On Tue, Nov 27, 2012 at 22:17 +0100, Mike Belopuhov wrote:
> apparently it works just fine. the number of clones is limited
> by the v_specbitmap which currently allows for 64 clones total
> (per system, not per process).
>
> apply the diff and see how all your apps are happi
apparently it works just fine. the number of clones is limited
by the v_specbitmap which currently allows for 64 clones total
(per system, not per process).
apply the diff and see how all your apps are happily using single
device node: /dev/bpf0.
thanks to pedro and thib for the spec_open_clone
As you might have already noticed, the diff was committed.
Thanks!
On Mon, Nov 26, 2012 at 8:34 PM, David Hill wrote:
> Any thoughts?
>
> On Tue, Nov 13, 2012 at 02:18:35PM -0500, David Hill wrote:
>>Hello -
>>
>>I originally asked mikeb if splnet was needed in net/pfkey.c. He added
>>onto my di
On Mon, Nov 26, 2012 at 8:54 PM, Mark Kettenis wrote:
>> Date: Mon, 26 Nov 2012 20:37:22 +0100
>> From: Mike Belopuhov
>>
>> We've discussed this with claudio and while there are might
>> be some other related issues with pr_usrreq and spls, I think
>> t
We've discussed this with claudio and while there are might
be some other related issues with pr_usrreq and spls, I think
this particular diff needs to be committed. OK?
On Tue, Nov 13, 2012 at 14:18 -0500, David Hill wrote:
> Hello -
>
> I originally asked mikeb if splnet was needed in net/pfke
On Sun, Nov 25, 2012 at 10:53 AM, Reyk Floeter wrote:
> Am Sonntag, 25. November 2012 schrieb Brad Smith :
>>
>> I don't think you're understanding what I am trying to say. I am not
>> switching
>> or removing anything per se. The MII framework already takes care of this
>> and
>> has for 12 years
On Fri, Nov 23, 2012 at 5:16 PM, Reyk Floeter wrote:
> On Fri, Nov 23, 2012 at 04:04:20PM +, Stuart Henderson wrote:
>> This adds an ioctl to retrieve if_hardmtu, and adds code to
>> display it via ifconfig hwfeatures.
>>
>> $ ifconfig em0 hwfeatures
>> em0: flags=8b43 mtu
>> 1500
>> hw
On Thu, Nov 22, 2012 at 16:05 +0100, Mike Belopuhov wrote:
> re pf bug on bugs@:
>
> apparently the crash is caused by the stack corruption that happens
> in pf_map_addr as it expects to get an array of struct pf_src_node
> pointers, not just one pointer. the bug was introd
On Thu, Nov 22, 2012 at 7:34 PM, mxb wrote:
> Hey tech@,
>
> I'll more than gladly test any diffs for .
>
> Regards,
> Maxim
>
both have rx checksums enabled.
tx checksums can't be enabled on i350 right now.
tx checksums on ix will need henning's diff to go in first.
re pf bug on bugs@:
apparently the crash is caused by the stack corruption that happens
in pf_map_addr as it expects to get an array of struct pf_src_node
pointers, not just one pointer. the bug was introduced about four
years ago, but somehow (stack layout?) went unnoticed.
the proper fix is to
On Wed, Nov 21, 2012 at 11:50 AM, Alexey E. Suslikov
wrote:
> Hello tech@.
>
> Following this
> http://blog.crowdstrike.com/2012/11/http-iframe-injecting-linux-rootkit.html
>
> Besides of doing "#option LKM", is there any other way to disable modload(8)?
>
> Cheers,
> Alexey
>
modules can't be lo
On Fri, Nov 16, 2012 at 18:44 +, sbienddr...@googlemail.com wrote:
> As requested.
>
> ddb{0}> uvm_fault(0xd0a50760, 0x0, 0, 1) -> e
> kernel: page fault trap, code=0
> Stopped atdb_read_bytes+0x14:movzbl0(%esi,%ecx,1),%eax
> ddb{0}> ddb{0}> db_read_bytes(0,1,f58ded94,0,2) at db_re
On Thu, Nov 15, 2012 at 5:53 PM, Reyk Floeter wrote:
> On Thu, Nov 15, 2012 at 5:11 PM, Marc Espie wrote:
>> external people regularly ask "but why you don't want to use GNU/m4 GNU/make
>> GNU/whatever ?"
