Subversion Repositories HelenOS

Rev

Go to most recent revision | Show changed files | Details | Compare with Previous | Blame | RSS feed

Filtering Options

Rev Age Author Path Log message Diff
205 7005 d 11 h jermar /SPARTAN/trunk/ Get rid of memcopy().
All we now have is memcpy() and _memcpy().
memcpy() is defined to be builtin.
Where not applicable, architectures must implement memcpy() code or call slowish _memcopy().		  
195 7006 d 1 h vana /SPARTAN/trunk/ 1) memcopy and _memcopy functions rewriten to ANSI C norm.
2) Repaired ia32,ia64 and mips version of SPARTAN to work with this memcopy functions
3) Warning for non declared funcions added and repaired ia32,ia64 and mips versions to pass build process with this warning and Werror option		  
167 7017 d 8 h jermar /SPARTAN/trunk/ Switch over to per vm_t page tables. (breaks ia32 userspace)

Define dummy memcopy() for amd64 and ppc to satisfy compiler.

Get rid of problematic #include <arch/mm/page.h> in mm/page.h.

Indentation fixes and small changes here and there.		  
125 7051 d 23 h jermar /SPARTAN/trunk/ Huge indentation fix.  
120 7066 d 22 h jermar /SPARTAN/trunk/ Generic 4-level page tables interface & implementation (review & test).
Implement more architecture dependant macros on IA-32.
Provide fake implementation on IA-64 and MIPS.
Convert map_page_to_frame() to use the new interface.
Move adjusted map_page_to_frame() from IA-32's arch/mm/page.c to the generic mm/page.c.		  
116 7087 d 5 h jermar /SPARTAN/trunk/ Add map_structure() to automate mapping of memory structures that can span multiple pages and/or cross page boundaries.
Change ACPI map_sdt() to use map_structure().

Small changes in MPS code.
The extra frame allocation for accessing frame 0 is unnecessary as it is possible to access frame 0 from kernel address space.
Zero TSS descriptor in the newly prepared GDT.		  
115 7087 d 21 h jermar /SPARTAN/trunk/ Implement several assembler functions in gcc's __asm__ notation instead of in .s or .S file.
Gain both better speed and size.		  
113 7090 d 20 h jermar /SPARTAN/trunk/ SMP recovery patch #2 (SMP is no longer broken !!!).
Fix missing KA2PA() operation in ap.S which was causing page faults during AP early initialization.
Fix bug in map_page_to_frame(): 'root' was interpretted as kernel address while read_dba() returns physical address.
Make references to page directory and page tables use kernel addresses instead of physical addresses.

Massive frame allocation code cleanup.
Basically revert to what we had had before implementation of userspace.

Usual cosmetics.		  
110 7092 d 6 h jermar /SPARTAN/trunk/ SMP recovery patch #1. (SMP still broken)
Move ap_boot() to K_TEXT_START_2 section.
Change ap_boot() to cope with mapping of kernel address space.
Change some addresses in mps.c from kernel to physical.

MIPS.
Move msim dprinter device to 0xB0000000.
This address is from kseg1 (unmapped, uncached) segment.		  
109 7094 d 1 h jermar /SPARTAN/trunk/ Discard 'copy' parameter of map_page_to_frame().
Introduce 'root' parameter of map_page_to_frame().
If 'root' is zero, page directory address is read from CPU.
Otherwise 'root' is used as page directory address.

Add missing word to new doxygen-style comments in scheduler.c.		  
106 7094 d 8 h jermar /SPARTAN/trunk/ IA-32 fixes.
Use kernel addresses instead of physical addresses in map_page_to_frame().
Physical addresses are supposed to only be exported to mm hardware.
Because of this fix, userspace is functional again.

Remap EGA videoram to (0x80000000 + videoram) and change the ega driver to work with the new address.

Minor cosmetics through out the code.
Changes in linker scripts.		  
105 7094 d 21 h jermar /SPARTAN/trunk/ Big changes in IA-32 address space map.
Now the kernel is mapped above 0x80000000. Finally!
Userspace address space starts at 0x00000000.
Changes in many places.
This improvement temporarily breaks SMP and most likely also other stuff.
Supported size of memory is now only 4M as it is the biggest size that can be mapped at once on IA-32.

Changes in linker script.
Changes required because of the above.
Do not patch hardcoded_* variables but assign to them instead.

Cosmetic changes here and there.		  
34 7128 d 4 h jermar /SPARTAN/trunk/ More ACPI work.
Initial MADT table parsing.

SMP renaming and reorganization to reflect there are more ways to bring SMP up.		  
11 7191 d 21 h jermar /SPARTAN/trunk/ Clean up (ia32 vs. i386).
Header files reorganization.		  
5 7195 d 23 h jermar /SPARTAN/trunk/ TLB shootdown.  
1 7199 d 3 h jermar /SPARTAN/ Initial import  

←Prev12Next→

Show All