piixpm(4) support for AMD FCH watchdog

The diff below adds support for the watchdog as found in the embedded AMD FCH (fusion controller hub) as found on APU2. Index: sys/dev/pci/piixpm.c === RCS file: /cvs/src/sys/dev/pci/piixpm.c,v retrieving revision 1.42 diff -u -p

em(4) reset I210 wake status

When I boot an APU2 using wake-on-lan and then attempt to power off with shutdown -hp it wakes itself up again. To prevent this it is necessary to clear the I210 PME_STATUS flag. This is described in section 5.6.2 of the I210 datasheet: The PE_WAKE_N remains asserted until the operating system

gpio(4) support for APU2

The diff below adds gpio(4) support to wbsio(4) for Nuvoton NCT5104D (pcengines APU2). It is based on Matt Dainty's diff posted to this list in November 2018: https://marc.info/?l=openbsd-tech=154134941027009=2 A key difference from Matt Dainty's original diff is the use of config_search()

Re: ld.so speedup (part 1)

On Mon 6, May 2019 at 00:16:57, Nathanael Rensen wrote: > On Sun, 5 May 2019 at 06:41, Martin Pieuchot wrote: [snip] > > Do I understand correctly that you're not recursing if not object is > > inserted? > > Yes. > > You have probably also noticed that

Re: ld.so speedup (part 2)

On Sun, 5 May 2019 at 06:26, Martin Pieuchot wrote: > > On 27/04/19(Sat) 21:55, Nathanael Rensen wrote: > > The diff below speeds up ld.so library intialisation where the dependency > > tree is broad and deep, such as samba's smbd which links over 100 libraries. > > &

Re: ld.so speedup (part 1)

On Sun, 5 May 2019 at 06:41, Martin Pieuchot wrote: > > On 27/04/19(Sat) 21:45, Nathanael Rensen wrote: > > The diff below speeds up ld.so library loading where the dependency tree > > is broad and deep, such as samba's smbd which links over 100 libraries. > > &

ld.so speedup (part 2)

The diff below speeds up ld.so library intialisation where the dependency tree is broad and deep, such as samba's smbd which links over 100 libraries. See for example https://marc.info/?l=openbsd-misc=155007285712913=2 See https://marc.info/?l=openbsd-tech=155637285221396=2 for part 1 that

ld.so speedup (part 1)

The diff below speeds up ld.so library loading where the dependency tree is broad and deep, such as samba's smbd which links over 100 libraries. See for example https://marc.info/?l=openbsd-misc=155007285712913=2 The timings below are for ldd /usr/local/sbin/smbd: Timing without diff: 2m02.50s

AMRR improvements for rt2860

I have been using an rt2860 hostap for a few years and I have discovered that AMRR does not work properly for this driver. The symptom is that some stations get stuck at 1 Mbps and do not progress up to faster rates. Unlike many drivers, rt2860 does not keep the ieee80211_amrr_node on its

Re: preserve /etc and /var file ownership during make build

On 7 March 2016 at 08:49, Theo de Raadt wrote: >> During make build the distrib/sets/makeetcset script does not preserve >> ownership when installing /etc and /var files from the base set. >> >> I've noticed that this leads to some discrepancies compared with a normal >>

maillog mode

About four months ago the mode for maillog was changed from 600 to 640 in newsyslog.conf, but the Makefile that creates this file was not updated. Nathanael Index: Makefile === RCS file: /cvs/src/etc/Makefile,v retrieving revision

sis(4) induced uvm faults on net4801

tl;dr: Zero-length packets from sis(4) on net4801 result in negative length mbufs causing uvm faults. I have observed uvm faults shortly after bringing up a sis(4) interface on a Soekris net4801: uvm_fault(0xd3adbbf0, 0xd3ee5000, 0, 1) -> e kernel: page fault trap, code=0 Stopped at

pfctl(8) detect multiple root queues on one interface

pfctl(8) doesn't check that there is at most one root queue per interface. For example: queue r0 on $if bandwidth 100M default queue r1 on $if bandwidth 100M default # pfctl -f /etc/pf.conf pfctl: DIOCXCOMMIT: Invalid argument Below is a diff for pfctl(8) to detect and report this situation

Avoid -nan for ping std-dev

Sometimes floating point imprecision in the calculation of the standard deviation of round-trip latency by ping(8) / ping6(8) leads to attempting to evaluate the sqrt(3) of a negative value: $ ping -c 1 www.google.com PING www.google.com (150.101.170.166): 56 data bytes 64 bytes from

Redundant logic in /etc/rc.d/spamlogd

The logic to disable spamlogd when spamd is not enabled was added to _rc_quirks() in /etc/rc.d/rc.subr. The corresponding logic that remains in /etc/rc.d/spamlogd became ineffective when additional code was added below it anyway. Index: spamlogd

RFC 5071 DHCP options

I use dhcpd(8) to boot some boxes with PXELINUX. The numbered options work but dhcpd.conf(5) is easier to maintain with names. These options are defined in RFC 5071. Nathanael Index: sbin/dhclient/tables.c === RCS file:

ntpd(8) constraint with a uniprocessor kernel

With ntpd(8) on a uniprocessor kernel (i386/amd64) I've found that poll(2) often reacts to the SIGCHLD from a constraint child process before the response message is received. Then constraint_check_child() closes the fd and the response is lost. # /usr/sbin/ntpd -dvs ntp engine ready constraint

Fix for smtpd offline enqueue

The smtpd enqueue -S option does not take an argument. An email enqueued offline with only the -t option (e.g. vi recovery) results in the options -S -t passed to smtpd enqueue. The -t is ignored resulting in: debug: smtpd: scanning offline queue... debug: smtpd: enqueueing offline message

