dont need splvm in pool_gc_pages

2017-02-06 Thread David Gwynne
jmatthew@ noticed that i still had this code in here, despite making it mandatory for pools to specify the IPL they operate at. ok? Index: subr_pool.c === RCS file: /cvs/src/sys/kern/subr_pool.c,v retrieving revision 1.205 diff -u -p

use mac address from fdt on raspberry pi

2017-02-06 Thread Jonathan Gray
The videocore part of the raspberry pi that boots the system and runs the mailbox interface and other functions afterwards knows about a MAC address that appears to be derived from a portion of a unique serial number along with the raspberry pi foundation oui B827EB (base 16)Raspbe

Re: specify curves via ecdhe statement in httpd.conf

2017-02-06 Thread Andreas Bartelt
On 02/06/17 14:44, Joel Sing wrote: On Sunday 05 February 2017 17:05:40 Andreas Bartelt wrote: - What type of public certificate are you using (RSA or ECDSA)? ECDSA with P-256. Certificate signed by letsencrypt (via RSA). Must-staple is enabled - that's why I'm also using the ocsp line for tes

Re: tcpstat percpu counters

2017-02-06 Thread Jeremie Courreges-Anglas
Jeremie Courreges-Anglas writes: [...] I've removed the TCPSTAT_INLINE code. If we want to experiment with inlined/not inlined counters, it is probably more interesting to make that choice elsewhere. > Anyway, tcpstat diff below, a bit ugly because that struct is a mix of > uint32_t and uint64

Re: Xorg vs wifi scanning

2017-02-06 Thread Stefan Sperling
On Mon, Feb 06, 2017 at 03:54:56PM +0100, Martin Pieuchot wrote: > tb@ found that another code path needs to be unlocked. Diff below does > that. These ioctl(2)s are only messing at the driver level, so it is > safe to drop the NET_LOCK() there as well. > > ok? I am happy with this. OK. And th

void crp_callback

2017-02-06 Thread Alexander Bluhm
Hi, The return code of crp_callback is never checked, so it is not useful to propagate the error. When an error occures in an asynchronous network path, incrementing a counter is the right thing. By removing the error handling, I have found 4 places where an error is not accounted. For now just

Re: PF & CPU hogging

2017-02-06 Thread Mike Belopuhov
On 6 February 2017 at 17:02, Martin Pieuchot wrote: > PF has its own home-brewed solution for dealing with CPU hogging. It > has been introduced in r1.88 of net/pf_table.c and I couldn't find any > explanation why it is different than the idiom we use in other places. > The idea was just to be a

Re: PF & CPU hogging

2017-02-06 Thread Mike Belopuhov
On 6 February 2017 at 17:02, Martin Pieuchot wrote: > PF has its own home-brewed solution for dealing with CPU hogging. It > has been introduced in r1.88 of net/pf_table.c and I couldn't find any > explanation why it is different than the idiom we use in other places. > > So let's use the same id

NFS: Kill so_upcall NET_LOCK() recursion

2017-02-06 Thread Martin Pieuchot
guenther@ pointed out during a2k17 that nfsrv_rcv() already has a way to not re-enter the socket layer. Instead of doing soreceive() directly in the receiving path, we notify and wakeup nfsd(8) to do it. I'd like to do that unconditionally to fix one of the blocking recursions. I just finished a

PF & CPU hogging

2017-02-06 Thread Martin Pieuchot
PF has its own home-brewed solution for dealing with CPU hogging. It has been introduced in r1.88 of net/pf_table.c and I couldn't find any explanation why it is different than the idiom we use in other places. So let's use the same idiom, I promise to introduce a macro an unify all of them once

Re: Scheduler ping-pong with preempt()

2017-02-06 Thread Bob Beck
Go for it mpi.. move forward. ok beck@ On Mon, Feb 6, 2017 at 7:48 AM, Martin Pieuchot wrote: > On 24/01/17(Tue) 13:35, Martin Pieuchot wrote: > > Userland threads are preempt()'d when hogging a CPU or when processing > > an AST. Currently when such a thread is preempted the scheduler looks >

Re: Xorg vs wifi scanning

2017-02-06 Thread Martin Pieuchot
On 06/02/17(Mon) 12:27, Martin Pieuchot wrote: > Paul and Antoine reported that, since the NET_LOCK() went in, doing a > channel scan with iwm(4) and iwn(4) freezes X. > > This is a deadlock due to the fact that wireless drivers sleep during > a long time holding the NET_LOCK() while the X server

Re: Scheduler ping-pong with preempt()

2017-02-06 Thread Martin Pieuchot
On 24/01/17(Tue) 13:35, Martin Pieuchot wrote: > Userland threads are preempt()'d when hogging a CPU or when processing > an AST. Currently when such a thread is preempted the scheduler looks > for an idle CPU and puts it on its run queue. That means the number of > involuntary context switch oft