>>
>
> External people seem to ask weird questions.
>
> I just had to dig into autoconf/aut
On Thu, Nov 15, 2012 at 12:18 PM, Martin Pieuchot
wrote:
> Simple diff to print UniNorth's revision so it becomes easy to know
> which AGP chipset people have without looking at the 'eeprom -p' output.
>
> before:
> memc0 at mainbus0: uni-n
> after:
> memc0 at mainbus0: uni-n rev 0
somehow splnet was forgotten in the receive path which leads to
the following splassert:
splassert: if_start: want 7 have 5
Starting stack trace...
splassert_check() at splassert_check+0x7e
if_start() at if_start+0x32
ieee80211_deliver_data() at ieee80211_deliver_data+0x1bd
ieee80211_decap() at ie
On Tue, Nov 06, 2012 at 12:15 +0100, Stefan Sperling wrote:
> On Tue, Nov 06, 2012 at 11:32:19AM +0100, Mike Belopuhov wrote:
> > attach fails early in case there's no firmware, but
> > athn_detach does ieee80211_ifdetach and if_detach
> > regardless of whether ifne
the diff syncs up some pieces with freebsd and fixes some bugs
(notably the one with the missing braces :)
ixgbe_phy_aq is removed cause it's a x540 which is handled now
by the separate module ixgbe_x540.c.
i'd like to get rid of this change quickly as it somewhat clashes
with other pending chang
Remove code that is confusing and wrong: we don't want to disable
rx checksumming if IFCAP_CSUM_IPv4 is set. Whatever this code used
to mean in FreeBSD it doesn't do any good here. IXGBE_RXCSUM_PCSD
is set when you do RSS (and is correctly handled in the visible RSS
chunk below).
OK?
Index: if_
attach fails early in case there's no firmware, but
athn_detach does ieee80211_ifdetach and if_detach
regardless of whether ifnet part got setup correctly
leading to a free of an unallocated memory and a
panic.
the following diff follows an established practice
in the other drivers and fixes the p
On Sun, Nov 4, 2012 at 2:21 PM, Christian Weisgerber wrote:
> Christian Weisgerber wrote:
>
>> Like bge(4), we previously couldn't enable TCP/UDP transmit checksum
>> offload on em(4). We can now.
>>
>> Works fine here on
>> em0 at pci1 dev 1 function 0 "Intel PRO/1000MT (82540EM)" rev 0x02
>>
Hi,
Flow director is described as "a large number of flow affinity
filters that direct receive packets by their flows to queues
for classification, load balancing, and matching between flows
and CPU cores." As we don't support anything like this the code
(that is still compiled in but is not call
On Tue, Oct 30, 2012 at 2:58 PM, Christiano F. Haesbaert
wrote:
> On 30 October 2012 14:36, rustyBSD wrote:
>> MMmhh...
>>
>> == /usr/src/usr.bin/mg/dired.c ==
>> Go look the line 729:
>>
>> if ((fopen(dname,"r")) == NULL) {
>> ...
>>
>> Now you can cry
>>
>
> What is your point ?
>
Hi,
In case you have a ubsec(4) or a hifn(4) crypto-accelerator
plugged in to an amd64 with >4Gb of RAM, this diff should fix
a crash for you in case you have enabled kern.usercrypto by
doing buffer allocations of the "DMA-accessible memory" as
most of the old^Wsupported crypto-accelerators can't
Hi,
Unless Miod objects, I'd like to kill this leftover.
