BSD Commit Log Search

   Rewrite description of BIOCLOCK, following discussion with guy harris
   ok florian

   Honor WAITOK semantic in uvm_pagermapin() by sleeping if pmap_enter(9) failed.

   Fix a bug reported by bluhm@ where a WAITOK allocation in the page fault
   handler would return a NULL kva instead of sleeping.

   ok jca@

   uvm_km_free(9) is gone.

   Use km_alloc(9) to allocate malloc(9) memory chunks.

   Get rid of uvm_km_kmemalloc_pla() and uvm_km_free() that are no longer used.

   This finishes the conversion to the "New unified allocator of kernel memory"
   introduced by art@ 14 years ago.

   ok tb@

   Do not allow root to bypass BIOCLOCK
   ok dlg florian, plus discussion with guy harris

   Use splbio() instead of splvm() to block interrupts in the page cache.

   Only the buffer flipper (incorrectly?) allocates & frees pages from
   interrupt context.

   Sync the comments with some pmap_copy_page(9) implementations to keep
   track of hacks required by the flipper.

   ok jca@

   Remove calls to OpenSSL_add_all_algorithms() and ERR_load_crypto_strings().
   These are no-ops in LibreSSL, and in Portable have been mostly replaced
   by a call to OPENSSL_init_crypto() in the compat layer.  ok tb@

   the structure produced has alignment, which contained uninitialized data.
   This is obviously fixed by using calloc(), but it was also observed that
   the precalculated storage requirement was higher than what is actually
   filled.  So all the math has been rewritten to be byte accurate.
   Temporarily, a syslog_r check is added to report if we ever get count
   inconsistancy again so that can be fixed quickly.
   ok claudio

   Simplify up_generate_addpath_all() by using up_generate_addpath() for
   the case where new and old are NULL.

   This makes it more clear what up_generate_addpath_all() actually does
   and up_generate_addpath() still handles the addpath send all case.
   OK tb@

   sync support for systems that lack __builtin_popcount() from portable

   unused on OpenBSD (nothing sets MISSING_BUILTIN_POPCOUNT), but it
   makes syncing much easier.

   update our ML-KEM implementation to upstream libcrux v0.0.4

   tested/ok tb@

   Check return value of trie_roa_add() in a consistant way.

   Also adjust the trie_add() call for IMSG_RECONF_PREFIX_SET_ITEM to follow
   the same logic.

   Fixes CID 492365
   OK benno@

   Make requests to the external terminal one queue instead of one by type,
   and include any other requests as well so that ordering is maintained.

   Rename mld6_fasttimeo() to mld6_fasttimo() for consistency with all
   other protocol timeout functions.

   Include the RDE timing statistics in the show rib mem output but
   also in the show metric output.
   OK tb@

   Track the amount of time spent in various parts of the RDE main event loop.

   Knowing which parts of the loop take a lot of time helps with performance
   tuning.
   OK tb@

   If display-popup is used inside a popup, modify that popup. From
   m-einfalt at gmx dot de in GitHub issue 4678.

   sync

   Correct locking documentation about fields updated only by the page daemon.

   please commit generated files afterwards

   Convert multicast address lookup macro into function.

   Convert macros IN_LOOKUP_MULTI() and IN6_LOOKUP_MULTI() into functions
   in_lookupmulti() and in6_lookupmulti().  Make some paramters const
   as suggested by kn@.

   OK kn@ mvs@

   vmboot: A tiny kernel for booting SEV VMs

   When using SEV-enabled VMs, we let vmd(8) launch the kernel directly.
   With this, we do not have to trust a firmware like OVMF.  However,
   with this approach, sysupgrade(8) can not be used as the kernel
   resides outside the VM in the host file system.  Thus maintainig
   SEV-enabled VMs is hard.

   With vmboot, we use a small ramdisk kernel that runs the actual
   kernel using the kexec facility.  vmboot can be launched directly
   by vmd(8).  The actual kernel is loaded from the disk image of the
   SEV-enabled VM.  With this, sysupgrade(8) can be used, and there
   is no need for additional firmware.

   This approach is similar to vmboot for powerpc64 and octeon.

   Discussed with bluhm@ for a while.

   ok mlarkin@

   Systopia team at UBC found 5 info leaks in the private privsep protocol
   between httpd programs, and provided a fix.
   ok claudio benno florian

   naddy points out that environment variable BLOCKSIZE might interfere
   with the df output, and recommends using -P to force a specific output
   format.

   Allow to compile a kernel without the NETHER attribute

   input kettenis@ deraadt@ cludwig@
   ok bluhm@

   retire wbinvd_on_all_cpus()

   ok mlarkin@

   Do not send theme updates to panes that have exited, GitHub issue 4671.

   remove another mention of the nixspam list, which has shut down.
   ok ratchov@

   if anyone still has this list configured in spamd.conf they should
   remove it (and maybe disable spamd-setup).

   it would be nice if the file could be removed or emptied on
   www.openbsd.org too

   Convert all keys for backspace, not just A-Z.

   Add a command to explcitly set the selection mode in copy mode, GitHub
   issue 3842.