On 10/02/15(Tue) 03:04, Claudio Jeker wrote:
There is no need to not allow the same network to be configured more then
once. Instead just rely on the multipath and priority handling of the
routing table to select the right route.
Additionally this removes cloned routes (arp/npd cache) when the
I'd love to hear from people using USB Ethernet adapters about the
if_input() conversion diff below.
Please make sure everything works as before, including tcpdump(8).
I also take ok on a per-driver basis :)
M
Index: if_aue.c
===
On 19/03/15(Thu) 14:11, Martin Pieuchot wrote:
When a host want to send packets to a destination whose Ethernet address
that has not been resolved yet, it puts such packet on a mbuf queue.
Right now this queue, linked to the corresponding ARP data structure, is
hand rolled. I wrote the diff
On 21/03/15(Sat) 17:48, David Gwynne wrote:
On 19 Mar 2015, at 11:11 pm, Martin Pieuchot m...@openbsd.org wrote:
When a host want to send packets to a destination whose Ethernet address
that has not been resolved yet, it puts such packet on a mbuf queue.
Right now this queue
When a host want to send packets to a destination whose Ethernet address
that has not been resolved yet, it puts such packet on a mbuf queue.
Right now this queue, linked to the corresponding ARP data structure, is
hand rolled. I wrote the diff below during s2k15 to make use of the
mq_enqueue(9)
On 12/02/15(Thu) 12:35, Martin Pieuchot wrote:
On 10/02/15(Tue) 03:04, Claudio Jeker wrote:
There is no need to not allow the same network to be configured more then
once. Instead just rely on the multipath and priority handling of the
routing table to select the right route.
Additionally
Hi Renato,
On 06/03/15(Fri) 15:03, Renato Westphal wrote:
Hi all,
I've done a lot of work on ldpd(8) a long time ago but only now I
found time to organize my patches and send them to review. The patches
can be found here:
https://github.com/rwestphal/openbsd-ldpd/commits/renato-2015
The
On 26/02/15(Thu) 01:51, Martin Pieuchot wrote:
On 22/02/15(Sun) 00:54, Martin Pieuchot wrote:
On 20/02/15(Fri) 23:15, Stefan Sperling wrote:
In the following configuration I can pretty easily trigger endless
screenfulls of scrolling ehci_idone: ex=%p is done! messages,
where %p
I'm progressively changing how pseudo-drivers are plugged into our
network stack with the goal to turning them MP-safe.
The diff below is a simple refactoring and should not introduce any
behavior change. It moves a bridge-specific vlan-related chunk of
code into the bridge(4) driver.
I'd like
On 05/03/15(Thu) 12:25, David Higgs wrote:
On Mar 3, 2015, at 8:44 AM, David Higgs hig...@gmail.com wrote:
With much help from mpi@, I have made a first big step towards improving
upd(4). I’m not sure when tree lock ends, but I’m still happy to accept
feedback if right now isn’t the
This spring the new trend has a name: if_input() !
If you are the owner of one of these USB Ethernet dongle, please do
me a favor a make sure they still work as expected with the diff
below.
Martin
Index: if_cdce.c
===
RCS file:
Even our ISA Ethernet drivers can be converted to if_input(). If you
still use some of these, I appreciate test reports.
I'm asking here because Miod said everybody can test them... hum hum.
Alternatively, if you think some drivers can go away, I'll summon
tedu@.
Index: isa/if_ef_isapnp.c
On 26/03/15(Thu) 08:00, Ted Unangst wrote:
Martin Pieuchot wrote:
Even our ISA Ethernet drivers can be converted to if_input(). If you
still use some of these, I appreciate test reports.
I'm asking here because Miod said everybody can test them... hum hum.
Alternatively, if you
How do people feel about printing the prefixlen in CIDR notation? I'm
annoyed about outputs not fitting in 80 chars when using autoconf magic:
-inet6 fd00::f2de:f1ff:fe6a:15d1 prefixlen 64 autoconf pltime 3594 vltime 7194
+inet6 fd00::f2de:f1ff:fe6a:15d1/64 autoconf pltime 3594 vltime 7194
I'd like to be able to easily identify how deep are the tentacles of a
pseudo-driver in your network stack. This diff takes the example of
gif(4) and move all the remaining blocks checking for IFT_GIF under
the appropriate #ifdef dance.