Fix for shutdown(8) +minutes

Errant semicolon preventing shutdown(8) +minutes from working: Index: shutdown.c === RCS file: /cvs/src/sbin/shutdown/shutdown.c,v retrieving revision 1.42 diff -u -p -r1.42 shutdown.c --- shutdown.c 18 Apr 2015 18:28:37 -

TLS_READ_AGAIN and TLS_WRITE_AGAIN

The tls_init(3) man page states: The tls_close(), tls_read() and tls_write() functions, along with the tls_accept() and tls_connect() function families, have two special return values: TLS_READ_AGAIN A read operation is necessary to continue.

autoinstall(8) tweaks

A couple of autoinstall(8) tweaks that I find useful. I find it convenient to be able to specify a path to the response file. I also prefer to use the DHCP supplied hostname rather than the MAC address. Index: install.sub === RCS

Re: Flag to set from address in mail(1)

On Wed, 07 Jan 2015 14:31:13 -0700, Todd C. Miller wrote: Here's a version that does not rely on non-standard sendmail -t behavior. - todd Much better. Works well for me, thanks. Nathanael Index: usr.bin/mail/cmd3.c ===

Re: Flag to set from address in mail(1)

On Mon, 05 Jan 2015 22:26:03 -0500, trondd wrote: I like this better. But I still want the set from=XXX in .mailrc and of course the manpage. I would like to have this option. The diff doesn't work, however. If you reply to a message, it messes up the header and replyall will crash. set

Re: Flag to set from address in mail(1)

On Sun, 04 Jan 2015 05:57:38 +, Martin Brandenburg wrote: Since the ability to pass arbitrary arguments to sendmail has been removed from mail(1), I have added a variable and flag to pass a from address to sendmail. I considered making mail take the same arguments for this as it would

Re: Perl build race building .bs files

On Thu, 1 Jan 2015 13:47:59 -0700, Andrew Fresh wrote: I had seen complaints but was unable to reproduce the problem myself. Nathanael Rensen sent me an email out of the blue with both the likely culprit as well as a patch to fix. A few days after receiving this patch, another person

IEEE80211_DEBUG

The IEEE80211_DEBUG kernel option needs a little help to compile. Index: ieee80211_pae_input.c === RCS file: /cvs/src/sys/net80211/ieee80211_pae_input.c,v retrieving revision 1.19 diff -u -p -r1.19 ieee80211_pae_input.c ---

sysmerge auto update

Since the switch to SHA256, sysmerge(8) has stopped auto-upgrading files without local changes. $ diff -u a b | grep -E '^\+' | sed '1d' +SHA256 (./etc/rc) = 0e1cb2e6c6d11941dab14853a8ab1888f1b14fad6588fa3e6230b8ef16b62ffd $ diff -u a b | grep -E '^\+' | sed '1d' | awk '{print $3}' = $ diff -u

tcpdump(8) not printing llc data for ieee80211

Sometimes tcpdump(8) does not print llc higher layer data when using -y IEEE802_11 or IEEE802_11_RADIO. Index: print-802_11.c === RCS file: /cvs/src/usr.sbin/tcpdump/print-802_11.c,v retrieving revision 1.13 diff -u -p -r1.13

Modifying queues while pf is disabled

When pf is disabled (pfctl -d), pf_remove_queues() detaches hfsc from the interface but the queues are not removed from pf_queues_active. The next time pf rules are loaded, pf_commit_queues() calls pf_remove_queues() and in turn hfsc_delqueue() for each queue in pf_queues_active. This fails since

ntpd(8) losing sync

I noticed that ntpd(8) was regularly reporting it was unsynced, which in turn impacts down stream clients. ntpd_adjtime() in usr.sbin/ntpd/ntpd.c uses the olddelta returned by adjtime(2) to determine whether the previous adjustment was completed. The clock is considered to be synced only if

pflow(4) with optional flowsrc

Some time ago I proposed a diff to allow pflow(4) to determine the src IP address based on the route table if flowsrc was not specified. That diff was not accepted because having multiple places look up route tables is undesirable. Since then henning@ moved UDP checksum calcs into ip_output. That

Re: 802.11 power save fixes

On 10 October 2012 17:54, Mark Kettenis mark.kette...@xs4all.nl wrote: Your RT2860-based AP still works reliably with power saving support enabled? If so I'll probably commit that part of the diff as well. Yes it's working well. Since enabling power saving on my RT2860 AP I haven't noticed any

802.11 power save fixes

I enabled power save for an rt2860 AP using the following diff: Index: sys/dev/ic/rt2860.c === RCS file: /cvs/src/sys/dev/ic/rt2860.c,v retrieving revision 1.65 diff -u -p -r1.65 sys/dev/ic/rt2860.c --- sys/dev/ic/rt2860.c 23 Oct

acpi ioapic related diffs

I have included some diffs below that I have been accumulating. a) sys/dev/acpi/acpi.c sys/dev/acpi/acpicpu.c This diff is to support ACPI in a Xen HVM. I posted this previously (http://marc.info/?l=openbsd-techm=128120981015035w=2), but my timing was poor - it was late in a release cycle and

ACPI systems without legacy mode

Some ACPI systems do not support legacy mode and do not provide an smi_cmd port within the FADT. This is mentioned in the following: http://www.acpi.info/DOWNLOADS/ACPIspec40a.pdf (p119, SMI_CMD) http://wiki.osdev.org/ACPI (Switching to ACPI Mode) OpenBSD aborts ACPI initialisation if the FADT

RT2860 GTK update

I am using a Ralink RT2860 with OpenBSD 4.7 in a Soekris net4801 in hostap mode as a wireless access point. After a group temporal key (GTK) update broadcasts stop being received by the attached stations. If the GTK update reuses the same RT2860 shared key slot then the problem goes away: Index: