BSD Commit Log Search

   stack(7): tweak diagrams for readability a bit

   lint, make: only enable strict bool mode when using GCC

   The GCC precompiler marks each token whether it comes from a system
   header. Lint uses this information to allow some expressions that aren't
   strictly bool, such as in __predict_true and __predict_false. The Clang
   preprocessor treats all tokens as not coming from system headers, thus
   being too strict.

   Use the new FCNTL mechanism to move blocks to clean segments, rather than
   duplicating the code in user space.  lfs_cleanerd still decides which
   segments' blocks to move but no longer identifies individual blocks
   for cleaning, and does not handle file data.

   Wait to determine inode-block size until the block and fragment sizes have
   been finalized.

   Add ibsize to the summary output.

   stack(7): fix stack base computation

   ok uwe@

   Note release of tz{code,data}2026b

   sysctl(7): document guard size macros

   Fix off-by-one in case user provided '\x80' in the auth string

   g/c

   pthread_attr(3): link to pthread_attr_get_np(3)

   evbarm: drop unused altera configuration files

   lint: recognize Clang's __c11_atomic builtins

   Needed by libpthread and several others when building with HAVE_LLVM=yes
   MKLLVM=yes MKGCC=no.

   doc/RESPONSIBLE: add myself

   Re-factor the various m68k platforms' identifycpu() functions, and bring
   some consistency to how the CPU/FPU/MMU/cache information is displayed.
   Each platform now provides a machine_set_model() function that is called
   from cpu_startup_common().  This function is respondible for calling
   cpu_setmodel() with platform-specific model description.  Then, common code
   is responsible for printing the model information, along with the CPU
   and cache info, but separately (and this is no longer included in the
   hw.model sysctl node; at least it's now consistent!)

   Some examples:

     Qemu 10.1.2 Virt platform
     MC68040+MMU+FPU, 4k+4k on-chip I/D caches

   (hw.model == "Qemu 10.1.2 Virt platform")

     Motorola MVME-147
     MC68030 CPU+MMU @ 25MHz, MC68882 FPU


    [43 lines not shown]

   fix clang build.

   Fix double free of uri (later handled by bozo_clean_request())

   Fix use-after-free in the "http://" case

   pmapmi: Define and use pmap_md_kernel_vaddr_{p,to_paddr} in pmap_extract.

   Allow an architecture to provide a function that extracts the PA from the
   VA that the kernel text/data/bss are mapped at. This allows pmap_extract
   to work for kernel mappings that aren't PAGE_SIZE sized.

   Fix iteration over protos[] to prevent out-of-bounds access

   switch everyone to jemalloc 5.3.1 again

   micphy: Fill out mii_mpd_model in the sc struct.

   The rest of the code expects this to be filled in and the driver
   seems stable with this change (at least on my machine).

   "ship it!" jmcneill@

   micphy: Apply errata workaround for KSZ9031

   This fixes ethernet on an olimex a64 board, and should be pulled up
   to -11, -10, -9...

   Thanks jmcneill for pointing out the problem.

   PR port-arm/60193: olimex a64 - no carrier with emac0 ethernet

   3RDPARTY: note newer Lua releases

   Add cast to fix the build on some platforms

   sunxidrm: suppress misleading error message

   Do not print "failed to activate endpoint" if the device tree leaves a
   CRTC unused.

   sunxi: Inject PMF display on/off events in CRTC-0 atomic_{enable,disable}

   Fixes "dpms" on Pinebook after update to drm atomic interface.

   catch up with gftfb and hyperfb:
   - support bold mono characters
   - use glyphcache_underline() where appropriate
   - set & support RI_FULLCLEAR

   use glyphcache_underline(), add & support RI_FULLCLEAR

   some minor fixes:
   - underling anti-aliased characters properly
   - set & support RI_FULLCLEAR