The #ifdef I'm adding below are not strictly needed to
On 31/03/15(Tue) 23:06, David Higgs wrote:
This was much more straightforward than expected.
- Replace an array with a LIST of allocated sensors.
- Remove or rescope variables counting sensors.
- Allocated sensors are always attached.
- Drop an unnecessary size calculation.
Do you need
On 31/03/15(Tue) 21:56, Kimberley Manning wrote:
Hi,
This diff makes the pcn driver use m_defrag for fragmented mbuf chains,
I like this kind of cleanups. As for vio(4) could you try the diff or
are you looking for testers?
Are you after something specific or are you changing the drivers
When if_input_process() will pass a mbuf to pseudo-interface handlers,
they will change the `rcvif` pointer in the packet header. That's why
we should not pass a ifp pointer to the handlers and instead let them
look at the value of `rcvif`.
Diff below change if_input() and ether_input() to no
On 26/03/15(Thu) 17:46, Henning Brauer wrote:
* Mike Belopuhov m...@belopuhov.com [2015-03-26 14:36]:
On 26 March 2015 at 14:27, Stuart Henderson st...@openbsd.org wrote:
seems reasonable. (I'd quite like that for v4 too, though it wouldn't
cope with non-contiguous netmask ;)
On 31/03/15(Tue) 18:06, Miod Vallat wrote:
What, no ec? No ep? Why play favorites???
You're asking for tricky ones!
Enjoy :)
I'm afraid my ISA ec(4) seems to no longer work. Blinks red during POST,
doesn't get detected by the kernel (yes, it's jumped correctly).
Does that mean we
[moved to tech@]
On 27/02/15(Fri) 11:40, patrick keshishian wrote:
Hi,
On 2/26/15, Ulf Brosziewski ulf.brosziew...@t-online.de wrote:
On 02/27/2015 03:31 AM, Ulf Brosziewski wrote:
...
It might be that the following patch to wsmouse.c solves the problem
with the new version of
On 19/02/15(Thu) 21:49, David Higgs wrote:
On Feb 13, 2015, at 7:29 AM, David Higgs hig...@gmail.com wrote:
On Friday, February 13, 2015, Martin Pieuchot mpieuc...@nolizard.org
wrote:
On 13/02/15(Fri) 00:28, David Higgs wrote:
I guess nobody else has tried calling
On 27/02/15(Fri) 12:07, Stefan Sperling wrote:
On Thu, Feb 26, 2015 at 10:45:01PM +0100, Stefan Sperling wrote:
There are other issues which are not fixed by this diff and which
my diff did not fix either.
But these are bugs we can investigate another day. Happy travels!
I suppose
On 26/02/15(Thu) 22:45, Stefan Sperling wrote:
On Thu, Feb 26, 2015 at 01:51:29AM +0100, Martin Pieuchot wrote:
On 22/02/15(Sun) 00:54, Martin Pieuchot wrote:
On 20/02/15(Fri) 23:15, Stefan Sperling wrote:
In the following configuration I can pretty easily trigger endless
screenfulls
On 20/02/15(Fri) 23:15, Stefan Sperling wrote:
In the following configuration I can pretty easily trigger endless
screenfulls of scrolling ehci_idone: ex=%p is done! messages,
where %p is to a constant pointer value (same in each message).
[ehci host]--[usb extension cable (hub 1)]--[usb
On 22/02/15(Sun) 00:54, Martin Pieuchot wrote:
On 20/02/15(Fri) 23:15, Stefan Sperling wrote:
In the following configuration I can pretty easily trigger endless
screenfulls of scrolling ehci_idone: ex=%p is done! messages,
where %p is to a constant pointer value (same in each message
On 26/03/15(Thu) 17:39, Florian Obser wrote:
On Thu, Mar 26, 2015 at 01:48:03PM +0100, Martin Pieuchot wrote:
How do people feel about printing the prefixlen in CIDR notation? I'm
annoyed about outputs not fitting in 80 chars when using autoconf magic:
-inet6 fd00::f2de:f1ff:fe6a:15d1
On 26/03/15(Thu) 17:35, Florian Obser wrote:
On Thu, Mar 26, 2015 at 05:46:12PM +0100, Henning Brauer wrote:
* Mike Belopuhov m...@belopuhov.com [2015-03-26 14:36]:
however I agree that if we do this for ipv6 we should do it for ipv4 as
well
but then do we care about tons of stuff out
With the increasing number of machines shipping with rate-matching
hubs instead of companion controllers to support USB Full and Low-
speed devices, a number of people asked me if it was possible to
add support for Full-speed isochronous transfers in order to use
USB1.1 uaudio(4) devices with
On 30/01/15(Fri) 01:25, Ulf Brosziewski wrote:
Probably I was too sceptical about synaptics.c. The bug I observed
with the ALPS touchpad seems to be due to a kind of mismatch between
the ALPS code in pms and the event handling in wsconscomm. The patch
below contains the initial change as well
On 14/04/15(Tue) 07:40, attila wrote:
[...]
