BSD Commit Log Search

   Use pmap_{pa_has,va_is}_static_mapping() rather than hard-coding the
   VIRT68K TT ranges.

   Introduce new functions:
   - pmap_pa_has_static_mapping() -- Does the specified physical address
     range have a static mapping (either TT or allocated in pmap_bootstrap1())?
   - pmap_va_is_static_mapping() -- Is the specified virtual address range
     one of those static mappings?

   These platforms don't yet have the necessary changes to use the new
   68k pmap, but declare an empty machine_bootmap[] for a future change
   to shared code that will reference it.

   Remove __HAVE_NEW_PMAP_68K for declaring machine_bootmap[].

   Fix previous; the old cpuspeed was in MHz by the end of cpu_startup(),
   not MHz*10 (despite being MHz*10 earlier in startup).

   Also, don't const-fold by hand; let the compiler do it so that it's
   more obvious what is going on.

   Oops, <m68k/mmu.h> isn't needed by user-land.  Include it when _KERNEL
   only.

   Fully duplicate the pmap_bootmap definitions from the new 68k pmap
   module to simplify the transitional phase and faciliate some future
   code sharing.

   Adapt to cpuspeed -> cpuspeed_khz changes.

   Add mmu_range_is_tt() helper function that consults the TT register
   configuration to determine if a physical address range is transparently-
   translated for the specified access.

   (missed "cvs add")

   Add mmu_range_is_tt() helper function that consults the TT register
   configuration to determine if a physical address range is transparently-
   translated for the specified access.

   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