We don't seem to be ditching AF_* defines, so AF_HYLINK
glory will be preserved. OK?
diff --git sys/netinet/in_proto.c sys/netinet/in_proto.c
index c146e49..0f7845f 100644
--- sys/netinet/in_proto.c
+++ sys/netinet/in_proto.c
@@ -322,25 +3
On Tue, Oct 9, 2012 at 1:23 PM, Reyk Floeter wrote:
> Hi,
>
> the iked.conf(5) manpage says: "If srcid is omitted, the default is to
> use the hostname of the local machine, see hostname(1) to set or print
> the hostname." This was true but I broke it with a commit about two
> years ago :(
>
> The
diff looks good to me. ok mikeb
On Tue, Sep 18, 2012 at 12:49 +0200, Gerhard Roth wrote:
> Hi,
>
> we observed mysterious panics while stopping vr interfaces. This was due
> to vr_stop() trying to stop the transfers but then not waiting for them
> to really finish but rather remove their DMA buf
On Mon, Sep 17, 2012 at 5:03 PM, Henning Brauer wrote:
> * mxb [2012-09-10 17:51]:
>> is there any plans to expand 'tagged' keyword in PF into list?
>
> not that I am aware of, but it would make sense to have list expansion
> there as well.
>
would it mean "tagged foo AND bar" or "tagged foo OR
On Wed, Aug 22, 2012 at 3:55 PM, Stefan Fritsch wrote:
> Here is the current version. Changes from v6:
>
> - net: rename device vioif -> vio
> - net: don't use watchdog for cleaning up tx descriptors
> (in fact, remove watchdog completely)
> - net: various fixes for RingEventIdx mode; should fix
On Fri, Aug 24, 2012 at 12:55 AM, Stefan Sperling wrote:
> For some bizarre reason emacs links to libossaudio instead ;)
>
yeah, an sndio backend is yet to be written...
another reason to migrate to the superior pulse-audio framework!
On Tue, Aug 21, 2012 at 1:35 PM, Stefan Fritsch wrote:
>
> What do yo suggest then? Use a separate timeout? Accept the lower
> performance?
>
what is the performance difference in terms of pps on small packets
and Mbps for large are we talking about?
> Resetting the device in the watchdog has an
On Mon, Aug 20, 2012 at 2:42 PM, Stefan Fritsch wrote:
> On Mon, 20 Aug 2012, Mike Belopuhov wrote:
>>
>> On Mon, Aug 20, 2012 at 1:42 PM, Stefan Fritsch wrote:
>>>
>>> Because the interrupt load increases by approx. 7000 per second. I think
>>> the
>
On Mon, Aug 20, 2012 at 1:42 PM, Stefan Fritsch wrote:
> On Mon, 20 Aug 2012, Mike Belopuhov wrote:
>>>
>>> Always using the tx interrupt decreases performance significantly. On my
>>> test system (in MBytes/s):
>>>
>>> Sending UDP from 75 t
On Mon, Aug 20, 2012 at 1:28 PM, Stefan Fritsch wrote:
> On Fri, 17 Aug 2012, Mike Belopuhov wrote:
>>
>> On Fri, Aug 17, 2012 at 1:17 PM, Stefan Fritsch wrote:
>>>
>>> On Thursday 16 August 2012, Mike Belopuhov wrote:
>>>>
>>>> cou
On Fri, Aug 17, 2012 at 1:17 PM, Stefan Fritsch wrote:
> On Thursday 16 August 2012, Mike Belopuhov wrote:
>> could you please tell me if you're using tx interrupt or not?
>> if yes, why do you need to have a watchdog code implement
>> a txeof path and not the a
On Thu, Aug 16, 2012 at 4:49 PM, Stefan Fritsch wrote:
> Here is the next iteration:
>
> Changes from V5:
> - Remove unused code for mutexes
> - Remove explicit hardware memory barriers. If the host uses non-temporal
> memory operations like movnti, it will hopefully issue the memory
> barrier
On Wed, Aug 15, 2012 at 5:36 PM, Mike Belopuhov wrote:
> On Wed, Aug 15, 2012 at 5:02 PM, Ted Unangst wrote:
>> The acpihpet timer is, in my testing, lots better than the acpitimer.
>> Faster to read and more precise. They should not have the same quality
>> value. Doubl
On Wed, Aug 15, 2012 at 5:02 PM, Ted Unangst wrote:
> The acpihpet timer is, in my testing, lots better than the acpitimer.
