sortix--sortix/kernel/net/fs.cpp

/*******************************************************************************

    Copyright(C) Jonas 'Sortie' Termansen 2013.

    This file is part of Sortix.

    Sortix is free software: you can redistribute it and/or modify it under the
    terms of the GNU General Public License as published by the Free Software
    Foundation, either version 3 of the License, or (at your option) any later
    version.

    Sortix is distributed in the hope that it will be useful, but WITHOUT ANY
    WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
    FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
    details.

    You should have received a copy of the GNU General Public License along with
    Sortix. If not, see <http://www.gnu.org/licenses/>.

    net/fs.cpp
    Filesystem based socket interface.

*******************************************************************************/

// TODO: Should this be moved into user-space?

#include <sys/un.h>

#include <assert.h>
#include <errno.h>
#include <stddef.h>
#include <stdint.h>
#include <string.h>

#include <sortix/fcntl.h>
#include <sortix/poll.h>
#include <sortix/socket.h>
#include <sortix/stat.h>

#include <sortix/kernel/descriptor.h>
#include <sortix/kernel/inode.h>
#include <sortix/kernel/ioctx.h>
#include <sortix/kernel/kernel.h>
#include <sortix/kernel/kthread.h>
#include <sortix/kernel/mtable.h>
#include <sortix/kernel/pipe.h>
#include <sortix/kernel/poll.h>
#include <sortix/kernel/process.h>
#include <sortix/kernel/refcount.h>

#include "fs.h"

// TODO: This is declared in the <sys/socket.h> header that isn't ready for
//       kernel usage, so we declare it here for now.
#ifndef AF_UNIX
#define AF_UNIX 3
#endif

namespace Sortix {
namespace NetFS {

class Manager;
class StreamSocket;

class Manager : public AbstractInode
{
public:
	Manager(uid_t owner, gid_t group, mode_t mode);
	virtual ~Manager() { }
	virtual Ref<Inode> open(ioctx_t* ctx, const char* filename, int flags,
	                        mode_t mode);

public:
	bool Listen(StreamSocket* socket);
	void Unlisten(StreamSocket* socket);
	Ref<StreamSocket> Accept(StreamSocket* socket, ioctx_t* ctx, uint8_t* addr,
	                         size_t* addrsize, int flags);
	int AcceptPoll(StreamSocket* socket, ioctx_t* ctx, PollNode* node);
	bool Connect(StreamSocket* socket);

private:
	StreamSocket* LookupServer(struct sockaddr_un* address);

private:
	StreamSocket* first_server;
	StreamSocket* last_server;
	kthread_mutex_t manager_lock;

};

class StreamSocket : public AbstractInode
{
public:
	StreamSocket(uid_t owner, gid_t group, mode_t mode, Ref<Manager> manager);
	virtual ~StreamSocket();
	virtual Ref<Inode> accept(ioctx_t* ctx, uint8_t* addr, size_t* addrsize,
	                          int flags);
	virtual int bind(ioctx_t* ctx, const uint8_t* addr, size_t addrsize);
	virtual int connect(ioctx_t* ctx, const uint8_t* addr, size_t addrsize);
	virtual int listen(ioctx_t* ctx, int backlog);
	virtual ssize_t recv(ioctx_t* ctx, uint8_t* buf, size_t count, int flags);
	virtual ssize_t send(ioctx_t* ctx, const uint8_t* buf, size_t count,
	                     int flags);
	virtual ssize_t read(ioctx_t* ctx, uint8_t* buf, size_t count);
	virtual ssize_t write(ioctx_t* ctx, const uint8_t* buf, size_t count);
	virtual int poll(ioctx_t* ctx, PollNode* node);

private:
	int do_bind(ioctx_t* ctx, const uint8_t* addr, size_t addrsize);

public: /* For use by Manager. */
	PollChannel accept_poll_channel;
	Ref<Manager> manager;
	PipeEndpoint incoming;
	PipeEndpoint outgoing;
	StreamSocket* prev_socket;
	StreamSocket* next_socket;
	StreamSocket* first_pending;
	StreamSocket* last_pending;
	struct sockaddr_un* bound_address;
	bool is_listening;
	bool is_connected;
	bool is_refused;
	kthread_mutex_t socket_lock;
	kthread_cond_t pending_cond;
	kthread_cond_t accepted_cond;

