This enables useful features such as directories, CRC32 checksums, and other
useful features. The initrdfs in the kernel is now hooked up against the new
API, although the kernel's current limited FS support is a problem for now.
To work around that, directories are not supported at runtime, although the
internal API understands them wonderfully. This will be fixed when the
kernel gets a real VFS.
This provides control over the caching of memory, which makes write-combined
IO possible. Graphics drivers can use this to transfer data at a much higher
rate to the video memory.
The implementation is a bit hacky but it'll do for now. It provides enough
support for the experimental VBE driver to work on the real computers I
tested it on, even if the BIOS uses screwed up default MTRRs.
The virtual memory layer now automatically uses the PAT feature if available
but in a backwards compatible manner and otherwise just tries to approximate
PAT features if they are asked for.
sfork(2) now calls sforkr(2) with the current registers.
This will prove useful in creating threads, where user-space now can fully
control what state the child will start in. This is unlike the Linux clone
system call that accepts a pointer to the child stack; this is more powerful
and somehow simpler. Note that this will create a rather raw thread; no
thread initization has been done by the standard thread API (when it is
implemented), so this feature shouldn't be used by programmers unless they
know what they are doing.
fork(2) now calls sfork(2) directly. Also removed fork(2) and sfork(2) from
the kernel as they are done using sforkr(2) now. So technically they aren't
system calls right now, but that could always change.
sfork is much like rfork except sharing is default for everything.
Eventually, I'll make a rfork(3) wrapper function around sfork(2) to
provide compatibility to BSD programs.
I don't like Linux clone(2): that's some messy function.
Any detected COM ports available as /dev/comN.
It currently utilizes AGAIN to do polling in user-space. This prevents it
from locking up the whole system and makes it respond to the SIGINT hack.
There is also a more reliable and faster polling-blocking mode, but it locks
up the entire system.
The main interrupt mode is broken, perhaps by a bug in VirtualBox.
C and C++ files are now kept together and so are the mxmpp declarations.
Header files are now stored in include/ and mxmpp'd into preproc/.
All other code now -I ../libmaxsi/preproc.
And other stuff to make this happen, including refactoring Makefile.
<features.h> declares _SORTIX_SOURCE if no conflicting macros are
declared, such as _GNU_SOURCE.
Fixed g++ automatically declaring _GNU_SOURCE, but Sortix isn't GNU.
Replaced SORTIX_UNIMPLEMENTED macro with __SORTIX_SHOW_UNIMPLEMENTED.
A bool is set when a thread is terminated, which may help detect it.
A cached version of the thread's pid is also kept around.
And lastly, the thread is unsubscribed from events upon destruction.
This will allow development of a better terminal providing stdin.
Added new system calls settermmode(2) and gettermmode(2) declared in
<sys/termmode.h>. They allow querying and changing the current mode of
terminals (enabling raw keyboard data, signal handling, line buffering,
UTF-8 encoding stdin, and more). However, all that is unsupported by the
current terminal device driver.
Added KBKEY_ENCODE and KBKEY_DECODE macros to <sys/keycodes.h> which allows
encoding the kbkey format in UTF-32 characters.
Caps lock now works as caps lock, not as shift lock.
This new design will allow implementing a working tty, such that stdin is
the only way to access the keyboard, instead of the current hacky way of
using a special system call to read from the keyboard.
Added a new system header file <sys/keycodes.h> defining the constants for
every key on the keyboard. This will be used in future APIs.
The main change is to split the keyboard driver into a class that reads
from the keyboard, while another class handles the translation into
printable characters (if possible). This allows a terminal driver based
on logical key presses and printable characters, instead of a terminal
driver based only on unicode-ish codes.
Read operations are enabled by default, but you must set DISKWRITE=1 in
makeflags before write operations are permitted. This protects against
accidentally corrupting the existing filesystems on the system.
When compiled with gcc 4.6.1, 32-bit Sortix would triple fault during
early boot: When the TLB is being flushed, somehow a garbage value had
sneaked into Sortix::Memory::currentdir, and a non-page aligned (and
garbage) page directory is loaded. (Triple fault, here we come!)
However, adding a volatile addr_t foo after the currentdir variable
actually caused the system to boot correctly - the garbage was written
into that variable instead. To debug the problem, I set the foo value
to 0: as long as !foo (hence the name nofoo) everything was alright.
After closer examination I found that the initrd open code wrote to a
pointer supplied by kernel.cpp. The element pointed to was on the
stack. Worse, its address was the same as currentdir (now foo).
Indeed, the stack had gone into the kernel's data segment!
Turns out that this gcc configuration stores variables in the data
segment in the reverse order they are defined in, whereas previous
compilers did the opposite. The hack used to set up the stack during
early boot relied on this (now obviously incorrect) fact.
In effect, the stack was initialized to the end of the stack, not
the start of it: completely ignoring all the nice stack space
allocated in kernel.cpp.
I did not see that one coming.
This commit fixes some instances of uninitialized memory.
In addition, the bootstrap tables for x64 are moved around a bit,
in this awful game of placing stuff where it won't collide with grub.
Jonas 'Sortie' Termansen