesen wrote:
Stephen Hemminger wrote:
+arp_notify - BOOLEAN
+ Define mode for notification of address and device changes.
+ 0 - (default): do nothing
+ 1 - Generate gratuitous
On Tue, Mar 06, 2007 at 02:20:55PM -0800, David Miller wrote:
From: Robert Olsson [EMAIL PROTECTED]
Date: Tue, 6 Mar 2007 14:26:04 +0100
David Miller writes:
Actually, more accurately, the conflict exists in how this GC
logic is implemented. The core issue is that hash table
Make sure not to copy_to_user more than user's buffer can handle (we
already checked the min, just use it) in ipv6_getsockopt_sticky. And
while there, minor whitespace cleanup now that ipv6_getsockopt_sticky
call can nicely fit on one line.
Signed-off-by: Chris Wright [EMAIL PROTECTED]
---
diff
On 25-02-2007 10:08, Simon Arlott wrote:
This happens on every boot if more information is needed:
[ 37.393715] =
[ 37.393830] [ INFO: inconsistent lock state ]
[ 37.393881] 2.6.21-rc1-git #146
[ 37.393929] -
[
77 matches
Mail list logo