Hi,
I want merge my diff that avoids queueing in the IP protocol input
path and mpi@'s diff that queues local packets to unlock the
forwarding path.
One step is to rename ip_local() to ip_deliver() and give it the
same parameters as the pr_input functions.
I also added an assert that IPv4 delive
On Sat, May 27, 2017 at 07:24:30PM -0400, Ted Unangst wrote:
> using pread and pwrite is a little shorter and half the syscalls.
>
ok mlarkin. thanks!
> Index: virtio.c
> ===
> RCS file: /cvs/src/usr.sbin/vmd/virtio.c,v
> retrieving
using pread and pwrite is a little shorter and half the syscalls.
Index: virtio.c
===
RCS file: /cvs/src/usr.sbin/vmd/virtio.c,v
retrieving revision 1.46
diff -u -p -r1.46 virtio.c
--- virtio.c11 May 2017 07:31:20 - 1.46
On Sat, May 27, 2017 at 10:58:40PM +0100, Jason McIntyre wrote:
> On Sat, May 27, 2017 at 11:45:43PM +0200, Adam Wolk wrote:
> > Index: chown.8
> > ===
> > RCS file: /cvs/src/bin/chmod/chown.8,v
> > retrieving revision 1.20
> > diff -u
On Sat, May 27, 2017 at 11:45:43PM +0200, Adam Wolk wrote:
> Index: chown.8
> ===
> RCS file: /cvs/src/bin/chmod/chown.8,v
> retrieving revision 1.20
> diff -u -p -r1.20 chown.8
> --- chown.8 31 Dec 2015 23:38:16 - 1.20
> ++
On Sat, May 27, 2017 at 11:01:29PM +0200, Adam Wolk wrote:
> On Sat, May 27, 2017 at 01:42:45PM -0600, Theo de Raadt wrote:
> > I agree with you. Maybe change the comment
> >
> > /* UID and GID are separated by a dot and UID exists. */
> >
> > to say a bit more on the matter, to
Hi tech@,
I've been talking to Rafael about this piece of code over the last
week a bit, and have tried it on my home TV setup. My ISP has a
multicast setup for watching television, with their set-top-box
connected to my TV. Up until now, I was using net/igmpproxy on my
OpenBSD gateway to get th
2017-05-27 22:45 GMT+03:00 Joerg Jung :
> Hi,
>
> I think devel/waf is gone since two years and may not come back, so no
> need to mention in port-modules(5).
>
> OK?
Yes, and don't forget to bring some gas for nice burning!
On Sat, May 27, 2017 at 01:42:45PM -0600, Theo de Raadt wrote:
> I agree with you. Maybe change the comment
>
> /* UID and GID are separated by a dot and UID exists. */
>
> to say a bit more on the matter, to prevent a zealot from arriving 2-3
> years from now and proposing remov
Steven McDonald wrote:
> Hi there,
>
> ext2fs currently has #if 0 around the size check for lengthening a
> file. This allows ftruncate(2) calls for very large files to succeed,
> but result in a smaller file than was requested. I noticed the problem
> while trying to create vmd(8) images on an ex
Hi,
I think devel/waf is gone since two years and may not come back, so no
need to mention in port-modules(5).
OK?
Regards,
Joerg
Index: share/man/man5/port-modules.5
===
RCS file: /cvs/src/share/man/man5/port-modules.5,v
retriev
I agree with you. Maybe change the comment
/* UID and GID are separated by a dot and UID exists. */
to say a bit more on the matter, to prevent a zealot from arriving 2-3
years from now and proposing removal. Just a few words to hint . support
will stay forever.
It seems the sen
Catch up with pfctl/pfctl_osfp.c, no binary change.
Index: usr.sbin/tcpdump/pfctl_osfp.c
===
RCS file: /cvs/src/usr.sbin/tcpdump/pfctl_osfp.c,v
retrieving revision 1.11
diff -u -p -r1.11 pfctl_osfp.c
--- usr.sbin/tcpdump/pfctl_osfp.c
Hi tech@,
I stumbled on SUPPORT_DOT while reading /usr/src/bin/chmod.c, got curious
and started doing some research.
POSIX changed the separator from . to : to make the utility properly work with
usernames containing a dot. The standard doesn't forbid keeping the dot handling
for backwards compat
Hi there,
ext2fs currently has #if 0 around the size check for lengthening a
file. This allows ftruncate(2) calls for very large files to succeed,
but result in a smaller file than was requested. I noticed the problem
while trying to create vmd(8) images on an ext2fs filesystem.
Below is a patch
This reduces the diff with usr.sbin/tcpdump/pfctl_osfp.c. The change
from tcpdump is newer, so change pfctl. No binary change.