};

static void QueueAppend(StreamSocket** first, StreamSocket** last,
                        StreamSocket* socket)
{
	assert(!socket->prev_socket);
	assert(!socket->next_socket);
	socket->prev_socket = *last;
	socket->next_socket = NULL;
	if ( *last )
		(*last)->next_socket = socket;
	if ( !*first )
		*first = socket;
	*last = socket;
}

static void QueueRemove(StreamSocket** first, StreamSocket** last,
                        StreamSocket* socket)
{
	if ( socket->prev_socket )
		socket->prev_socket->next_socket = socket->next_socket;
	else
		*first = socket->next_socket;
	if ( socket->next_socket )
		socket->next_socket->prev_socket = socket->prev_socket;
	else
		*last = socket->prev_socket;
}

StreamSocket::StreamSocket(uid_t owner, gid_t group, mode_t mode,
                           Ref<Manager> manager)
{
	inode_type = INODE_TYPE_STREAM;
	dev = (dev_t) manager.Get();
	ino = (ino_t) this;
	this->type = S_IFSOCK;
	this->stat_uid = owner;
	this->stat_gid = group;
	this->stat_mode = (mode & S_SETABLE) | this->type;
	this->prev_socket = NULL;
	this->next_socket = NULL;
	this->first_pending = NULL;
	this->last_pending = NULL;
	this->bound_address = NULL;
	this->is_listening = false;
	this->is_connected = false;
	this->is_refused = false;
	this->manager = manager;
	this->socket_lock = KTHREAD_MUTEX_INITIALIZER;
	this->pending_cond = KTHREAD_COND_INITIALIZER;
	this->accepted_cond = KTHREAD_COND_INITIALIZER;
}

StreamSocket::~StreamSocket()
{
	if ( is_listening )
		manager->Unlisten(this);
	delete[] bound_address;
}

Ref<Inode> StreamSocket::accept(ioctx_t* ctx, uint8_t* addr, size_t* addrsize,
                                int flags)
{
	ScopedLock lock(&socket_lock);
	if ( !is_listening )
		return errno = EINVAL, Ref<Inode>(NULL);
	return manager->Accept(this, ctx, addr, addrsize, flags);
}

int StreamSocket::do_bind(ioctx_t* ctx, const uint8_t* addr, size_t addrsize)
{
	if ( is_connected || is_listening || bound_address )
		return errno = EINVAL, -1;
	size_t path_offset = offsetof(struct sockaddr_un, sun_path);
	size_t path_len = (path_offset - addrsize) / sizeof(char);
	if ( addrsize < path_offset )
		return errno = EINVAL, -1;
	uint8_t* buffer = new uint8_t[addrsize];
	if ( !buffer )
		return -1;
	if ( ctx->copy_from_src(buffer, addr, addrsize) )
	{
		struct sockaddr_un* address = (struct sockaddr_un*) buffer;
		if ( address->sun_family == AF_UNIX )
		{
			bool found_nul = false;
			for ( size_t i = 0; !found_nul && i < path_len; i++ )
				if ( address->sun_path[i] == '\0' )
					found_nul = true;
			if ( found_nul )
			{
				bound_address = address;
				return 0;
			}
			errno = EINVAL;
		}
		else
			errno = EAFNOSUPPORT;
	}
	delete[] buffer;
	return -1;
}

int StreamSocket::bind(ioctx_t* ctx, const uint8_t* addr, size_t addrsize)
{
	ScopedLock lock(&socket_lock);
	return do_bind(ctx, addr, addrsize);
}

int StreamSocket::connect(ioctx_t* ctx, const uint8_t* addr, size_t addrsize)
{
	ScopedLock lock(&socket_lock);
	if ( is_listening )
		return errno = EINVAL, -1;
	if ( is_connected )
		return errno = EISCONN, -1;
	if ( addr && do_bind(ctx, addr, addrsize) != 0 )
		return -1;
	if ( !bound_address )
		return errno = EINVAL, -1;
	return manager->Connect(this) ? 0 : -1;
}

int StreamSocket::listen(ioctx_t* /*ctx*/, int /*backlog*/)
{
	ScopedLock lock(&socket_lock);
	if ( is_connected || is_listening || !bound_address )
		return errno = EINVAL, -1;
	if ( !manager->Listen(this) )
		return -1;
	return 0;
}

