Diable a few options unlikely to be all that useful on a NeXT:
- COMPAT_NOMID, COMPAT_SUNOS
- file-system MFS (TMPFS is preferred)
- dse, se (SCSI Ethernet)
- sl (SLIP), ppp / pppoe
Shrinks the kernel by ~167KB.
Properly account for the display frame buffer in machine_bootmap[].
New pmap now at least get close to single-user on my slab. Issue
with Ethernet driver?
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.
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.
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.
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.