Feedback most welcome.
See below.
/* -*- mode:c; tab-width:8; indent-tabs-mode:t; c-basic-offset:8 -*- */
We do not include editor settings in files, the first line should
contain:
/* $OpenBSD$ */
Which will be expanded by CVS.
/*
*
On 15/04/15(Wed) 23:00, Claudio Jeker wrote:
On Wed, Mar 18, 2015 at 05:46:34AM +0100, Claudio Jeker wrote:
On Tue, Mar 17, 2015 at 05:35:21PM +0100, Martin Pieuchot wrote:
On 12/02/15(Thu) 12:35, Martin Pieuchot wrote:
On 10/02/15(Tue) 03:04, Claudio Jeker wrote:
There is no need
On 17/04/15(Fri) 16:47, Dimitris Papastamos wrote:
Hi,
This patch adds an option to usbdevs(8) to disable USB bus probing
at runtime. The operation is restricted to the root user.
It would be nice to show if probing is on or off, for example
# usbdevs -p
bus probing: on
# usbdevs -p off
Here's a new diff to convert carp_input() to use ether_input_mbuf()
instead of ether_input(). This is a necessary step to be able to
convert other pseudo-drivers to if_input() but it is just needed for
the transition.
This version only call M_PREPEND(9) on mbufs that will be passed to
This diff adds the necessary glue to bridge(4) to be able to convert
other pseudo-drivers to if_input(). This will also help bridge(4)'s
own conversion.
Since bridge_input() already returns a mbuf, I believe this change is
less error prone than the carp(4) one. But I appreciate reviews,
tests
On 21/04/15(Tue) 12:35, Martin Pieuchot wrote:
This diff adds the necessary glue to bridge(4) to be able to convert
other pseudo-drivers to if_input(). This will also help bridge(4)'s
own conversion.
Since bridge_input() already returns a mbuf, I believe this change is
less error prone
On 24/04/15(Fri) 20:48, David Higgs wrote:
This is the final patch in the series.
Utilize the pending flags and report callback for their intended purpose - to
process async behavior.
Apply splusb() to ensure report callbacks can't fire before their data
structures have been properly
On 21/04/15(Tue) 17:15, Vitaliy Makkoveev wrote:
Now fd_getfile() function returns unacquired struct file instance
[...]
It is unacceptable on multiprocessor machine because the instance referenced
by fp can be destroyed between fd_getfile() and FREF() calls. So I want
fd_getfile() returns
On 22/04/15(Wed) 10:29, Dimitris Papastamos wrote:
[...]
On a side note, What do you think of using a sysctl instead of an ioctl?
I guess an ioctl will do but it would make it easy for people to disable
this at boot time by simply adding the relevant knob in /etc/sysctl.conf.
Otherwise I
On 24/04/15(Fri) 21:21, Johan Huldtgren wrote:
a few hours after I sent the previous e-mail the backup
(April 23rd snap) took over and became the master, at
that point I could not reach the carp interfaces anymore.
Reverting roles so the host running the April 12th snap
became the master
On 02/05/15(Sat) 08:44, attila wrote:
Hi tech@,
This patch adds man pages for usbd_open_pipe, usbd_open_pipe_intr,
usbd_close_pipe and usbd_abort_pipe, done as two files:
usbd_open_pipe.9 and usbd_close_pipe.9. It also adds these two new .9
files to the appropriate Makefile and tweaks
Hello Johan and thanks for your great report!
On 27/04/15(Mon) 11:54, Johan Huldtgren wrote:
If you try 1.250 and 1.253 and tell me if you can reproduce the problem
that would be really helpful. In case you see something weird, Could
you include the routing table netstat -rnf inet in your
Now that the stack is ready, here's a diff to take the first victim,
trunk(4), out of ether_input().