> Faster to read and more precise. They should not have the same quality
> value. Double acpihpet.
>
as long as acpi subsystem attaches acpitimer earlier we don't
need tha
On Mon, Aug 13, 2012 at 5:41 PM, Mike Belopuhov wrote:
> On Mon, Aug 13, 2012 at 5:36 PM, Mike Belopuhov wrote:
>> On Mon, Aug 13, 2012 at 5:30 PM, Stefan Fritsch
>> wrote:
>>> On Monday 13 August 2012 17:07:41 you wrote:
>>>> > * Note: the i386 does
On Mon, Aug 13, 2012 at 5:36 PM, Mike Belopuhov wrote:
> On Mon, Aug 13, 2012 at 5:30 PM, Stefan Fritsch
> wrote:
>> On Monday 13 August 2012 17:07:41 you wrote:
>>> > * Note: the i386 does not currently require barriers, but we must
>>> > * provide the f
On Mon, Aug 13, 2012 at 5:30 PM, Stefan Fritsch wrote:
> On Monday 13 August 2012 17:07:41 you wrote:
>> > * Note: the i386 does not currently require barriers, but we must
>> > * provide the flags to MI code.
>> >
>> > This is not correct for virtio. We need a memory barrier.
>>
>> sure, copy i
On Mon, Aug 13, 2012 at 4:24 PM, Stefan Fritsch wrote:
> Hi,
>
> here is the next iteration of my patch.
>
> Changes from V4 include:
>
> - virtio: support RING_EVENT_IDX
> - virtio: use lfence/sfence because __sync_synchronize() is broken
> on gcc < 4.4
> - net: rework memory ha
On Mon, Aug 13, 2012 at 4:24 PM, Stefan Fritsch wrote:
> - virtio: use lfence/sfence because __sync_synchronize() is broken
> on gcc < 4.4
please don't. use bus_space_barrier.
Driver was obtained from FreeBSD, but was heavily massaged.
Test reports (if any) and OKs are welcome (:
The diff is a just too large to inline, please find it here:
http://theapt.org/~mike/files/oce.diff
Cheers
Obtained from FreeBSD. Test reports (on any other ix cards as well)
and OKs are welcome (:
The diff is a bit too large to inline, please find it here:
http://theapt.org/~mike/files/x540.diff
Cheers
On Tue, Jul 31, 2012 at 18:20 +0200, Mike Belopuhov wrote:
> Hi,
>
> As it was recently discovered some newer machines have 40 bit
> physical address size which means that the math that is good
> for 36 bit paddr just doesn't cut it anymore. In fact the
> memory region
Hi,
As it was recently discovered some newer machines have 40 bit
physical address size which means that the math that is good
for 36 bit paddr just doesn't cut it anymore. In fact the
memory region lengths we program into the MSRs are completely
off.
The following diff has some code ported from
On Sun, Jul 29, 2012 at 5:47 PM, Hrvoje Popovski wrote:
>
> Hello,
>
> i have tested your patches with torrent box behind 2 firewalls (nat,
> carp,pfsync). One firewall is patched and second box isn't. Both of them
> quite often export flows with duration of 429496*
>
is it possible that these flo
On Sun, Jul 22, 2012 at 11:03 +0100, Stuart Henderson wrote:
> On 2012/07/21 21:52, Mike Belopuhov wrote:
> > On Sat, Jul 21, 2012 at 19:25 +0100, Stuart Henderson wrote:
> > > On 2012/07/21 18:49, Mike Belopuhov wrote:
> > > > this sppp_clear_ip_addrs_args dance lo
On Wed, Jul 18, 2012 at 4:16 PM, Gerhard Roth wrote:
>> same here, wouldn't it be possible to match the ipsec.conf grammar and
>> ignore the SNMPv3 naming a bit?