Index: sbin/pfctl/pfctl_osfp.c
===
RCS file: /cvs/src/sbin/pfctl/pfctl_osfp.c,v
retrieving revision 1.23
di
Hello,
On Sat, May 27, 2017 at 08:45:31PM +0200, Martin Pieuchot wrote:
> On 27/05/17(Sat) 17:33, Alexandr Nedvedicky wrote:
> > Hello,
> >
> > On Fri, May 26, 2017 at 04:54:57PM +0200, Martin Pieuchot wrote:
> > > The global list of softc is *not* accessed in the input path, so it
> > > doesn't
On 27/05/17(Sat) 12:58, Sebastian Benoit wrote:
> (benno_pflow_try4_task.diff)
>
> ok?
Two nits below with that ok mpi@
> diff --git sys/net/if_pflow.c sys/net/if_pflow.c
> index a40fe23862b..20ab4e0e88d 100644
> --- sys/net/if_pflow.c
> +++ sys/net/if_pflow.c
> @@ -67,6 +67,7 @@ struct pflowsta
On 27/05/17(Sat) 17:33, Alexandr Nedvedicky wrote:
> Hello,
>
> On Fri, May 26, 2017 at 04:54:57PM +0200, Martin Pieuchot wrote:
> > The global list of softc is *not* accessed in the input path, so it
> > doesn't need splnet().
> >
> > ioctl(2) handlers are already executed with the NET_LOCK() he
On Sat, May 27, 2017 at 08:30:01PM +0200, Claudio Jeker wrote:
> Agreed, new diff
OK bluhm@
>
> --
> :wq Claudio
>
> Index: kern/uipc_socket.c
> ===
> RCS file: /cvs/src/sys/kern/uipc_socket.c,v
> retrieving revision 1.184
> diff
On Sat, May 27, 2017 at 08:14:37PM +0200, Michal Mazurek wrote:
> print_ioctl() is unused if not debugging. Found with clang, after
> marking the function "static". Use OSFP_DEBUG, in a similar
> fashion to OPT_DEBUG (pfctl_optimize.c).
>
> OK?
OK bluhm@
>
> Index: sbin/pfctl/pfctl_osfp.c
> ===
On 23/05/17(Tue) 17:41, Mark Kettenis wrote:
> So here is a diff that implements what I proposed recently. This
> recognizes that atomic instructions on amd64 already include an
> implicit memory barrier and allows us to write optimized code that
> avoids a redundant memory barrier.
>
> Note that
On Sat, May 27, 2017 at 07:45:49PM +0200, Alexander Bluhm wrote:
> On Sat, May 27, 2017 at 06:37:48PM +0200, Claudio Jeker wrote:
> > Similar to the routing socket the pfkey socket only needs the
> > netlock further down (in pfkeyv2_send()). With this all users of
> > raw_usrreq are now not using t
This improves our AMS-IX border router (200 configured sessions, plus
Route Servers) starting performance.
>From 45 minutes to bring everyone up, to 3 minutes, and bgpctl stays
acceptably responsive during the thundering herd. Memory pressure is
also far more relaxed.
OK
On 2017 May 26 (Fri) a
On Sat, May 27, 2017 at 06:55:52PM +0200, Sebastian Benoit wrote:
> Sebastian Benoit(be...@openbsd.org) on 2017.05.27 18:35:06 +0200:
> >
> > in a bgpd.conf config with
> >
> > network 2a00:15a8:6:100::/56 set ext-community soo 23:42
> > network 2a00:15a8:6:100::/56
> >
> > the second line "over
print_ioctl() is unused if not debugging. Found with clang, after
marking the function "static". Use OSFP_DEBUG, in a similar
fashion to OPT_DEBUG (pfctl_optimize.c).
OK?
Index: sbin/pfctl/pfctl_osfp.c
===
RCS file: /cvs/src/sbin/pfc
On Sat, May 27, 2017 at 06:37:48PM +0200, Claudio Jeker wrote:
> Similar to the routing socket the pfkey socket only needs the
> netlock further down (in pfkeyv2_send()). With this all users of
> raw_usrreq are now not using the NET_LOCK which should simplify to push
> the NET_LOCK down.
In sorflu
Mark local functions as static. Also remove some double newlines between
function definitions.
Index: sbin/pfctl/pfctl_osfp.c
===
RCS file: /cvs/src/sbin/pfctl/pfctl_osfp.c,v
retrieving revision 1.23
diff -u -p -r1.23 pfctl_osfp.c
--
Sebastian Benoit(be...@openbsd.org) on 2017.05.27 18:35:06 +0200:
>
> in a bgpd.conf config with
>
> network 2a00:15a8:6:100::/56 set ext-community soo 23:42
> network 2a00:15a8:6:100::/56
>
> the second line "overwrites" the first line.