If you glanced over previous versions of this diff, the differences
are mostly in sys/net/trunklacp.c and include some printf(9) fixes
in bonus.
Ok?
Index: net/if_ethersubr.c
On 20/04/15(Mon) 18:37, Mike Belopuhov wrote:
On Tue, Apr 14, 2015 at 22:08 +0300, Lauri Tirkkonen wrote:
On Tue, Apr 14 2015 20:40:58 +0200, Mike Belopuhov wrote:
According to 3.2 in RFC 7323:
Once TSopt has been successfully negotiated, that is both SYN and
SYN,ACK contain
This diff is a first step towards removing all pseudo-driver #ifdef
in ether_output(). As for ether_input() the goal of this work is to
provide an elegant design to make it easier to turn pseudo-drivers
MP-safe.
So instead of including some bridge(4), vlan(4) and carp(4) specific
code in
On 07/05/15(Thu) 20:58, Mike Belopuhov wrote:
As I've pointed out before, on panic we can be running on any
CPU and our disk controller's interrupts can interrupt on the
other one. Since we'll most likely be holding a kernel lock,
dealing with unlocking it might get hairy very fast. Instead
On 2015-05-07 04:19, David Higgs wrote:
On Apr 30, 2015, at 7:09 AM, Martin Pieuchot m...@openbsd.org wrote:
[...]
Your tweaks were good, so I tweaked it further:
- When submit fails, invalidate affected sensors as described above.
- When invalidating sensors, do it recursively.
- When battery
Diff below convert vlan(4) to the new if_input() framework which means
that vlan_input() will now be executed before ether_input().
Compared to trunk(4) multiple vlan(4)s can be attached to the same
parent interface. When such thing happens only one input handler is
added to keep the if_inputs
On 07/05/15(Thu) 11:50, Martin Pieuchot wrote:
This diff is a first step towards removing all pseudo-driver #ifdef
in ether_output(). As for ether_input() the goal of this work is to
provide an elegant design to make it easier to turn pseudo-drivers
MP-safe.
So instead of including some
There's only one #if NTRUNK left in the tree, this diff gets rid of it
by overwriting trunk ports' if_output().
ok?
Index: net/if_ethersubr.c
===
RCS file: /cvs/src/sys/net/if_ethersubr.c,v
retrieving revision 1.196
diff -u -p
On 14/05/15(Thu) 08:20, sven falempin wrote:
On Thu, May 14, 2015 at 6:27 AM, Martin Pieuchot m...@openbsd.org wrote:
one level of indentation.
pre
if (X == null ) {dprintf(you broke this); return ;}
unindented code;
/pre
Same gain, more dev ease and corner case check.
I disagree
If we change the rcvif pointer of a packet we need to run if_input()
again otherwise we might skip the handlers on the new interface.
Ultimately it would be nice to only assign rcvif in if_input(), but
that's for another diff.
This fix one case I left out in my previous conversion, ok?
Index:
I have one setup with multiple interfaces in a bridge and on some of
these interfaces some vlan(4)s. But there's currently a bug that
prevent us to send (receive is fine) VLAN packets in such config.
Diff below fixes that.
The problem is that vlan_output() does not pass its parent interface
to
On 12/05/15(Tue) 12:15, Martin Pieuchot wrote:
On 07/05/15(Thu) 11:50, Martin Pieuchot wrote:
This diff is a first step towards removing all pseudo-driver #ifdef
in ether_output(). As for ether_input() the goal of this work is to
provide an elegant design to make it easier to turn pseudo
ifp can never be NULL in carp_set_ifp(), so remove the check and gain
one level of indentation.
ok?
Index: netinet/ip_carp.c
===
RCS file: /cvs/src/sys/netinet/ip_carp.c,v
retrieving revision 1.254
diff -u -p -r1.254 ip_carp.c
---
the situation.
Can I have oks?
On 15 maj 2015, at 13:14, Martin Pieuchot m...@openbsd.org wrote:
I have one setup with multiple interfaces in a bridge and on some of
these interfaces some vlan(4)s. But there's currently a bug that
prevent us to send (receive is fine) VLAN packets in such config
On 15/05/15(Fri) 15:53, Martin Pieuchot wrote:
Some more if_output() conversion. The xl bits are here because I'd
like to reduce the number of places where IFQ_ENQUEUE() is used.
After applying this diff you should only have a couple left.
Anyone?
Ok?