>>
>> auth hmac-sha1 authkey "fooobar" enc aes enckey "dkjdkj"
>> - instead of -
>> hmac sha authpass "foobar" cipher aes privpass "dkjdk
On Wed, Jul 11, 2012 at 13:13 +0200, Henning Brauer wrote:
> ..because now you had to initialize both set_prio in pf_rule to it
> everywhere. we did that, at least in some parts of our tree...
> problem being of course that 0 is a valid value there and can\t easily
> be used as "don't touch" indica
On Fri, Jun 29, 2012 at 7:20 PM, Christian Weisgerber
wrote:
> And the corresponding bits for iked(8). I don't think we need to keep
> "aes-ctr" for compatibility there, do we?
>
i'm a sitting on a fence a little bit. i don't really mind either way
hence my OK for the diff.
On Fri, Jun 29, 2012 at 1:36 PM, Henning Brauer
wrote:
> now it's very unclear what your actual problem is - the struct is
> called ifbreq and used in a number of places, most notably of course
> the ioctls.
>
he's trying to add patterns to the mac address matching code
and pretends to be done wi
On Fri, Jun 29, 2012 at 5:41 AM, Lawrence Teo wrote:
> This diff adds support for the Winbond W83627UHG chip, as found on the
> Lanner FW-7539 appliance. This diff was ported from DragonFly BSD:
>
>
http://gitweb.dragonflybsd.org/dragonfly.git/commitdiff/245ec76bc1613b22cf282
526fa9931e4c16e4237
On Fri, Jun 29, 2012 at 12:09 AM, Christian Weisgerber
wrote:
> Is there a particular reason we only support AES-128-CTR ("AESCTR")
> with isakmpd(8), but not the 192- and 256-bit variants like we do
> for AES-CBC and AES-GCM?
>
> If not, and I assume it's just a historic oversight, how about this
On Thu, Jun 28, 2012 at 16:07 +0200, Peter Hessler wrote:
> So we have been running with this diff in the office, and it is working
> well for us.
>
>
> bnx0 at pci7 dev 0 function 0 "Broadcom BCM5709" rev 0x20: apic 0 int 6
> bnx0: address 1c:c1:de:e4:5e:b0
> brgphy0 at bnx0 phy 1: BCM5709 10/10
hi,
one of the last pfsync "improvements" (r1.180) by yours truly
introduced some nasty regressions for those running pfsync over
the crossover cable. i apologize for all the inconveniences.
peter hessler, kapeatanakis giannis and myself have been trying
very hard to get this fixed.
we have iden
On Sat, Jun 9, 2012 at 12:41 PM, Stuart Henderson
wrote:
> On 2012/06/09 14:09, Alexander Polakov wrote:
>> > I appreciate that the defaults would stay the same, but really,
>> > what is the point in doing this? cwm can't be everything to
>> > everyone.
>>
>> The point is: when you want tiling fr
On Sat, Jun 2, 2012 at 11:02 AM, Stuart Henderson wrote:
> As with other entries in /etc/services, this adds to the ports to
> net.inet.tcp.baddynamic, preventing using the associated ports as a
> valid dynamic source port number.
>
> OK?
>
sounds like a reasonable thing to do. ok mikeb
On Thu, May 31, 2012 at 10:53 +0200, Paul de Weerd wrote:
> I've recently bought a new Sony laptop, VPCZ23C5E. Some things on
> this machine work fine (sound, suspend, wired network), some things
> don't (X, resume, wireless). One of the problems I have is that
> -current doesn't find any disks b
gt;>
>> Yes, I got a couple of private mails stating the same, which explains
>> why I hadn't seen any others. Anyway, Mike Belopuhov suggested pchpcib
>> as an alternative name so I've renamed the driver to that along with
>> the man pages, etc.
>
> I'
t; why I hadn't seen any others. Anyway, Mike Belopuhov suggested pchpcib
> as an alternative name so I've renamed the driver to that along with
> the man pages, etc.