>
> this prints a warning, because i wasted 30 minutes wo
Similar to the routing socket the pfkey socket only needs the
netlock further down (in pfkeyv2_send()). With this all users of
raw_usrreq are now not using the NET_LOCK which should simplify to push
the NET_LOCK down.
OK?
--
:wq Claudio
Index: kern/uipc_socket2.c
in a bgpd.conf config with
network 2a00:15a8:6:100::/56 set ext-community soo 23:42
network 2a00:15a8:6:100::/56
the second line "overwrites" the first line.
this prints a warning, because i wasted 30 minutes wondering why i could not
see the community on the wire...
ok?
(benno_bgpd_parse_dup
Christian Weisgerber wrote:
> Move sha224_initial_hash_value[] under !SHA2_SMALL, else it is unused
> if sha2.c is built with SHA2_SMALL, as pointed out by clang.
>
> This applies to the sha2.c copies in libc and libsa. The one in
> sys/crypto doesn't have SHA224.
>
> ok?
does anything anywhere
Hello,
On Fri, May 26, 2017 at 05:11:22PM +0200, Martin Pieuchot wrote:
> The global array of interfaces is accessed in the input path and need
> the NET_LOCK().
>
> Ok?
>
looks good to me.
OK sashan@
OK florian@
On Sat, May 27, 2017 at 10:15:45AM +0200, Florian Riehm wrote:
> Hi,
>
> This patch fixes the carp mode 'balancing ip-stealth'.
>
> Problem:
> System A
> carp1: flags=8843 mtu 1500
> lladdr 00:00:5e:00:01:01
> description: Carp-intern
> index 7 priority 15 llp
On Sat, May 27, 2017 at 10:15:45AM +0200, Florian Riehm wrote:
> We musst set the link state UP if at least one vhid is in state MASTER.
OK bluhm@
>
> friehm
>
> Index: netinet/ip_carp.c
> ===
> RCS file: /cvs/src/sys/netinet/ip_ca
Hello,
On Fri, May 26, 2017 at 04:54:57PM +0200, Martin Pieuchot wrote:
> The global list of softc is *not* accessed in the input path, so it
> doesn't need splnet().
>
> ioctl(2) handlers are already executed with the NET_LOCK() held, so
> splnet() is superfluous.
changes look good to me, but s
We already build the amd64 kernel with -ffreestanding. Should we
also do this for the boot code? It would make clang happier, and
conceptually it makes sense.
I've successfully tested the resulting gcc-built pxeboot(8) as well
as boot(8) with root on crypto softraid.
Index: arch/amd64/stand/Mak
This makes MULTIPROCESSOR kernels use copyin32(9) to guarantee
atomicity. This will break m88k GENERIC.MP; shouldn't be too
difficult to fix for someone whu understands m88k assembly.
Long-term goal is to implement copyin32(9) on all our architectures
and remove the #define.
ok?
Index: kern/sy
Move sha224_initial_hash_value[] under !SHA2_SMALL, else it is unused
if sha2.c is built with SHA2_SMALL, as pointed out by clang.
This applies to the sha2.c copies in libc and libsa. The one in
sys/crypto doesn't have SHA224.
ok?
Index: lib/libc/hash/sha2.c
reads ok
and yes on sthens suggestion
Peter Hessler(phess...@openbsd.org) on 2017.05.27 14:50:25 +0200:
> Allow us to receive our own AS paths from a neighbor.
>
> Like several of the related diffs, this also invites dragons and grues
> into your network.
>
> Probably needs the most love in th
On 2017/05/27 14:50, Peter Hessler wrote:
> Allow us to receive our own AS paths from a neighbor.
>
> Like several of the related diffs, this also invites dragons and grues
> into your network.
>
> Probably needs the most love in the man page, as usual ;).
>
> OK?
> Index: bgpd.conf.5
> ===
Allow us to receive our own AS paths from a neighbor.
Like several of the related diffs, this also invites dragons and grues
into your network.
Probably needs the most love in the man page, as usual ;).
OK?
Index: bgpd.conf.5
===
On Sat, May 27, 2017 at 11:55:48AM +0200, Frederic Cambus wrote:
> On Fri, May 26, 2017 at 05:37:25PM +0100, Jason McIntyre wrote:
>
> > the examples are expected to be typed as shown. so i think "name" should
> > be changed to an actual font name, and the description adjusted to show
> > that.
