snmpd improvements

2016-12-21 Thread Franco Fichtner
Hi, Switching from net-snmp to OpenBSD's snmpd raised two issues and I'd like to know if they make sense to address: A pid file is missing. Would a patch for this be accepted? The snmpd.conf can contain static values. If these values are rewritten/changed over time by rewriting the config,

ospf6d: handle interface MTU changes

2016-12-21 Thread Jeremie Courreges-Anglas
Hi, After ospfd here's a diff to make ospf6d refresh his view of an interface's MTU at runtime. This needs a fresh kernel. The parent should pass the IFINFO message to its children first, and then decide to react to a possible interface change. Like for ospfd, the engine runs the FSM only if

Re: sleep.1: "while true;" -> "while :;"

2016-12-21 Thread Jason McIntyre
On Wed, Dec 21, 2016 at 09:59:18AM +0100, Michal Mazurek wrote: > On 22:49:38, 20.12.16, Jason McIntyre wrote: > > On Tue, Dec 20, 2016 at 10:58:40PM +0100, Michal Mazurek wrote: > > > While there is nothing wrong with "while true", "while :" is better > > > and used a lot more often in the source

vndcompress et al import?

2016-12-21 Thread Franco Fichtner
Hi, Is anyone aware or interested in porting vndcompress et al from NetBSD to OpenBSD? Is there any technical reason against inclusion? We have a budget for this. If anyone is interested please let me know. Cheers, Franco

Re: sleep.1: "while true;" -> "while :;"

2016-12-21 Thread Michal Mazurek
On 22:49:38, 20.12.16, Jason McIntyre wrote: > On Tue, Dec 20, 2016 at 10:58:40PM +0100, Michal Mazurek wrote: > > While there is nothing wrong with "while true", "while :" is better > > and used a lot more often in the source tree. > > > > OK? > > > > i'm not sure why "while :" is better in

ND6 and splsoftnet()

2016-12-21 Thread Martin Pieuchot
This diff get rids of all splsoftet() in nd6 code. I converted timers to the NET_LOCK() even if ip6_output() doesn't assert for it for the moment. That's simply the way to go. Also because these functions iterates on global data structures that will need the NET_LOCK() or a rewrite. ok?

Re: if attach/detach netlocks

2016-12-21 Thread Alexander Bluhm
On Wed, Dec 21, 2016 at 12:45:37PM +0100, Mike Belopuhov wrote: > Anyways, OK for the diff below? OK bluhm@ > diff --git sys/net/if_pflow.c sys/net/if_pflow.c > index 0df0b69fd8a..8592d98a743 100644 > --- sys/net/if_pflow.c > +++ sys/net/if_pflow.c > @@ -265,14 +265,11 @@

clang amd64 libm: declare copysign() correctly

2016-12-21 Thread Christian Weisgerber
libm uses copysign() and copysignf() internally, but fails to declare the amd64 assembly versions that way. When built with clang, this results in undefined references to _libm_copysign etc. Presumably gcc replaces those calls to copysign with a builtin, but clang doesn't. Index:

Re: clang amd64 libm: declare copysign() correctly

2016-12-21 Thread Mark Kettenis
> Date: Wed, 21 Dec 2016 13:28:26 +0100 > From: Christian Weisgerber > > libm uses copysign() and copysignf() internally, but fails to declare > the amd64 assembly versions that way. When built with clang, this > results in undefined references to _libm_copysign etc. > >

Re: if attach/detach netlocks

2016-12-21 Thread Mike Belopuhov
On Wed, Dec 21, 2016 at 00:13 +0100, Alexander Bluhm wrote: > On Tue, Dec 20, 2016 at 06:47:31PM +0100, Mike Belopuhov wrote: > > @@ -1109,11 +1115,16 @@ if_clone_destroy(const char *name) > > s = splnet(); > > if_down(ifp); > > splx(s); > > } > > > > -

Re: sleep.1: "while true;" -> "while :;"

2016-12-21 Thread Christian Weisgerber
On 2016-12-20, Jason McIntyre wrote: > i'm not sure why "while :" is better in this example, but "while true" > is clearer, i think. allowing for differences in preference, is their a > good reason to change what's there? I agree that "true" is clearer and I see no reason to

NET_LOCK() recursion during netboot

2016-12-21 Thread Visa Hankala
Network boot triggers a netlock recursion: panic rw_enter sosend <- NET_LOCK() nfs_send nfs_request nfs_lookup VOP_LOOKUP vfs_lookup namei unp_connect uipc_usrreq soconnect <- NET_LOCK() sys_connect An XXXSMP workaround seems appropriate here. OK? Index: kern/uipc_usrreq.c

