Rev 1.153 moved the initialization of buckstring[] from lazily at first
sysctl kern.malloc.bucket retrieval, to kmeminit(), but forgot to remove the
buckstring_init multiple initialization guard. Remove it.
Extremely old PA-1.0 processors did not have integer multiply and divide
instructions, and relied upon SFU (coprocessor) instructions to do so.
The floating-point emulation code written by HP for OSF/1 had support to
emulate these instruction as well.
That code eventually made its way to Utah' HPBSD, and later to
PA-RISC MkLinux; but both code bases had the appropriate logic to check for
these SFU instructions and dispatch to the emulation code commented out,
because none of these code bases were supporting PA-1.0 (PCX or older)
processors anymore.
When this code was carried away in OpenBSD, the unneeded dispatch logic was
left out, but it was not noticed that this had caused some code in spmath/
to become dead code. Let it rot in the Attic/ now.
Fix loading networks via mrt and bulk endpoints.
When refactoring the filterset code in the RDE these endpoints were
not adjusted to use imsg_send_filterset() and so the transmitted
network was dropped in the RDE.
Reported and initial fix provided by René Ammerlaan
OK tb@
Handle protected frames in the rx and tx path for MT7925
In mt7925_mac_fill_rx() set IEEE80211_RXI_HWDEC if the frame was decrypted
in HW. Also apply the same workaround for the stripped IV and MIC as done
for the MT7921 side.
In mt7925_mac_write_txwi() set the protected frame bit in the txd if the
packet is marked as protected. Also sync the function with the linux code
and add all the NOTYET bits.
With this my mt7925 seems to be happy on both encrypted and unencrypted
networks.
Add refresh-client -q to query terminal size, useful for terminals which
do not respond to SIGWINCH (such as serial). GitHub issue 5766 from
Charlie Stanley.
rpki-client: tweak sha256 parameter handling
This makes cbs_is_sha256_algorithmIdentifier_value() use the same logic
as is used for for sha256WithRSAEncryption a few lines below. For the
latter this is now simply an RFC requirement as ARIN fixed their objects
with absent parameters in sha256WithRSAEncryption. However, they still
have tons of sha256 with incorrectly encoded parameters (as do many
others).
ok claudio job
Refactor and rename variables & functions for clarity
This refactor should make it clearer that CAs are prioritised only if
they were confirmed in the previous run, and improves logic that hinged
upon a timestamp being non-zero.
OK tb@
Correctly check whether CAs were previously confirmed
The intention was to check using the state from the previous run, but
this got lost in a refactor.
Reported by Ties de Kock
OK tb@
Prevent fq_codel panic from drop counter wrap
Sustained traffic through a congested fq_codel queue can wrap the
per-flow drop counter to zero, triggering the control_law() assertion
and panicking the kernel. Saturate drops and decrement a nonzero ldrops
to preserve the grace-period delta and prevent the wrap.
Issue reported by Acts1631.
OK sashan@, claudio@
pf(4) yet another race between pf_test() and pf_purge_states()
pf_counters_inc() running on behalf of pf_test() may dereference a
dead memory left behind pf_purge_states(). The reference to state
object is still valid, but other items such as radix tables might
be gone. To stay on the safe side the pf_counters_inc() must
grab state's mutex and check if state itself is on its death
row (->timeout == PFTM_UNLINKED). If it is the case then
pf_counters_inc() must bail out without trying to update counters.
Issue was kindly reported by Acts1631.
OK dlg@
Fix out-of-bounds read on on missing right bracket after character class.
The original code checked for a closing bracket before parsing the
bracket expression. When character class support was added, that
closing bracket can now be part of the character class instead.
Add another check for the closing bracket after parsing character
classes and roll back if we don't find one.
Adapted from an OpenSSH bug report. OK djm@
