[SysCalls] Implement System API

[wake-0]

Implement Sytem API e.g. MINIX SysCalls.

Atm. the following calls should be implemented:

system.h:

1.) execute(char* command) 2.) yield() which switches to the next process 3.) exit() which quits the current process 4.) sleep(int) which sets a process's state to SLEEPING for the given amount of ms 5.) shutdown()

process.h:

1.) read_processes(..) returns a list of the running processes including meta-information 2.) kill_process(int) kills the process with the given id 3.) get_process_count() returns the number of running processes

filesystem.h:

1.) read_cwd(char* buf, int len) returns the current working directory into the given buf 2.) set_cwd(char* path) sets the current working directory to the given path 3.) read_file(char* name, char* buf, int len) opens the given file and reads len bytes of the file 4.) read_directory(char* name, ...) reads the directory content and returns containing directories and files including meta information

ipc.h:

1.) open_ipc_channel(..) opens a namespace in the IpcManager 2.) send_ipc_message(..) sends a message to a namespace 3.) has_ipc_message(..) returns true or false

stdio.h:

1.) print(char*, int) prints the given string to the console (this is a helper function only for getting printfto work in ext. processes)

devices.h:

1.) open_device(char* devicename) opens the device and returns the handle as int 2.) close_device(int) closes the device 3.) ioctl_device(int handle, int mode, int cmd, int len, char* buf) calls ioctl on the device 4.) ...

[trylimits]

Some clarifications on the system API:

We should have somewhere a folder api which contains all necessary header files and source files for the system calls. A system call immediately switches into supervisor/privileged mode and executes the corresponding kernel function and switches back to user mode. The memory mapping of TTBR1 (kernel) has to be read/write-protected for user mode thus triggering a interrupt if a user process tries to access the kernel by bypassing the system API. This ISR should immediately kill the process.

[trylimits]

Above comment is not entirely true. We discuss the concept in detail at Wed. See #75. If I find some time I will probably implement a prototype for a system call.

[trylimits]

In the above commit I have refactored some code (according to our conventions) and implemented a working system call from the console to the kernel to start a process.

Discussion tomorrow I guess.

[ghost]

I'm sorry to say that the last commit is not 100% correct. Further explanation at our meeting.

[trylimits]

I refactored the System API to a more useful structure. The main idea is that all system call functions are wrapped into a user API, such as execute(char* command) and within that function a systemCallMessage_t is assembled. The user functions should be placed into header files which fulfill the POSIX standard, e.g. execute(..) is placed into system.h.

[wake-0]

Acc. to todays coaching readfile should only read chunks from the sd card so that there is not the problem that reading needs to much time.

[trylimits]

Above commits implement the API for accessing the filesystem.

[trylimits]

The function SystemCall(..) in systemcall.c includes a switch case handling the different system call ids. However, the systemCallMessage_t struct already contains this information. In the kernel implementation we use the system call id from the latter, thus removing this switch case would make the implementation cleaner.

[trylimits]

Above commit removes this handling and calls the SWI instruction with a constant number. All the system calls still work.

[trylimits]

System API functions are done and are working well. Addtional functions should be discussed in new tickets.