BSD Commit Log Search

   openssh-10.2

   The only change since 10.1 is the channels.c fix

   use multiple txqs to spread traffic handling over softnet threads.

   implement "checksum offload" between rport pairs.

   this allows the kernel to skip ip/tcp/udp checksum calculation for
   packets between rdomains.

   rport interfaces advertise checksum offload capabilities to the
   stack, so the stack will set CSUM_OUT mbuf flags rather than do the
   checksum calculation for packets being transmitted on an port
   interface. when these packets are sent back into the stack by the
   partner rport interface, the CSUM_OUT flags get mapped to equivalent
   CSUM_IN_OK flags. this is modelled on how lo(4) does the same thing.

   ice(4): print no. of queues during attach

   ok bluhm@, stsp@

   sys/time.h needed before sys/futex.h

   When we move to more partitions, MAKEDEV will need sysctl kern.maxpartitions.
   The ramdisk sysctl is a minimal version which only knows certain variables,
   so add this one into the list.

   Revert previous. Let's deal with it when the portable release is out.

   test framework: allow overriding the "/tmp/" directory

   Windows is super picky when it comes to paths, so it needs some special
   massaging. Will let us avoid a patch or hack in portable.

   Drop variable names from the swapctl(2) prototype, per style(9).

   ok tb@

   propagate errors from bnxt_up() back to userspace

   Reveals errors that can occur during 'ifconfig bnxt0 down up', such as:
   ifconfig: bnxt0: SIOCSIFFLAGS: Cannot allocate memory

   help & ok jmatthew@ sf@ jan@

   don't reuse c->isatty for signalling that the remote channel
   has a tty attached as this causes side effects, e.g. in
   channel_handle_rfd(). bz3872

   ok markus@

   unbreak vport to vport communication on the same veb.

   veb is now set up to support port interfaces that can do checksum
   offloading on behalf of the network stack (or virtual machines via
   tun_hdr offloads). to allow packets leaving the kernel ip stack via
   a vport interface to use hardware offloading, vport interfaces
   advertise ip/tcp/udp offloads. veb then fixes up the packet if
   the outgoing port doesn't support the necessary offload.

   however, veb and vport can be used to support communication between
   rdomains. this means packets can leave the stack via a vport interface
   in one rdomain, and then enter the stack again on another vport.
   in this situation the packets (re)entering the stack don't have
   checksums calculated because vport advertises csum offloads, so the
   packets get dropped.

   this has vport map the CSUM_OUT flags to CSUM_IN flags before pushing
   the packets into the stack again, despite the checksums not being
   calculated. this is modelled on the behaviour of lo(4) and tun(4)/

    [4 lines not shown]

   oops, accidental commit, backout

   openssh-10.1

   update documentation for btrace(8) and dt(4) so it matches
   the code we have in tree. Lot of suggestions from claudio@
   and ingo@.

   OK deraadt@

   Make -mserialize-volatile the default. Correctly, this time.

   Add newly required #include <sys/disklabel.h>

   ok deraadt@

   few more places to use DL_PARTNUM2NAME / DL_PARTNAME2NUM

   oops; typo

   Make async IOs daemon aware of pmemrange's OOM mechanism.

   Fix a missing wakeup.

   reported and tested by miod@, ok kettenis@

   Lower bound for how long to prefer IPv6 at least.

   RFC 8925 suggests 5 minutes, so that's what we are going with.

   kn noticed that if the IPv6-Only preferred option is send from the
   server with a value of 0, dhcpleased(8) would still request a lease
   while apple devices would not.

   There were two problems with that:
   1. A value of 0 would be treated as turning the feature of, which is
   not correct.
   2. The value would be used for a timeout, so a very small value would
   create a timeout that would constantly fire.

   OK kn

   typos: a ssh* -> an ssh*

   ok dtucker@

   rpki-client: improve error message for duplicate TALs

   Using several TALs with the same SPKI doesn't really make sense and no
   longer works due to the way the CCR is constructed. Since the current
   error is very confusing, hint more clearly at the problem.

   Prompted by a report by Gregory Edigarov on misc.

   ok job

   disable POOL_DEBUG for release
   ok deraadt@

   Make -mserialize-volatile the default. Missed during the gcc3->gcc4 backend
   conversion.

   In the expansion of zero_extendqisi2, remove code which is not found in the
   other zero_extend* expansions and is identical to the fallback MI code.

   In pseudo-assembly statements, %L implies a trailing %., so constructs using
   %L%. are wrong.
   The fact that this never caused any assembler complaints hints that the two
   occurrences of this wrong construct are never hit - and in fact, commenting
   them out entirely does not appear to change anything in generated code.

   Let cmmu_apr_cmode() also provide the value of CACHE_GLOBAL if needed, rather
   than have pmap use an incorrect heuristic to decide whether it can be cleared.

   If write() returned short, the subsequent write would restart from the
   beginning of the buffer not the end of what was written.  Fix, since
   we want modpipe to corrupt data for testing purposes deliberately not
   accidentally.  ok djm@

   stray newline