ssize_t StreamSocket::recv(ioctx_t* ctx, uint8_t* buf, size_t count,
                           int /*flags*/)
{
	ScopedLock lock(&socket_lock);
	if ( !is_connected )
		return errno = ENOTCONN, -1;
	return incoming.read(ctx, buf, count);
}

ssize_t StreamSocket::send(ioctx_t* ctx, const uint8_t* buf, size_t count,
                           int /*flags*/)
{
	ScopedLock lock(&socket_lock);
	if ( !is_connected )
		return errno = ENOTCONN, -1;
	return outgoing.write(ctx, buf, count);
}

ssize_t StreamSocket::read(ioctx_t* ctx, uint8_t* buf, size_t count)
{
	return recv(ctx, buf, count, 0);
}

ssize_t StreamSocket::write(ioctx_t* ctx, const uint8_t* buf, size_t count)
{
	return send(ctx, buf, count, 0);
}

int StreamSocket::poll(ioctx_t* ctx, PollNode* node)
{
	if ( is_connected )
	{
		PollNode* slave = node->CreateSlave();
		if ( !slave )
			return -1;
		int incoming_result = incoming.poll(ctx, node);
		int outgoing_result = outgoing.poll(ctx, slave);
		return incoming_result == 0 || outgoing_result == 0 ? 0 : -1;
	}
	if ( is_listening )
		return manager->AcceptPoll(this, ctx, node);
	return errno = ENOTCONN, -1;
}

Manager::Manager(uid_t owner, gid_t group, mode_t mode)
{
	inode_type = INODE_TYPE_UNKNOWN;
	dev = (dev_t) this;
	ino = 0;
	this->type = S_IFDIR;
	this->stat_uid = owner;
	this->stat_gid = group;
	this->stat_mode = (mode & S_SETABLE) | this->type;
	this->manager_lock = KTHREAD_MUTEX_INITIALIZER;
	this->first_server = NULL;
	this->last_server = NULL;
}

static int CompareAddress(const struct sockaddr_un* a,
                          const struct sockaddr_un* b)
{
	return strcmp(a->sun_path, b->sun_path);
}

StreamSocket* Manager::LookupServer(struct sockaddr_un* address)
{
	for ( StreamSocket* iter = first_server; iter; iter = iter->next_socket )
		if ( CompareAddress(iter->bound_address, address) == 0 )
			return iter;
	return NULL;
}

static StreamSocket* QueuePop(StreamSocket** first, StreamSocket** last)
{
	StreamSocket* ret = *first;
	assert(ret);
	QueueRemove(first, last, ret);
	return ret;
}

bool Manager::Listen(StreamSocket* socket)
{
	ScopedLock lock(&manager_lock);
	if ( LookupServer(socket->bound_address) )
		return errno = EADDRINUSE, false;
	QueueAppend(&first_server, &last_server, socket);
	socket->is_listening = true;
	return true;
}

void Manager::Unlisten(StreamSocket* socket)
{
	ScopedLock lock(&manager_lock);
	while ( socket->first_pending )
	{
		socket->first_pending->is_refused = true;
		kthread_cond_signal(&socket->first_pending->accepted_cond);
		socket->first_pending = socket->first_pending->next_socket;
	}
	socket->last_pending = NULL;
	QueueRemove(&first_server, &last_server, socket);
	socket->is_listening = false;
}

int Manager::AcceptPoll(StreamSocket* socket, ioctx_t* /*ctx*/, PollNode* node)
{
	ScopedLock lock(&manager_lock);
	if ( socket->first_pending &&
	    ((POLLIN | POLLRDNORM) & node->events) )
		return node->master->revents |= ((POLLIN | POLLRDNORM) & node->events), 0;
	socket->accept_poll_channel.Register(node);
	return errno = EAGAIN, -1;
}

Ref<StreamSocket> Manager::Accept(StreamSocket* socket, ioctx_t* ctx,
	                              uint8_t* addr, size_t* addrsize, int /*flags*/)
{
	ScopedLock lock(&manager_lock);

	// TODO: Support non-blocking accept!
	while ( !socket->first_pending )
		if ( !kthread_cond_wait_signal(&socket->pending_cond, &manager_lock) )
			return errno = EINTR, Ref<StreamSocket>(NULL);

