Summary of changes:
- kernel/mem renamed to kernel/mm
- consolidated page/frame defs into one file which now lives in the
kernel/mm package and is referenced by both pmm and vmm pkgs
- consolidated parts of the vmm code (e.g. PDT+PTE)
- memcopy/memset helpers moved to the kernel package
- physical allocators moved to the kernel/mm/pmm package
- break vmm -> pmm pkg dependency by moving AllocFrame() into the mm
package.
This reverts commit 8f04deadc1f1620dbd0aaea07f04a6c7f82f58c0.
The actual issue was triggered by a memory corruption due to the fact
that the exception handling gate code did not preserve XMM regs.
The bug manifested itself as a series of null bytes appearing to the
tty device when the ring buffer gets linked to it and its contents are
flushed to the tty.
Using gdb, I was able to track the problem into the ringbuf's Read
method and specifically to the call to copy(). I have replaced the call
with a for loop that the compiler would anyway optimize into a rep stosb
or equivalent asm instruction.
Due to some changes in the runtime init functions between go 1.7 and
later versions (e.g. runtime.modulesInit() defined after go 1.7), the
kernel code that bootstraps the go runtime had to be split into separate
files which are conditionally compiled using +build flags.
The code responsible for mapping the kernel sections worked under the
assumption that the linker would align all sections on a page boundary.
To figure out the page extents for a particular section, the
implementation would round the section VMA down to the nearest page and
add to that the section length rounded up to the nearest page size.
However, some sections (e.g. noptrbss) use 32 byte alignment. Depending
on the section length, the original implementation could in some cases
skip mapping of the last page in the section which would cause a page
fault when its contents were accessed.
The fix for this issue is quite simple: calculate the end page by
rounding up (section start addr + section length) to the next page.
vmm.IdentityMapRegion can be used by device drivers that want to
establish an identity mapping for a contiguous physical memory block in
order to access some hardware or table.
The hal package automatically selects the best font from the list of
available fonts based on the console dimensions and the font priorities. The font
selection can be overriden by passing the "consoleFont" boot commandline
parameter to the kernel (e.g. consoleFont=terminus8x16)
Ths page offset is defined in arch/XXX/constants.inc and needs to be
passed to the kernel so we can correctly calculate the physical frame
addresses that correspond to the ELF section virtual memory addresses.
This commit removes the HWProbes() function from the console and tty
packages and replaces it with a global ProbeFuncs slice which is fetched
by the hal package when the hardware autodetection code runs.
Each driver should provide an init() function that appends a probe function
to the global ProbeFuncs slice.
This approach allows us to support conditional compilation of drivers in
the future (e.g. using build tags)
The implementation of Printf has been moved from the early package to
the kfmt package. The dependency to ActiveTerminal has been removed and
the code now uses an io.Writer for its output. As Go interfaces cannot
be used before bootstrapping the Go runtime, the code uses a ring-buffer
fallback for storing any kernel output emitted before that point.
The ring-buffer implements both io.Reader and io.Writer and uses a fixed
size of 2048 bytes (set by the ringBufferSize constant). This provides
enough space to hold a standard 80x25 screen's output.
This commit refactors the old VT implementation to work with the revised
TTY interface and adds support for:
- scrollback
- terminal state handling
When a terminal becomes activated, it overwrites the attached console
contents with the contents of its viewport.
Things like error messages (e.g in the io pkg) are actually allocated
when init() is executed. Unless we trigger a call to init(), values like
this will be nil causing various problems when we try to use functions
from the stdlib.
By setting up pwd as a Go workspace, we can trim import paths from
something like "github.com/achilleasa/gopher-os/kernel" to just
"kernel".
These changes make forking easier and also allows us to move the code to
a different git hosting provider without having to rewrite the imports.