>
(benno_pflow_try4_task.diff)
ok?
diff --git sys/net/if_pflow.c sys/net/if_pflow.c
index a40fe23862b..20ab4e0e88d 100644
--- sys/net/if_pflow.c
+++ sys/net/if_pflow.c
@@ -67,6 +67,7 @@ struct pflowstats pflowstats;
void pflowattach(int);
intpflow_output(struct ifnet *ifp, struct mbuf
Hi tech@,
Add missing comments for CAN and SUB, for consistency.
Comments? OK?
Index: sys/dev/wscons/ascii.h
===
RCS file: /cvs/src/sys/dev/wscons/ascii.h,v
retrieving revision 1.5
diff -u -p -r1.5 ascii.h
--- sys/dev/wscons/ascii.h
On Sat, May 27, 2017 at 11:07:00AM +0200, Stefan Fritsch wrote:
> NVM_ADMIN_DEL_IOCQ does not need prp1 (just as NVM_ADMIN_DEL_IOSQ).
> Remove what is likely a cut'n'paste error from the *_ADD_* code.
>
> ok?
Tested on X270. No regression noticed.
>
> --- sys/dev/ic/nvme.c
> +++ sys/dev/ic/nvme
On Sat, May 27, 2017 at 11:05:42AM +0200, Stefan Fritsch wrote:
> On Fri, 26 May 2017, Claudio Jeker wrote:
>
> > Testing it on my X270. I get:
> > nvme0: unable to delete q, disabling
> >
> > Apart from that it seems to work (eventhough without inteldrm not very
> > helpful since I lose the disp
On Sat, May 27, 2017 at 09:38:28AM +0200, Claudio Jeker wrote:
> The MLEN check is for n which never has m_type MT_HEADER. n is allocated
> with MGET() and we want to put remain bytes into it. If that is bigger
> than MLEN it will not fit and we allocate a cluster.
You are right. I was confused b
Peter Hessler(phess...@openbsd.org) on 2017.05.27 12:17:56 +0200:
> Right now, we have 'neighbor-as' that can be used as magic syntax sugar
> for filter rules. I want to have magic syntax sugar for 'local-as',
> which expands to our ASN. It already plays nicely with the local-as
> diff I sent, an
Right now, we have 'neighbor-as' that can be used as magic syntax sugar
for filter rules. I want to have magic syntax sugar for 'local-as',
which expands to our ASN. It already plays nicely with the local-as
diff I sent, and can be committed in either order.
match in from $neighbor set commu
On 2017 May 27 (Sat) at 11:05:30 +0200 (+0200), Peter Hessler wrote:
:On 2016 Sep 26 (Mon) at 20:09:13 +0200 (+0200), Peter Hessler wrote:
::We already have a local AS saved per peer. Let's use it. This is very
::useful when one needs to change their local AS.
::
::"
::neighbor 192.0.2.1 {
::
On Fri, May 26, 2017 at 05:37:25PM +0100, Jason McIntyre wrote:
> the examples are expected to be typed as shown. so i think "name" should
> be changed to an actual font name, and the description adjusted to show that.
>
> otherwise it may be neccessary to document it more fully if it's not so
>
On 27/05/17(Sat) 11:34, Sebastian Benoit wrote:
> (benno_pflow_try3_2_locked_arg.diff)
>
> Introduce a 'locked' argument to sobind(), socreate() and soclose() to
> indicate if its called with a lock held.
> In pflow in the ioctl path, these functions can now be used without
> givin
On 27/05/17(Sat) 11:33, Sebastian Benoit wrote:
> (benno_pflow_try3_1_task.diff)
>
> move sending of pflow packet into a task, seperated from the data
> collection by a mbuf queue.
Comments inline.
> diff --git sys/net/if_pflow.c sys/net/if_pflow.c
> index a40fe23862b..8cfffa1e4e7 100644
Hello,
this looks good to me too.
OK sashan@
On Fri, May 26, 2017 at 04:22:29PM +0200, Martin Pieuchot wrote:
> The global list of softc is used in the input path and need to be
> protected by the NET_LOCK().
>
> ok?
>
(benno_pflow_try3_2_locked_arg.diff)
Introduce a 'locked' argument to sobind(), socreate() and soclose() to
indicate if its called with a lock held.
