| Age | Commit message (Collapse) | Author |
|---|
|
Add support for managing the page table entry dirty bit. This will allow
vm(9) to know when to page-out data to backing store, among other
things.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
* Switched to using macros for the GDT entry bits, to improve clarity
and make the code easier to modify. Also got rid of some junk values
in a few of the entries.
* Shrunk the GDT data array from 256 entries to 7 (GDT_ENTRY_COUNT)
since the GDT will only ever use 7 of the entries for now.
* Aligned the GDT using __cacheline_aligned, because the GDT entries
will be read every time a segment selector is loaded.
* Modified code dealing with the GDTR to just use `g_gdtr`, since the
same GDT/GDTR is always used.
* In `gdt_load()`, changed `eax` -> `ax`, since segment registers are
only 16 bits. Also replaced the `gdtr` to just use `g_gdtr`.
Additionally, used the inline assembly formatting to supply segment
selectors with KERNEL_CS/DS macros.
Signed-off-by: Quinn Stephens <quinn@osmora.org>
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Replace intr_alloc_vector() with a cleaner and more machine independent
intr_register()
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Ensure interrupts are disabled at the start of panic()
to prevent any interrupts from further degrading the
state of the system. Also adds a newline before "panic: "
to clean up the logs.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Simplicity is divine, fork() may be powerful but is no longer simple. It
became a thing in the late 70s during the early days of UNIX when computing
was simple, when CPUs were only 16-bits, MMUs were not prevalent and RAM was
only 512 KB. However it isn't 1971 anymore, process management, CPUs and memory
architecture has advanced significantly since. This commit ceases work on the
fork() syscall as implementing it would only introduce unnecessary complexity,
security issues, hefty processing overhead and would perpetuate what should
now be considered legacy. The current best alternative would be providing
a form of process spawning as well as a mechanism to wait for the child
process to complete.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Kernel logs must now all be lowercase for consistency. A new style
called Peripheral Description Notation (PDN) has also been introduced
to describe devices.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
This commit introduces a bus resource abstraction. A bus resource
encapsulates bus specific information and semantics. Along with
abstracting away bus specific details and providing an overall
smoother interface, this additionally provides flexibility, access
control and security by keeping track of permissions and allowing
things to be turned on or off.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Was supposed to happen on Jan 1, sorry! Happy late new year!
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Added more defines and comments to the UART driver, based on the 16550
chip datasheet, to make it more clear how each operation works.
Signed-off-by: Quinn Stephens <quinn@osmora.org>
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Ensure all loads preceding SWAPGS are serialized using LFENCE. This
mitigates the possibility of SWAPGS being speculatively bypassed which
would lead to security issues.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Maintain naming consistency between public machine-dependent and
machine-independent functions.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Instead of writing cpu_info to IA32_KERNEL_GS_BASE then using SWAPGS,
write the cpu_info pointer directly to IA32_GS_BASE
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
- Use only defines in frame.h
- Improve handling of trap error codes
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
This commit changes how reboot() handles REBOOT_HALT. Now instead of one
core being halted, all cores will be halted through an IPI to a halt
vector.
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
|
Signed-off-by: Ian Moffett <ian@osmora.org>
|