Re: if attach/detach netlocks

2016-12-21 Thread Alexander Bluhm
On Tue, Dec 20, 2016 at 07:33:35PM +0100, Martin Pieuchot wrote: > On 20/12/16(Tue) 18:47, Mike Belopuhov wrote: > > On Tue, Dec 20, 2016 at 17:06 +0100, Martin Pieuchot wrote: > > > > > > You'll need to release the lock before calling ifc->ifc_create in > > > if_clone_create() and do the same

Re: ND6 and splsoftnet()

2016-12-21 Thread Alexander Bluhm
On Wed, Dec 21, 2016 at 05:21:53PM +0100, Alexander Bluhm wrote: > > @@ -430,15 +430,16 @@ nd6_llinfo_timer(void *arg) > > void > > nd6_timer_work(void *null) > > { > > - int s; > > struct nd_defrouter *dr, *ndr; > > struct nd_prefix *pr, *npr; > > struct in6_ifaddr *ia6, *nia6; >

Re: NET_LOCK() recursion during netboot

2016-12-21 Thread Alexander Bluhm
On Wed, Dec 21, 2016 at 04:52:32PM +, Visa Hankala wrote: > Network boot triggers a netlock recursion: > > panic > rw_enter > sosend <- NET_LOCK() > nfs_send > nfs_request > nfs_lookup > VOP_LOOKUP > vfs_lookup > namei > unp_connect > uipc_usrreq > soconnect <- NET_LOCK() > sys_connect > >

Re: snmpd improvements

2016-12-21 Thread Jan Klemkow
On Wed, Dec 21, 2016 at 10:40:48AM +0100, Franco Fichtner wrote: > Hi, > > Switching from net-snmp to OpenBSD's snmpd raised two > issues and I'd like to know if they make sense to address: > > A pid file is missing. Would a patch for this be accepted? As far as I know, the OpenBSD project

Re: ND6 and splsoftnet()

2016-12-21 Thread Martin Pieuchot
On 21/12/16(Wed) 18:22, Alexander Bluhm wrote: > On Wed, Dec 21, 2016 at 05:21:53PM +0100, Alexander Bluhm wrote: > > > @@ -430,15 +430,16 @@ nd6_llinfo_timer(void *arg) > > > void > > > nd6_timer_work(void *null) > > > { > > > - int s; > > > struct nd_defrouter *dr, *ndr; > > > struct

Re: clang amd64 libm: declare copysign() correctly

2016-12-21 Thread Philip Guenther
On Wed, Dec 21, 2016 at 4:49 AM, Mark Kettenis wrote: >> Date: Wed, 21 Dec 2016 13:28:26 +0100 >> From: Christian Weisgerber >> >> libm uses copysign() and copysignf() internally, but fails to declare >> the amd64 assembly versions that way. When

Re: BFD: route get and route monitor

2016-12-21 Thread Sebastian Benoit
Hrvoje Popovski(hrv...@srce.hr) on 2016.12.21 22:03:56 +0100: > On 17.12.2016. 14:05, Peter Hessler wrote: > > Updated output, requested by Theo. A normal get will show just the bfd > > state, use "-bfd" to get all of the information. > > > > OK? > > > > $ route -n get 203.0.113.9 > >route

Re: BFD: route get and route monitor

2016-12-21 Thread Hrvoje Popovski
On 17.12.2016. 14:05, Peter Hessler wrote: > Updated output, requested by Theo. A normal get will show just the bfd > state, use "-bfd" to get all of the information. > > OK? > > $ route -n get 203.0.113.9 >route to: 203.0.113.9 > destination: 203.0.113.9 >mask: 255.255.255.255 >

Re: n_time in trpt(8)

2016-12-21 Thread Stefan Sperling
On Wed, Dec 21, 2016 at 10:01:28PM -0500, Lawrence Teo wrote: > In 2014, mpi@ substituted n_time, n_long, and n_short with their equivalent > u_int_* types throughout the network stack to remove the dependency on > : > > http://marc.info/?l=openbsd-tech=140523875001860=2 > > As mentioned in his

n_time in trpt(8)

2016-12-21 Thread Lawrence Teo
In 2014, mpi@ substituted n_time, n_long, and n_short with their equivalent u_int_* types throughout the network stack to remove the dependency on : http://marc.info/?l=openbsd-tech=140523875001860=2 As mentioned in his mail, trpt(8) is the only program in userland that uses n_time. The