>
> Here's the dmesg now:
>
> ---8<---
> pchpcib0 at pci0 dev 31 function 0 "Intel E600
esn is 90% supported in iked and this just adds some missing pieces.
works with strongswan and windows (although windows doesn't propose
ESN it's also a good test).
diff --git sbin/iked/iked.h sbin/iked/iked.h
index dc31da4..ce25bb3 100644
--- sbin/iked/iked.h
+++ sbin/iked/iked.h
@@ -156,6 +156,7
this allows ipsecctl to print 'esn' flag nicely:
# ipsecctl -ssa
ah tunnel from 10.2.0.6 to 10.2.0.5 esn spi 0xbeefdead auth hmac-md5
ah tunnel from 10.2.0.5 to 10.2.0.6 esn spi 0xdeadbeef auth hmac-md5
there's no real value in confi
the following diff adds support for the 64-bit extended sequence
numbers specified by RFC 4303. briefly, interesting points are:
- only lower 32-bit part is transmitted so packet is not changed;
- high-oder part is kept private and is used differently by HMAC
and other MAC hashes (such as GC
On Thu, May 10, 2012 at 15:22 +0200, Erik Lax wrote:
> Hi,
>
> I came across a real-world scenario where network traffic stopped due to
> interrupt starvation with the em driver, tested with a few different
> cards (on a network/interface where no packets were received). This is
> what caused it.
On Sat, May 05, 2012 at 16:07 +, Miod Vallat wrote:
> > add support for the WSATTR_REVERSE attribute to sti(4) to make mg,
> > less and others a bit prettier. tested on visualize fx 10 pro where
> > this email was written.
> >
> > tests on other sti's are welcome as well as oks (:
>
> This is
add support for the WSATTR_REVERSE attribute to sti(4) to make mg,
less and others a bit prettier. tested on visualize fx 10 pro where
this email was written.
tests on other sti's are welcome as well as oks (:
Index: dev/ic/sti.c
===
sequence numbers in esp and ah are unsigned numbers, so we
shouldn't ever be printing them with a minus sign :-)
ok?
Index: usr.sbin/tcpdump/print-ipsec.c
===
RCS file: /home/cvs/src/usr.sbin/tcpdump/print-ipsec.c,v
retrieving revisi
there's no apparent need to take a file descriptor table lock
before we've done allocating pipe structures and buffers. ok?
Index: sys/kern/sys_pipe.c
===
RCS file: /home/cvs/src/sys/kern/sys_pipe.c,v
retrieving revision 1.61
diff -u
I'm still looking for OKs on this.
On Mon, Apr 02, 2012 at 14:42 +0200, Mike Belopuhov wrote:
> When route priorities were added to the RA/redirects code, all
> of them got set to RTP_CONNECTED. Most importantly, rt6_flush
> was made to flush only RTP_CONNECTED routes.
I'm still looking for OKs on this.
On Mon, Apr 02, 2012 at 14:28 +0200, Mike Belopuhov wrote:
> RFC 4861 says:
>
> Target link-layer address
> The link-layer address for the target, i.e., the
> sender of the advertisement.
I'm still looking for OKs on this.
On Mon, Apr 02, 2012 at 14:18 +0200, Mike Belopuhov wrote:
> So when TLLA/SLLA is not specified we get into a weird state where
> we don't know a LL address and don't ask for it. Apparently, this
> is not what some IPv6 compliance te
On Sat, Apr 14, 2012 at 13:35 +0200, Mike Belopuhov wrote:
> some software needs fully functional per-thread %gs base address,
> hence the diff. limited testing shows no regressions.
> INTR_RESTORE_SELECTORS changes include:
> - %r11 is used to store curpcb instead of rdx that
On Sat, Apr 14, 2012 at 3:42 PM, Mark Kettenis
wrote:
>> > some software needs fully functional per-thread %gs base address,
>> > hence the diff. limited testing shows no regressions.
>> > INTR_RESTORE_SELECTORS changes include:
>> > - %r11 is used to store curpcb instead of rdx that is used
Er EFAULT
+.Fa base
+points outside the process's allocated address space.
