BSD Commit Log Search

   test the right thing, doofus

   avoid possible NULL deref if ssh_packet_check_rekey_blocklimit()
   called before the encrypted transport is brought up.

   make rdpkru() return unsigned, PKRU is all access/write disable bits
   no change in behaviour for existing callers
   ok deraadt@ mlarkin@

   Remove bug compatibility for implementations that don't support
   rekeying. AFAIK this is only an ancient Sun SSH version.

   If such an implementation tries to interoperate with OpenSSH, it
   will eventually fail when the transport needs rekeying.

   This is probably long enough to use it to download a modern SSH
   implementation that lacks this problem :)

   ok markus@ deraadt@

   Enforce maximum packet/block limit during pre-authentication phase

   OpenSSH doesn't support rekeying before authentication completes to
   minimise pre-auth attack surface.

   Given LoginGraceTime, MaxAuthTries and strict KEX, it would be
   difficult to send enough data or packets before authentication
   completes to reach a point where rekeying is required, but we'd
   prefer it to be completely impossible.

   So this applies the default volume/packet rekeying limits to the
   pre-auth phase. If these limits are exceeded the connection will
   simply be closed.

   ok dtucker markus

   unit tests for sshbuf_consume_upto_child()

   Add sshbuf_consume_upto_child(), to similify particular parsing
   patterns using parent/child buffer; ok markus@

   Serialize concurrent access to command queues with a mutex.  Fixes
   corruption visible on the Orion O6 with multiple devices and after
   the secondary cores spun up.  Command queue handling/synchronization
   can probably the improved further.

   We reserve regions in the IOVA map to ensure that we don't accidentally
   hand out addresses to the device that terminate in other places, e.g PCI
   BARs.  The IOVA address space, and thus its extent, can be smaller than
   real physical address space, in which those PCI BARs are.  As the extent
   susbsystem greets us with a panic if we attempt to allocate outside its
   coverage, ensure that we skip or clamp those.  This was visible on mpi@'s
   Ampere Altra with SMMUv3 turned on.

   Purge the ifq if the link is down.  This aligns the behaviour with other
   drivers, like em(4).  This fixes an issue where dwqe(4), e.g. on a veb(4),
   doesn't recover when the link is done but packets are bridged, filling up
   the ifq.

   ok dlg@

   Ampere Altra ACPI IORT table does not define SPIs for the different IRQ
   types.  Instead I assume we'll have to make use of MSIs, which I'll yet
   have to implement.

   It's not much of a problem though because these interrupts are mostly used
   in case something goes wrong, e.g. due to an invalid access; if everything
   is working fine, we don't need interrupts right now.

   Unfortunately though we bailed out at an unfortunate moment:  After we had
   already initialized the SMMU to intercept, but before the driver is hooked
   up as IOMMU, thus not mapping anything but the SMMU enforcing boundaries.

   For now we can simply log that the interrupts are not there and proceed.
   Note that SMMU on ACPI isn't enabled by default yet anyway.

   If SMMUv3's COHACC override flag is set in the IORT, we can force the SMMU's
   DMA tag to be coherent.  This tag is used for its command and event queues.

   Reclaim ownership of the page when retrying flushing it synchronously.

   Note that IOs in uvn_put() are always synchronous and the !PGO_SYNCIO case
   doesn't really make sense.

   Prevent a panic found by sthen@ and landry@

   rsync.1: sort flags; from Matthew Martin (phy1729 at gmail), thanks!

   Fix super-speed enablement by initializing the controller properly and
   then enabling USB3 support on the controller (instead of disabling it
   like we do right now).  This makes my umass(4) come up as super-speed
   instead of high-speed.

   ok kettenis@

   Display the new added rde_event_adjout_usec metric
   OK tb@

   Track the usec spent in processing the adj-rib-out as an own metric.

   OK tb@

   Bump version to 9.0

   Introduce a rib entry queue used to split the processing of UPDATES into
   two steps.

   One of the biggest latency inducing bottlenecks in bgpd is that UPDATE
   messages are processed to completion.  Updating the Adj-Rib-Out is a
   loop over all peers and is therefor super heavy.  By taking this part
   of the pipeline into its own step we can drop the latency inside the
   main poll loop by a lot.

   Introduce a per-peer update queue that enqueues the rib entries after
   the decision process and then handles those one by one. This also has
   a benefit that a single peer can not monopolize the processing in bgpd.
   So a single flapping peer should result in far less noticeable delays
   on all other UPDATES from other peers.

   For now disable the "addpath send all" optimisation since for that
   an extra queue of updates needs to be kept. Instead just use the regular
   addpath codepath that is less optimised but produces the same results.

   With and OK tb@

   Crude tr /t support.

   Some USB devices have a iManufacturer property that is just a single space.
   Handle this (and substitute the vendor name from usbdevs) by checking the
   length of the string after trimming spaces.

   ok jsg@

   regen

   Add D-Max as a USB vendor.  Seems the company doesn't exist anymore, but
   it is still listed in the official USB vendors list and it is unclear what
   company is currently using the vendor ID in question.

   ok jsg@

   print the IA_NA option out.

   print Boot File URL options

   sync

   handle printing of the dns servers option.

   Ignore whitespace-only or whitespace-then-comment lines from -f sysctl.conf

   " " or "  # comment" would be parsed as variable name rather than skipped.
   Found and fixed by "Alf" on bugs@, thanks.

   New regress/sbin/sysctl spc and spc-com targets now pass with this.
   "looks good" deraadt

   run sysctl

   exercise sysctl -f