Index: dev/usb/if_upl.c
On 13/04/15(Mon) 22:39, Rafal Brodewicz wrote:
Hi tech@
Enabling usb3 support in my ThinkPad T440s causes dock station usb ports
to stop working. I see uhub0: device problem, disabling port 3 message
in dmesg.
dmesg with usb3 on and off included.
Seems that the hub in your dock fails to
On 16/04/15(Thu) 10:02, Dimitris Papastamos wrote:
ping
A bit too soon to ping ;) You could wait at least a week!
On 15/04/15(Wed) 10:46, attila wrote:
Martin Pieuchot m...@openbsd.org writes:
On 14/04/15(Tue) 15:22, attila wrote:
Martin Pieuchot m...@openbsd.org writes:
static const struct usb_devno ualea_devs[] = {
{ USB_VENDOR_ARANEUS, USB_PRODUCT_ARANEUS_ALEA
On 13/04/15(Mon) 17:49, Dimitris Papastamos wrote:
Hi,
Attached below an updated patch that makes the layout a bit more sensible.
It's nice it works.
Now like various students that applied for the GSoC you've done the
first step. The interesting part of that GSoC is also the boring part:
On 16/04/15(Thu) 16:06, attila wrote:
[...]
After this was committed I received a critique of the driver from the
person behind the Alea II (Andreas Gustafsson) who made a few pretty
good points. He felt trying to pull all the entropy off of the device
that would theoretically be available
because most of it is run under the KERNEL_LOCK?
From d7b381d438aa98a80e66c4aceeb905ef0a3ab9f3 Mon Sep 17 00:00:00 2001
From: Martin Pieuchot m...@openbsd.org
Date: Mon, 2 Mar 2015 16:33:04 +0100
Subject: [PATCH 1/5] Check for flags inside fd_getfile(), will help for
upcoming locking.
---
sys/arch
On 14/04/15(Tue) 15:22, attila wrote:
Martin Pieuchot m...@openbsd.org writes:
static const struct usb_devno ualea_devs[] = {
{ USB_VENDOR_ARANEUS, USB_PRODUCT_ARANEUS_ALEA }
};
Is it possible to match your device based on the content of the device
descriptor instead
Here's a diff to convert these drivers to if_input(). They all make
use of m_devget(9) which takes an ifp as argument and I'd like to
change that. But first I need to make sure all these drivers are
converted.
If you can try this diff on one or more chips, please let me know how
it goes.
On 08/04/15(Wed) 10:03, David Gwynne wrote:
On 8 Apr 2015, at 01:38, Martin Pieuchot m...@openbsd.org wrote:
Here's a diff to convert these drivers to if_input(). They all make
use of m_devget(9) which takes an ifp as argument and I'd like to
change that. But first I need to make
On 08/04/15(Wed) 12:15, Mike Belopuhov wrote:
OK to rename gettdbbyaddr to gettdbbydst since that's what it does
and it aligns well with gettdbbysrc?
Makes sense to me. This is also coherent to the comments in ip_ipsp.h.
One comment though, you're moving a splsoftnet() in tdb_delete() but it
Now that all the drivers using m_devget(9) have been converted to
if_input() we no longer need to pass an interface pointer to set
`rcvif'.
This is a small step towards the removal of an interface pointer
in the mbuf header.
While here, m_clget(9) also no longer needs an ifp.
Ok?
Index:
One of the interesting things we've done during s2k15 was to redesign
ether_input(). The name of this function is clearly misleading. Back
in the old days [0] it was simply used to put a packet on a protocol
queue. Today ether_input() still does that, but before it does a lot
of different hacks
On 02/04/15(Thu) 18:43, Ulf Brosziewski wrote:
On 04/02/2015 03:39 AM, Fasse wrote:
On Wed, 01 Apr 2015 21:23:15 +0200
Ulf Brosziewskiulf.brosziew...@t-online.de wrote:
Yes, without some refactoring there won't be an elegant way.
pms_sync_elantech_v2 encodes some sync state in the 'flags'
Same problem as with vlan(4). carp(4) also need to stop calling
ether_input() directly and that implies having a mbuf with the
correct Ethernet header prepended.
Index: netinet/ip_carp.c
===
RCS file:
Changes done by pseudo-driver *after* ether_ifattach() must be undone
*before* ether_ifdetach(). Otherwise it is impossible to ensure we're
leaving a stacked pseudo-interface in a correct state.