Re: specify curves via ecdhe statement in httpd.conf

2017-02-06 Thread Joel Sing
On Sunday 05 February 2017 17:05:40 Andreas Bartelt wrote: > > - What type of public certificate are you using (RSA or ECDSA)? > > ECDSA with P-256. Certificate signed by letsencrypt (via RSA). > Must-staple is enabled - that's why I'm also using the ocsp line for > testing. Ah, this was the miss

ssh-keygen: vfprintf %s NULL

2017-02-06 Thread Sebastien Marie
Hi, While changing comments on ssh-key, I saw some vfprintf in log: Feb 6 14:24:44 clyde ssh-keygen: vfprintf %s NULL in "Key now has comment '%s' " Steps to reproduce: $ ssh-keygen -f test -C '' Generating public/pr

rip6 and icmp6stat percpu counters

2017-02-06 Thread Jeremie Courreges-Anglas
In icmp6_errcount() we could save a few function calls but I preferred to keep the conversion as mechanical as possible. Works fine here on amd64 (I can test armv7 soon). ok? Index: netinet/icmp6.h === RCS file: /d/cvs/src/sys/net

Re: make sosetopt responsible for m_free

2017-02-06 Thread Martin Pieuchot
On 03/02/17(Fri) 11:02, David Hill wrote: > On Fri, Feb 03, 2017 at 09:50:40AM +0100, Martin Pieuchot wrote: > > On 02/02/17(Thu) 12:12, David Hill wrote: > > > On Thu, Feb 02, 2017 at 09:34:07AM +0100, Martin Pieuchot wrote: > > > > On 01/02/17(Wed) 19:27, David Hill wrote: > > > > > Hello - > > >

Re: Xorg vs wifi scanning

2017-02-06 Thread Martin Pieuchot
On 06/02/17(Mon) 12:59, Stefan Sperling wrote: > On Mon, Feb 06, 2017 at 12:44:52PM +0100, Martin Pieuchot wrote: > > On 06/02/17(Mon) 12:39, Theo Buehler wrote: > > > This fixes the issue on my iwn0, but re-introduces the problem on my iwm0. > > > > That means another ioctl(2) triggers the proble

Re: Xorg vs wifi scanning

2017-02-06 Thread Stefan Sperling
On Mon, Feb 06, 2017 at 12:44:52PM +0100, Martin Pieuchot wrote: > On 06/02/17(Mon) 12:39, Theo Buehler wrote: > > This fixes the issue on my iwn0, but re-introduces the problem on my iwm0. > > That means another ioctl(2) triggers the problem with iwm(4). It would > help if you could figure out w

Re: Xorg vs wifi scanning

2017-02-06 Thread Martin Pieuchot
On 06/02/17(Mon) 12:39, Theo Buehler wrote: > On Mon, Feb 06, 2017 at 12:27:15PM +0100, Martin Pieuchot wrote: > > Paul and Antoine reported that, since the NET_LOCK() went in, doing a > > channel scan with iwm(4) and iwn(4) freezes X. > > > > This is a deadlock due to the fact that wireless drive

Re: Xorg vs wifi scanning

2017-02-06 Thread Theo Buehler
On Mon, Feb 06, 2017 at 12:27:15PM +0100, Martin Pieuchot wrote: > Paul and Antoine reported that, since the NET_LOCK() went in, doing a > channel scan with iwm(4) and iwn(4) freezes X. > > This is a deadlock due to the fact that wireless drivers sleep during > a long time holding the NET_LOCK() w

Xorg vs wifi scanning

2017-02-06 Thread Martin Pieuchot
Paul and Antoine reported that, since the NET_LOCK() went in, doing a channel scan with iwm(4) and iwn(4) freezes X. This is a deadlock due to the fact that wireless drivers sleep during a long time holding the NET_LOCK() while the X server tries to communicate with unix sockets, that still need t

Re: Password corruption in adduser

2017-02-06 Thread Theo Buehler
On Mon, Feb 06, 2017 at 11:04:02AM +, Raf Czlonka wrote: > Hi all, > > How about doing it throughout the tree[0]? > > [0] http://marc.info/?m=142689311221135 Thanks. Most of those are already fixed (or have been removed). I found and fixed the one in smtpd/table.5 right after fixing this one

Re: Password corruption in adduser

2017-02-06 Thread Raf Czlonka
Hi all, How about doing it throughout the tree[0]? [0] http://marc.info/?m=142689311221135 Cheers, Raf On Mon, Feb 06, 2017 at 05:53:22AM GMT, Theo Buehler wrote: > On Sun, Feb 05, 2017 at 09:47:35PM -0800, Philip Guenther wrote: > > On Sun, 5 Feb 2017, John McGuigan wrote: > > > I've noticed

Re: mount(8): some code shuffling to avoid a pledge problem

2017-02-06 Thread Sebastian Benoit
Theo Buehler(t...@math.ethz.ch) on 2017.02.06 08:17:43 +0100: > ping > > On Sat, Jan 28, 2017 at 03:25:53PM +0100, Theo Buehler wrote: > > The problem: > > > > $ mount /dev/tty /tmp > > Abort trap (core dumped) > > > > The relevant kdump snippet: > > > > 45441 mountCALL open(0x7f7eb58