In pflow in the ioctl path, these functions can now be used without
giving up the netlock first.
diff --git sys/kern/sys_socket.c sy
(benno_pflow_try3_1_task.diff)
move sending of pflow packet into a task, seperated from the data
collection by a mbuf queue.
diff --git sys/net/if_pflow.c sys/net/if_pflow.c
index a40fe23862b..8cfffa1e4e7 100644
--- sys/net/if_pflow.c
+++ sys/net/if_pflow.c
@@ -67,6 +67,7 @@ struct pflows
> We get called twice on suspend, once with DVACT_SUSPEND and once with
> DVACT_POWERDOWN. So, here is a patch that does it like in ahci.c an does
> everything in the DVACT_POWERDOWN path and nothing in the DVACT_SUSPEND
> path.
I suppose so.
> @@ -469,6 +512,11 @@ nvme_activate(struct nvme_so
NVM_ADMIN_DEL_IOCQ does not need prp1 (just as NVM_ADMIN_DEL_IOSQ).
Remove what is likely a cut'n'paste error from the *_ADD_* code.
ok?
--- sys/dev/ic/nvme.c
+++ sys/dev/ic/nvme.c
@@ -1120,7 +1120,6 @@ nvme_q_delete(struct nvme_softc *sc, struct nvme_queue *q)
memset(&sqe, 0, sizeof(sq
On Fri, 26 May 2017, Claudio Jeker wrote:
> Testing it on my X270. I get:
> nvme0: unable to delete q, disabling
>
> Apart from that it seems to work (eventhough without inteldrm not very
> helpful since I lose the display).
Thanks for testing.
We get called twice on suspend, once with DVACT_SU
On 2016 Sep 26 (Mon) at 20:09:13 +0200 (+0200), Peter Hessler wrote:
:We already have a local AS saved per peer. Let's use it. This is very
:useful when one needs to change their local AS.
:
:"
:neighbor 192.0.2.1 {
: remote-as 65530
: local-as 131000
:}
:"
:
:OK?
:
Updated!
Now it
ok
Florian Obser(flor...@openbsd.org) on 2017.05.27 08:06:04 +:
> It's probably best to use basename(3) here like for the cert file
> instead of incorrectly handrolling the functionality ;)
>
> OK?
>
> diff --git main.c main.c
> index f8dc22c8756..9a7c3c5568c 100644
> --- main.c
> +++ main.c
Hi,
This patch fixes the carp mode 'balancing ip-stealth'.
Problem:
System A
carp1: flags=8843 mtu 1500
lladdr 00:00:5e:00:01:01
description: Carp-intern
index 7 priority 15 llprio 3
carp: carpdev vio2 advbase 1 balancing ip-stealth
state MASTER vhi
It's probably best to use basename(3) here like for the cert file
instead of incorrectly handrolling the functionality ;)
OK?
diff --git main.c main.c
index f8dc22c8756..9a7c3c5568c 100644
--- main.c
+++ main.c
@@ -128,23 +128,19 @@ main(int argc, char *argv[])
}
if(domain->chai
While looking at implementing disk hotplug for ahci, I saw that the ahci spec
suggests putting disconnected ports in listen mode, which allows the controller
to put the phy in a low power state while still allowing it to generate
interrupts
when a device is attached. As a first step towards hotpl
On Fri, May 19, 2017 at 04:19:57PM +0200, Ingo Schwarze wrote:
> Hi Anton,
>
> Anton Lindqvist wrote on Fri, May 19, 2017 at 02:11:37PM +0200:
>
> > Hi,
> > Another UTF-8 related bug reported by tb@. How to re-produce:
> >
> > 1. Enable vi mode:
> >
> >$ set -o vi
> >
> > 2. Input the foll
There is no need to have 3 buttons to enable pfkey protocol.
Especially a button that is only enabling the protocol.
Found by jmc@ because of the manpage bits
--
:wq Claudio
Index: share/man/man4/options.4
===
RCS file: /cvs/src/sha
On Fri, May 26, 2017 at 10:24:23PM +0200, Alexander Bluhm wrote:
> On Fri, May 26, 2017 at 05:38:07PM +0200, Claudio Jeker wrote:
> > m_makespace() is using an old way to build up the mbuf chain. By using
> > MCLGETI() a lot of the code can be simplified and results at least in my
> > opinion in be
On Thu, 30 Mar 2017 22:17:46 +1030
Jack Burton wrote:
> One of our sites has a need to require/verify TLS client certs,
> without the overhead & complexity of apache / nginx, etc.
>
> OpenBSD's httpd seemed the obvious candidate, and I figured that the
> feature would be useful to others too -- s
Found the other day:
`route -T 10 add default 127.0.0.1` fails with:
add net default: gateway 127.0.0.1: Network is unreachable
The workaround to make it work is to add -ifp lo0
The problem is in ifa_ifwithroute() which does an rtalloc() call to get
the gateway route and it fails for rtables (not
70 matches
Mail list logo