Since I don't want to modify every single driver calling ether_ifdetach()
I'm using a new function,
Like vlan(4) and carp(4), bridge(4) also need to be tweaked to still
work during the if_input() transition. Removing the ether_input()
calls in this driver was a bit tricky *because* bridge_input() is
called after ether_input().
See how 3 M_PREPEND() are converted to 1? And soon none will be
Right now vlan_input() is called *after* ether_input(). More precisely
it is called after the mbuf has been m_adj(9)'usted to skip the Ethernet
header. This is not a problem in se but...
To make sure vlan(4) keeps working during the if_input() transition, it
has to stop calling ether_input()
On 19/05/15(Tue) 15:28, Maxime Villard wrote:
-- netinet/ip_icmp.c --
925 rt = rtalloc(sintosa(sin), RT_REPORT|RT_RESOLVE, rtableid);
if (rt == NULL)
return (NULL);
/* Check if the route is actually usable */
On 20/05/15(Wed) 07:40, Henning Brauer wrote:
* Johan Ymerson johan.ymer...@transmode.com [2015-05-19 19:25]:
On Tue, 2015-05-19 at 11:16 +, Johan Ymerson wrote:
On Tue, 2015-05-19 at 11:24 +0100, Stuart Henderson wrote:
On 2015/05/19 10:10, Johan Ymerson wrote:
Yes I understand
On 20/05/15(Wed) 14:14, Johan Ymerson wrote:
[...]
The patch did not apply cleanly to OPENBSD_5_7, I rewrote the patch a
bit:
Thanks, I committed my diff to -current.
With this patch everything (almost) work. At least as good as my patch
did. OSPFd still does something wrong with the link
On 05/06/15(Fri) 15:04, Claudio Jeker wrote:
On Thu, Jun 04, 2015 at 12:19:10PM +0200, Martin Pieuchot wrote:
I'd like to put the link-layer address back into the gateway field of
RTF_LOCAL addresses. The problem is that RTF_LOCAL routes are also
marked as RTF_LLINFO and a lot of code
bridge_ifenqueue() does not need any spl protection, if_output()
already raises it.
ok?
Index: net/if_bridge.c
===
RCS file: /cvs/src/sys/net/if_bridge.c,v
retrieving revision 1.241
diff -u -p -r1.241 if_bridge.c
--- net/if_bridge.c
Diff below moves bridge_output() to if_output(). It fixes the case I
already described some weeks ago where you have a physical interface
in a bridge and a vlan on top of it which is not in the bridge.
It also change the loop prevention code to use M_PROTO1 like in the
input path.
Tests,
This is the last pseudo-driver conversion.
The idea is to run bridge_input() *before* any ifih on an interface.
Doing so allow us to remove the hack between vlan(4) and bridge(4)
and simplify the logic for stacked ifih.
With that net/if_ethersubr.c is now free from #ifdef NPSEUDODRIVER.
As
On 02/06/15(Tue) 22:36, David Higgs wrote:
Here are some new sensors for upd(4) devices. All exist on my device except
AtRateTimeToEmpty, which still seemed a logical addition given that
AtRateTimeToFull is already present.
- AtRateTimeToEmpty
- RunTimeToEmpty
- NeedReplacement
-
I'd like to put the link-layer address back into the gateway field of
RTF_LOCAL addresses. The problem is that RTF_LOCAL routes are also
marked as RTF_LLINFO and a lot of code assume (correctly) that such
routes contain valid ARP or ND information.
I believe we decided to use an ``empty'' lladdr
On 22/06/15(Mon) 17:12, Alexander Bluhm wrote:
On Wed, Jun 17, 2015 at 02:08:14PM +0200, Martin Pieuchot wrote:
diff -u -p -r1.340 if.c
--- net/if.c16 Jun 2015 11:09:39 - 1.340
+++ net/if.c17 Jun 2015 12:03:36 -
@@ -530,6 +530,15 @@ if_input_process(void *xmq
When an interface is detached or destroyed the CPU executing if_detach()
removes all the mbufs received by this interface on three queues:
ARP, IPv4 and IPv6 protocol queues.
