.. coderef:: src/console.c :first_of_file: :anchor: typedef struct console { /* Initialise a console object. */ int (*open)(struct console *obj); /* Finish a console object. */ int (*close)(struct console *obj); /* Read from a console - if no data is available, return zero without blocking. Else return as much data as is available, up to 'len'. Return the number of bytes read, or -1 on error. */ int (*read)(struct console *obj, char *buf, int len); /* Write to a console - return the number of bytes written, or -1 on failure. */ int (*write)(struct console *obj, const char *buf, int len); /* Flush the write buffer of a console. */ void (*flush)(struct console *obj); /* Intrusive linked list, for HAL's use only. */ struct console *prev, *next; /* Implementation dependent data. */ void *data; } console_t; /* Register 'c' as a new console. Returns zero on success. */ int register_console(console_t *c); /* Unregister 'c' from the set of consoles. */ void unregister_console(console_t *c); /* Write to all consoles. */ void write_console(const char *buf, int len); /* Read from the default console into a buffer. If there is no data available, this call will block. Returns the number of bytes actually read, or -1 on error. */ int read_console(char *buf, int len); The kernel console ================== After two rather long chapters - one explaining the intricacies of how this project is structured, the other how to build bare metal - I'm happy to say that this part of the walkthrough is rather short. The initial aim with a kernel should be getting output out of it as quickly as possible - so you know what's actually going on. There are several ways of having the kernel communicate with the outside world - you can print to the screen (the most common way), send characters down a serial line (very useful in headless situations), or perhaps write to a logfile (perhaps less useful until the kernel has booted sufficiently!) Our aim for this chapter is to build the mechanism that allow us to write to one of these devices, and to read from input devices (keyboard, serial). The idea is to create a multiplexer - modules can register with this multiplexer and say "I know how to output data". Then, when the user wants to log some information, this is sent to all registered modules, broadcast style. Similarly modules can register that they can receive input, so when input is required, it can be gained from a number of sources. In "hal.h", which we introduced in the first chapter, there is a structure for this and associated helper functions - ``console_t`` and friends. .. raw:: html
.. coderef:: src/console.c :anchor: .. raw:: html
.. coderef:: src/console.c :anchor: #include "hal.h" /* The first console in a linked list. */ static console_t *consoles = NULL; /* Lock for all console operations. */ static spinlock_t lock = SPINLOCK_RELEASED; /* Registers a new console - declared in hal.h */ int register_console(console_t *c) { spinlock_acquire(&lock); if (consoles) consoles->prev = c; c->next = consoles; c->prev = NULL; consoles = c; /* If an open() function was provided, call it. */ if (c->open) c->open(c); spinlock_release(&lock); return 0; } So let's start defining our console multiplexer. The ``console_t`` structure is a doubly-linked-list, so to register a new console all we have to do is push the given object onto the front of that list. Note we also follow our usual practice of using stuff before we've defined it - in this case spinlocks, which will be used once we have multithreading to synchronise calls to the console functions. .. raw:: html
.. coderef:: src/console.c :anchor: /* Unregisters a console - declared in hal.h */ void unregister_console(console_t *c) { spinlock_acquire(&lock); console_t *prev = NULL; console_t *this = consoles; /* Scan through the linked list looking for 'c'. */ while (this) { if (this == c) { if (this->next) this->next->prev = prev; if (this->prev) this->prev->next = this->next; if (!prev) consoles = c; /* Found - call flush() then close() if they exist. */ if (this->flush) this->flush(this); if (this->close) this->close(this); break; } prev = this; this = this->next; } spinlock_release(&lock); } Similarly for unregistering a console - just search through the list of consoles and remove the offending item. Again we require thread synchronisation. .. raw:: html
.. coderef:: src/console.c :anchor: /* Writes to a console - declared in hal.h */ void write_console(const char *buf, int len) { spinlock_acquire(&lock); console_t *this = consoles; while (this) { if (this->write) this->write(this, buf, len); this = this->next; } spinlock_release(&lock); } Then we get to define writing and reading from the console. Writing is a simple broadcast operation - take the input and write it to all registered consoles. .. raw:: html
.. coderef:: src/console.c :anchor: /* Reads from a console - declared in hal.h */ int read_console(char *buf, int len) { if (len == 0) return 0; spinlock_acquire(&lock); console_t *this = consoles; while (this) { if (this->read) { int n = this->read(this, buf, len); if (n > 0) { spinlock_release(&lock); return n; } } this = this->next; if (!this) this = consoles; } spinlock_release(&lock); return -1; } Reading is slightly different - the ``read()`` functions defined in ``console_t`` are supposed to be non-blocking. So we cycle through all registered consoles trying to find one for whom ``read()`` is defined and returns a number of bytes greater than zero (0 return value means no data was available). .. raw:: html
.. coderef:: src/console.c :anchor: /* Flush and close all consoles. */ static int shutdown_console() { console_t *this = consoles; while (this) { if (this->flush) this->flush(this); if (this->close) this->close(this); this = this->next; } return 0; } static module_t x run_on_startup = { .name = "console", .required = NULL, .load_after = NULL, .init = NULL, .fini = &shutdown_console, }; Finally we define the function that will clean up any consoles active at time of shutdown and flush their contents, and define the ``module_t`` structure that will register us as a module! .. raw:: html
That's it! But it currently doesn't do anything, so you should swiftly advance onto one of the next chapters to implement an output method. .. raw:: html