	StreamSocket* client = socket->first_pending;

	struct sockaddr_un* client_addr = client->bound_address;
	size_t client_addr_size = offsetof(struct sockaddr_un, sun_path) +
	                          (strlen(client_addr->sun_path)+1) * sizeof(char);

	if ( addr )
	{
		size_t caller_addrsize;
		if ( !ctx->copy_from_src(&caller_addrsize, addrsize, sizeof(caller_addrsize)) )
			return Ref<StreamSocket>(NULL);
		if ( caller_addrsize < client_addr_size )
			return errno = ERANGE, Ref<StreamSocket>(NULL);
		if ( !ctx->copy_from_src(addrsize, &client_addr_size, sizeof(client_addr_size)) )
			return Ref<StreamSocket>(NULL);
		if ( !ctx->copy_to_dest(addr, client_addr, client_addr_size) )
			return Ref<StreamSocket>(NULL);
	}

	// TODO: Give the caller the address of the remote!

	Ref<StreamSocket> server(new StreamSocket(0, 0, 0666, Ref<Manager>(this)));
	if ( !server )
		return Ref<StreamSocket>(NULL);

	QueuePop(&socket->first_pending, &socket->last_pending);

	if ( !client->outgoing.Connect(&server->incoming) )
		return Ref<StreamSocket>(NULL);
	if ( !server->outgoing.Connect(&client->incoming) )
	{
		client->outgoing.Disconnect();
		server->incoming.Disconnect();
		return Ref<StreamSocket>(NULL);
	}

	client->is_connected = true;
	server->is_connected = true;

	// TODO: Should the server socket inherit the address of the listening
	//       socket or perhaps the one of the client's source/destination, or
	//       nothing at all?

	kthread_cond_signal(&client->accepted_cond);

	return server;
}

bool Manager::Connect(StreamSocket* socket)
{
	ScopedLock lock(&manager_lock);
	StreamSocket* server = LookupServer(socket->bound_address);
	if ( !server )
		return errno = ECONNREFUSED, false;

	socket->is_refused = false;

	QueueAppend(&server->first_pending, &server->last_pending, socket);
	kthread_cond_signal(&server->pending_cond);
	server->accept_poll_channel.Signal(POLLIN | POLLRDNORM);

	while ( !(socket->is_connected || socket->is_refused) )
		if ( !kthread_cond_wait_signal(&socket->accepted_cond, &manager_lock) &&
		    !(socket->is_connected || socket->is_refused) )
		{
			QueueRemove(&server->first_pending, &server->last_pending, socket);
			return errno = EINTR, false;
		}

	return !socket->is_refused;
}

// TODO: Support a poll method in Manager.

Ref<Inode> Manager::open(ioctx_t* /*ctx*/, const char* filename,
                             int /*flags*/, mode_t /*mode*/)
{
	if ( !strcmp(filename, "stream") )
	{
		StreamSocket* socket = new StreamSocket(0, 0, 0666, Ref<Manager>(this));
		return Ref<StreamSocket>(socket);
	}
	return errno = ENOENT, Ref<Inode>(NULL);
}

void Init(const char* devpath, Ref<Descriptor> slashdev)
{
	ioctx_t ctx; SetupKernelIOCtx(&ctx);
	Ref<Inode> node(new Manager(0, 0, 0666));
	if ( !node )
		PanicF("Unable to allocate %s/net/fs inode.", devpath);
	// TODO: Race condition! Create a mkdir function that returns what it
	// created, possibly with a O_MKDIR flag to open.
	if ( slashdev->mkdir(&ctx, "net", 0755) < 0 && errno != EEXIST )
		PanicF("Could not create a %s/net directory", devpath);
	if ( slashdev->mkdir(&ctx, "net/fs", 0755) < 0 && errno != EEXIST )
		PanicF("Could not create a %s/net/fs directory", devpath);
	Ref<Descriptor> mpoint = slashdev->open(&ctx, "net/fs", O_READ | O_WRITE, 0);
	if ( !mpoint )
		PanicF("Could not open the %s/net/fs directory", devpath);
	Ref<MountTable> mtable = CurrentProcess()->GetMTable();
	// TODO: Make sure that the mount point is *empty*! Add a proper function
	// for this on the file descriptor class!
	if ( !mtable->AddMount(mpoint->ino, mpoint->dev, node) )
		PanicF("Unable to mount filesystem on %s/net/fs", devpath);
}

} // namespace NetFS
} // namespace Sortix