This made sense to avoid referencing a dangling rcvif pointer. But now
mbufs contain unique interface indexes and protocol
Time goes by and things must be cleaned. Thanks to claudio@'s work
to support multiple connected routes carp(4) now have its own default
priority. So I audited the remaining iterations on finet and I
couldn't find any good reason to force carp(4) interfaces at a special
position in the list of
On 17/06/15(Wed) 14:07, Martin Pieuchot wrote:
On 08/06/15(Mon) 15:58, Martin Pieuchot wrote:
Diff below moves bridge_output() to if_output(). It fixes the case I
already described some weeks ago where you have a physical interface
in a bridge and a vlan on top of it which
On 24/06/15(Wed) 16:45, Mark Kettenis wrote:
Date: Wed, 24 Jun 2015 16:11:08 +0200
From: Martin Pieuchot m...@openbsd.org
Use only one ipi counter just like other archs do.
ok?
Problem is that the event counters aren't really MP safe. By keeping
them per-CPU you circumvent any
Use only one ipi counter just like other archs do.
ok?
Index: dev/openpic.c
===
RCS file: /cvs/src/sys/arch/macppc/dev/openpic.c,v
retrieving revision 1.81
diff -u -p -r1.81 openpic.c
--- dev/openpic.c 24 Jun 2015 11:58:06
On 20/06/15(Sat) 18:50, frantisek holop wrote:
trying to use a replacement usb enclosure for an old 2.5 IDE drive.
[...]
Controller /dev/usb0:
addr 1: high speed, self powered, config 1, EHCI root hub(0x),
Intel(0x8086), rev 1.00
port 2 addr 6: high speed, self powered, config 1, USB
On 26/06/15(Fri) 16:00, Alexandr Nedvedicky wrote:
Hello Martin,
I accept or your comments. I just have few quick notes/questions now.
2) I saw that you found some ALTQ leftovers, you have some Solaris
(2) I think ALTQs leftovers are still in CVS repo, will double check
anyway. Stack
On 26/06/15(Fri) 17:19, Alexandr Nedvedicky wrote:
On Fri, Jun 26, 2015 at 04:34:06PM +0200, Martin Pieuchot wrote:
On 26/06/15(Fri) 16:00, Alexandr Nedvedicky wrote:
2) I saw that you found some ALTQ leftovers, you have some Solaris
(2) I think ALTQs leftovers are still in CVS repo
During clone/destroy stress tests on pseudo-interfaces I found a double
free easily reproducible with dhclient(8) running on top of a trunk(4).
The problem comes from trunk_ether_delmulti() which is almost identical
to carp's version except that it always free mc.
So when you do # ifconfig
On 08/06/15(Mon) 15:58, Martin Pieuchot wrote:
Diff below moves bridge_output() to if_output(). It fixes the case I
already described some weeks ago where you have a physical interface
in a bridge and a vlan on top of it which is not in the bridge.
It also change the loop prevention code
On 08/06/15(Mon) 16:11, Martin Pieuchot wrote:
This is the last pseudo-driver conversion.
The idea is to run bridge_input() *before* any ifih on an interface.
Doing so allow us to remove the hack between vlan(4) and bridge(4)
and simplify the logic for stacked ifih.
With that net
It makes sense to not retransmit a packet on the interface it was
received from but we should still call bridge_localbroadcast() on
this interface otherwise the network stack never see this packet.
This fix the configuration where you have a vlan(4) on top of an
interface in a bridge(4) and the
On 28/05/15(Thu) 11:23, Martin Pieuchot wrote:
Let's use if_input() rather than ether_input_mbuf(). This simplifies
the bpf(4) bits and removes some duplicate rcvif assignments.
Once ether_input_mbuf() is no more my plan is to start merging various
ifp counters and put an interface index
On 31/05/15(Sun) 22:05, mxb wrote:
Hello,
any ideas regarding subject ?
Yes various. Any idea about how to reproduce it?
What you're seeing is a side-effect of r1.151 of netinet/if_ether.c.
I see this in ‘dmesg’.
Also all local (on machine itself) arp entries are incomplete:
Yes
On 23/05/15(Sat) 19:08, mxb wrote:
[...]
so far no problems.
Here's an updated version to match recent changes. It also includes a
nitpick fix from bluhm@.
Tests and oks welcome.
Index: net/if_ethersubr.c
===
RCS file:
Replace the last ether_input_mbuf() by if_input().
Index: net/if_tun.c
===
RCS file: /cvs/src/sys/net/if_tun.c,v
retrieving revision 1.144
diff -u -p -r1.144 if_tun.c
--- net/if_tun.c26 May 2015 11:36:26 - 1.144
+++
401 - 500 of 2703 matches
Mail list logo