+.El
+.Sh SEE ALSO
+.Xr __tfork 2 ,
+.Xr fork 2
Index: lib/libarch/amd64/amd64_get_gsbase.c
===
RCS file: lib/libarch/amd64/amd64_get_gsbase.c
diff -N
On Thu, Apr 12, 2012 at 22:40 +0900, SASANO Takayoshi wrote:
> Hello,
>
> > I have a system with an asrock a75m-itx motherboard and an amd a6-3500
> > processor. I notice there is a 40 second delay after the 'ahci0 at pci0
> > dev 17
> > ...' line.
>
> I have IBM's ThinkPad Edge E525, AMD A8
On Wed, Apr 11, 2012 at 19:30 +0200, Mike Belopuhov wrote:
> This moves the accounting flag to struct process where it belongs.
> The main rationale is that accounting flags are being collected on
> the program exit. Ok?
>
don't set AFORK when forking threads pointed out by
This moves the accounting flag to struct process where it belongs.
The main rationale is that accounting flags are being collected on
the program exit. Ok?
Index: kern/kern_acct.c
===
RCS file: /cvs/src/sys/kern/kern_acct.c,v
retriev
guenther@ has pointed out that sometimes suser is passed an incorrect
flags argument. Currently, there's only one flag that's defined:
SUSER_NOACCT (0x1). Unfortunately it aligns too well with AFORK flag
used with p_acflag. This diff cleans up the tree. OK?
Index: dev/pci/if_san_obsd.c
==
On Thu, Apr 5, 2012 at 9:21 AM, Alexander Polakov wrote:
> * Mike Belopuhov [120404 17:51]:
>> i agree that this is a great find. i don't really like the diff though.
>> i see no point in introducing this macro. what do others think?
>
> Your diff looks better to me.
On Wed, Apr 04, 2012 at 14:42 +0400, Alexander Polakov wrote:
> This is a diff from NetBSD pr.34583:
> http://gnats.netbsd.org/cgi-bin/query-pr-single.pl?number=34583
>
> Quoting the author:
>
> "I noticed that when writing large file (hundreds of megabytes)
> to an msdos disk, the wr
On Tue, Apr 03, 2012 at 13:15 +0200, Mike Belopuhov wrote:
> According to the RFC 2460 and RFC 5095, ICMP Parameter Problem, Code 0
> should be sent, not Code 2. OK?
>
> (while at the spot, kill a trailing whitespace)
>
we have improved the diff with sperreault@ to set a pointe
According to the RFC 2460 and RFC 5095, ICMP Parameter Problem, Code 0
should be sent, not Code 2. OK?
(while at the spot, kill a trailing whitespace)
diff --git sys/netinet6/ip6_input.c sys/netinet6/ip6_input.c
index 6afde60..e554ad7 100644
--- sys/netinet6/ip6_input.c
+++ sys/netinet6/ip6_inpu
When route priorities were added to the RA/redirects code, all
of them got set to RTP_CONNECTED. Most importantly, rt6_flush
was made to flush only RTP_CONNECTED routes. The purpose of
rt6_flush is to flush *all* routes going via a specified gw.
Now the problem is that the code in rtrequest1/mpa
RFC 4861 says:
Target link-layer address
The link-layer address for the target, i.e., the
sender of the advertisement. This option MUST be
included on link layers that have addresses when
responding to multi
So when TLLA/SLLA is not specified we get into a weird state where
we don't know a LL address and don't ask for it. Apparently, this
is not what some IPv6 compliance tests expect. For example Tahi
really wants you to ask for a LLA by sending a neighbor solicitation
which means state should be INC
no need to do route lookup and perform other expensive checks
if redirect destination is not unicast. this just moves this
check a bit up. ok?
diff --git sys/netinet6/icmp6.c sys/netinet6/icmp6.c
index 1501766..40361f4 100644
--- sys/netinet6/icmp6.c
+++ sys/netinet6/icmp6.c
@@ -2212,6 +2212,13
On Thu, Mar 15, 2012 at 19:34 +0900, YASUOKA Masahiko wrote:
> On Thu, 15 Mar 2012 09:16:42 +0100
> Alexander Bluhm wrote:
> > On Wed, Mar 14, 2012 at 03:32:08PM +0900, YASUOKA Masahiko wrote:
> >> In ip_input(), there is a filter to disable all packets to 127.0.0.0/27.
