We currently implement `TCGETS`, `TCSETS` and `TCSETSW` which operate on
a `Termios` struct. glibc uses the `Termios2` variants of these
syscalls to additionally get the terminal baud rate.
Use a `Termios2` as the main state of the tty. Return that struct
verbatim for the newly implemented `Termios2` ioctls. Convert to/from
the legacy `Termios` struct for the currently implemented ioctls.
If a task has called one of `sys_exit` or `sys_exit_group`, don't follow
the standard syscall exit path. Since the task is dead and it will never
be rescheduled, there's no point in processing the result of sys_exit.
We currently union the UNMASKABLE_SIGNALS set with the new signal mask.
This does the complete opposite of what we wnat, we want to *remove*
those signals from the newly computed signal mask.
This patch removes the UNMASKABLE_SIGNALS set from any newly computed
signal mask.
Make it such that when a ptrace event is hit:
- The current regset is saved in the ptrace state.
- The current task is put to sleep.
- Arrange for a SIGCHLD to be set to the parent.
- Notify any waiters with the appopriate signal.
This commit refactors the signal handling subsystem to address a bug in
process creation and to improve the separation of concerns between
signal delivery and signal disposition.
1. Fix Action Table Sharing:
Previously, the signal action table was wrapped in an `Arc`, causing
forked child processes to inadvertently share signal dispositions with
the parent. `SignalActionState` now contains a direct `SigActionSet` and
implements Clone, ensuring that processes receive a private copy of the
action table upon fork/clone.
2. Decouple Signal Selection from Execution:
The logic for selecting a pending signal has been moved from the
disposition state into a new `take_signal` method on `OwnedTask`. This
decouples the "taking" of a signal (respecting masks and priorities)
from the "actioning" of that signal. This is a prerequisite for
implementing ptrace.
3. Various cleanup bits:
- Renamed `SignalState` to `SignalActionState` to more accurately
reflect that it manages signal dispositions.
- Removed the `pending` signal set out of the action state and directly
into the `ThreadGroup` and `OwnedTask`.
- Renamed `set_pending` to `set_signal` for consistency.
- Simplified the signal delivery loop in `dispatch_userspace_task`
using the new `take_signal` API.
Add a new option `--init-arg` which allows arguments to be passed to
init when invoked by the kernel. This allows the `-i` parameter to be
passed through to bash such that it will be started in interactive mode.
Add the proper inode and fs id to the `FileAttr` struct returned by
`getattr`. This prevents the dynamic loader from skipping files which it
considers to be hard-links (share the same inode ID).
Implement FUTEX_{WAKE,WAIT}_BITSET by using a `u32` as a data
discriminant value.
Also add a bunch of tests to `usertest` to ensure proper functionality.
Currently, when a kernel stack overflow occures, the exception handler
blindly attempts to write the current context to the stack. If the SP
isn't valid this causes another fault, and so on - locking up the
system.
This commit re-arranges the stack layout, performs SP validation before
usage and switches to an emergency stack when SP isn't valid. This
allows the handler to run and panic gracefully.
Fixes: #98
Currently, each task implements it's own priority value. In Linux, each
thread group (process) has a default process which all tasks in that
group inherit. Tasks can, however, override the default process
priority.
Implement that logic here which also fixes the current compilation error
on master.
