PERFORCE change 32207 for review

[Peter Wemm]

http://perforce.freebsd.org/chv.cgi?CH=32207

Change 32207 by peter at peter_hammer on 2003/05/31 12:55:51

	update

Affected files ...

.. //depot/projects/hammer/TODO.txt#2 edit

Differences ...

==== //depot/projects/hammer/TODO.txt#2 (text+ko) ====

@@ -1,4 +1,8 @@
-$P4: //depot/projects/hammer/TODO.txt#1 $
+$P4: //depot/projects/hammer/TODO.txt#2 $
+
+reread the logs etc and check out the "XXX worry about this later" stuff
+
+performance tuning (segment registers etc at context switch time, pmap)
 
 threads (get/setusercontext/swapcontext/makecontext etc and kse kernel bits)
 (peter: started work on libc_r, but it crashes still)
@@ -6,26 +10,21 @@
 USER_LDT (however, there is a syscall to set the %fs and %gs base values
   to arbitary 64 bit addresses).
 
-ACPI
-(compiles/runs.. test more)
+SMP (acpi MADT table based etc, ignore mptable stuff)
 
-SMP
+kld module support (the elf_machdep.c relocator is wrong, it uses the old value
+rather than addends like it should.  See alpha/sparc64 for clues.)
 
-kld module support (the elf_machdep.c relocator is wrong).
-
 finish the missing relocation types in rtld-elf/amd64/reloc.c (easy)
 
-boot blocks (need gcc -m32 to work)
-(fixed in p4 tree, needs binutils patches etc)
+make release (should be mostly there, with the bootblock patches)
 
-performance tuning (segment registers etc at context switch time, pmap)
+X server
 
-'make release'
+add userland access to hardware debug registers via ptrace/procfs.
 
-X server
+gdb (doesn't look *that* hard, once debug regs are done)
 
-gdb
-
 ddb (needs a dwarf2 based unwinder to get stack traces and a disassembler)
 (peter: basic functionality, uses -fno-omit-frame-pointer.. disassembler
  still expects 32 bit code but compiles)
@@ -35,21 +34,27 @@
 
 crashdumps?
 
-teach libkvm about crashdumps.
+teach libkvm about crashdumps. (need to lookup KPML4, 4 level page tree etc)
 
-see if openssl can use the assembler x86-64 routines.
+see if openssl can use the assembler x86-64 routines. (easy)
 
 128TB of user process VM space (the support is there, but there is
 a bug somewhere as it crosses the 512G mark into a new pml4e)
 
 full 2MB pmap page support in pmap (ifdef'ed out, some stuff uses vtopte()
-where it shouldn't)
+where it shouldn't, and the direct pte map interface needs to be made
+aware of it)
 
 SSE/SSE2 support in libc (the fpget/set*() functions need to adjust the
-    SSE MXCSR register as well as the x87 control word).
+SSE MXCSR register as well as the x87 control word - netbsd have already
+done this, we can look at theirs)
 
 increase direct map base size to 4GB to cover all of pci mapping space
-so that pmap_mapdev can use it.
+so that pmap_mapdev can use it without consuming kvm.
+
+factor out common compat/freebsd32/* and compat/ia32/* from ia64 and amd64
+trees.  freebsd32 should be for running 32 bit binaries on a 64 bit kernel,
+while ia32 is for special MD bits.  Lots of missing syscalls.
 
 make ia32 binary support more robust, including things that do important
 ioctls etc.
@@ -59,3 +64,9 @@
 ======= DONE =======
 ppp(8) (compiler bug, hack in http://people.freebsd.org/~peter/hammer.diff,
   but gcc-3.3 would be much better)
+
+ACPI
+(compiles/runs.. test more)
+
+boot blocks (need gcc -m32 to work)
+(fixed in p4 tree, needs binutils patches etc)