> >> That filter drops a pa
we had this discussion recently and the outcome is: show the *actual* diff
that "was close to getting committed". so far there hasn't been any.
On Thu, Mar 15, 2012 at 11:31 AM, Mitja MuE>eniD
wrote:
> Hey,
>
> there was a longish thread here roughly 4 years ago titled "[patch] pf PPTP
> nat pa
On Tue, Mar 13, 2012 at 16:51 +, Stuart Henderson wrote:
> On 2012/03/13 16:41, Mike Belopuhov wrote:
> > this allows us to bridge two ipv4 networks over an ipv6 link with gif.
> > any objections? oks? the change was well tested by phessler.
> >
>
> Not that it
this allows us to bridge two ipv4 networks over an ipv6 link with gif.
any objections? oks? the change was well tested by phessler.
diff --git netinet6/in6_proto.c netinet6/in6_proto.c
index 7575055..8cf3de4 100644
--- netinet6/in6_proto.c
+++ netinet6/in6_proto.c
@@ -105,6 +105,7 @@
#include
On Tue, Mar 13, 2012 at 15:11 +0200, Joosep wrote:
> On Tue, Mar 13, 2012 at 2:20 PM, Mike Belopuhov wrote:
>
> > heh. does this diff make a difference?
> >
> Hi!
>
> No difference. Previously given packet sizes were with SHA1+3DES not with
> AES as said earlier.
s.
> Unfortunately i didn't had the chance to pinpoint the exact packet sizes. I
> haven't tested with MD5 yet.
>
> With the diff, previously sent off-list by Mike, there was no verification
> errors(with SHA1+AES and SHA1+3DES) in 24 hours.
>
> Thanks,
>
ubsec gets stuck after receiving a packet of a particular
length. for example, in my tests it was a tcp packet with
59 bytes of payload. interestingly, something breaks horribly
in the hardware and it stops processing any other packets.
the exact cause is believed to be a difference between the
s
ftp-proxy has all the code to support "on rdomain" feature
in place, just not used. the change below uses an rdomain
obtained via the SO_RTABLE socket option of the accepted
socket. OK?
Index: filter.c
===
RCS file: /home/cvs/src/us
On Sun, Feb 26, 2012 at 17:00 +0100, Mike Belopuhov wrote:
> On Sat, Feb 25, 2012 at 15:46 +0100, Mike Belopuhov wrote:
> > hi,
> >
> > here's a diff that adds support for the SMBus controller found
> > on AMD CS5536:
> >
> > glxpcib0 at pci0 dev 15
This is a well-known from thib and dlg originally with a length fix
from yours truly, that marginally doubles througput (from 300kpps to
500-600kpps on selected hardware).
The idea is to save an IP header and 8 bytes of payload (good enough
for tcp state tracking) instead of recommended 68 bytes.
On Sat, Feb 25, 2012 at 15:46 +0100, Mike Belopuhov wrote:
> hi,
>
> here's a diff that adds support for the SMBus controller found
> on AMD CS5536:
>
> glxpcib0 at pci0 dev 15 function 0 "AMD CS5536 ISA" rev 0x03: rev 3, 32-bit
> 3579545Hz timer, watchdo
hi,
here's a diff that adds support for the SMBus controller found
on AMD CS5536:
glxpcib0 at pci0 dev 15 function 0 "AMD CS5536 ISA" rev 0x03: rev 3, 32-bit
3579545Hz timer, watchdog, gpio, i2c
gpio0 at glxpcib0: 32 pins
iic0 at glxpcib0
maxtmp0 at iic0 addr 0x4c: lm86
# sysctl hw.sensors
hw.s
Still looking for ok's...
On Sat, Feb 11, 2012 at 01:27 +0100, Mike Belopuhov wrote:
> Hi,
>
> As it became evident, ix is driven by Low Latency Interrupts
> on 82599 to do all sorts of processing instead of the regular
> Rx/Tx queue interrupts. LLI is an additional facilit
801 - 900 of 1074 matches
Mail list logo