Re: ospf6d: handle interface MTU changes

2016-12-21 Thread Sebastian Benoit
i think this is ok Jeremie Courreges-Anglas(j...@wxcvbn.org) on 2016.12.21 12:08:23 +0100: > > Hi, > > After ospfd here's a diff to make ospf6d refresh his view of an > interface's MTU at runtime. This needs a fresh kernel. > > The parent should pass the IFINFO message to its children first,

Re: NET_LOCK() pflow panic

2016-12-21 Thread Darren Tucker
On Thu, Dec 22, 2016 at 11:41 AM, Alexander Bluhm wrote: > On Thu, Dec 22, 2016 at 11:32:26AM +1100, Darren Tucker wrote: >> > I don't have a solution for the moment and I want to be sure we know all >> > recursions before trying to write a fix. So here's a diff that

Re: snmpd improvements

2016-12-21 Thread Emille Blanc
On 16-12-21 9:41 AM, Jan Klemkow wrote: On Wed, Dec 21, 2016 at 10:40:48AM +0100, Franco Fichtner wrote: The snmpd.conf can contain static values. If these values are rewritten/changed over time by rewriting the config, snmpd needs to be restarted. Is there a technical reason for not

crtbegin.o and .ctors

2016-12-21 Thread Mark Kettenis
Our crtbegin.o has: [ 8] .ctorsPROGBITS 000248 04 00 A 0 0 4 whereas crtend.o has: [ 8] .ctorsPROGBITS 48 04 00 WA 0 0 4 So in crtbegin.o the .ctors section is read-only while in crtend.o it is writable. Our

Re: NET_LOCK() pr_sysctl

2016-12-21 Thread Mateusz Guzik
On Tue, Dec 20, 2016 at 05:37:20PM +, Alexander Bluhm wrote: > Obviosly a NET_LOCK() is missing in tcp_sysctl(). > > I think it is better to place the lock into net_sysctl() where all > the protocol sysctls are called via pr_sysctl. Then we don't have > to decide each case individually. As

Re: NET_LOCK() pflow panic

2016-12-21 Thread Alexander Bluhm
On Thu, Dec 22, 2016 at 11:32:26AM +1100, Darren Tucker wrote: > > I don't have a solution for the moment and I want to be sure we know all > > recursions before trying to write a fix. So here's a diff that mark the > > recursions with a XXXSMP like in the NFS case. Looks like visa@'s crash. I

Re: Minor improvement to socket(2) man page

2016-12-21 Thread Ingo Schwarze
Hi, Michael McConville wrote on Wed, Dec 21, 2016 at 08:17:20PM -0500: > Before I do the busy-work: do the man page gurus have a preferred > phrasing? I prefer fhopen(2)'s (listed above), which seems to be the > most common. The preferred version is something close to what .Rv produces:

splassert: ip_output: want 1 have 0

2016-12-21 Thread Mark Kettenis
splassert: ip_output: want 1 have 0 Starting stack trace... ip_output() at ip_output+0x7d ipsp_process_done() at ipsp_process_done+0x2ad esp_output_cb() at esp_output_cb+0x135 taskq_thread() at taskq_thread+0x6c end trace frame: 0x0, count: 253 End of stack trace. This makes no sense to me since

Re: NET_LOCK() pflow panic

2016-12-21 Thread Darren Tucker
Hi. On Tue, Dec 20, 2016 at 03:28:38PM +0100, Martin Pieuchot wrote: [...] > I don't have a solution for the moment and I want to be sure we know all > recursions before trying to write a fix. So here's a diff that mark the > recursions with a XXXSMP like in the NFS case. I think I have another

Re: Minor improvement to socket(2) man page

2016-12-21 Thread Michael McConville
Philip Guenther wrote: > On Wed, Sep 21, 2016 at 6:23 PM, Michael McConville wrote: > > The current version is somewhat awkward and forgets to mention that > > errno is set. I adapted the paragraph found in most other system call > > man pages. > > There are six syscalls that

Re: ND6 and splsoftnet()

2016-12-21 Thread Alexander Bluhm
On Wed, Dec 21, 2016 at 01:52:45PM +0100, Martin Pieuchot wrote: > ok? OK bluhm@, but comments inline > @@ -430,15 +430,16 @@ nd6_llinfo_timer(void *arg) > void > nd6_timer_work(void *null) > { > - int s; > struct nd_defrouter *dr, *ndr; > struct nd_prefix *pr, *npr; >