error handling #30
Description
cottage/kernel/src/socket/socket.c
Line 178 in 5825b5a
#include <net/arp.h>
#include <net/route.h>
#include <proc/proc.h>
#include <file/file.h>
#include <fs/fs.h>
#include <errors/errno.h>
#include <socket/socket.h>
#include <mem/malloc.h>
#include <string.h>
#include <net/endian.h>
#include <net/network.h>
#include <net/ipv4.h>
#include <cpu/cpu.h>
lock_t udp_port_table_lock;
socket_resource_t* udp_port_table[65536] = {0};
uint16_t get_next_ephemeral_port()
{
uint16_t start = EPHEMERAL_PORT_MIN + (cpu_rdrand32() % (EPHEMERAL_PORT_MAX - EPHEMERAL_PORT_MIN + 1));
uint16_t range = EPHEMERAL_PORT_MAX - EPHEMERAL_PORT_MIN + 1;
for(uint16_t i = 0; i < range; i++)
{
uint16_t port = EPHEMERAL_PORT_MIN + ((start - EPHEMERAL_PORT_MIN + i) % range);
if (udp_port_table[port] == NULL) return port;
}
return 0;
}
void socket_init()
{
// function body left deliberately empty
}
uint64_t sys_socket(uint64_t domain, uint64_t type, uint64_t protocol,
uint64_t, uint64_t, uint64_t)
{
if(domain != AF_IPV4)
{
return -EAFNOSUPPORT;
}
if(type != SOCK_DGRAM)
{
return -EINVAL;
}
if(protocol != PROTO_UDP)
{
return -EINVAL;
}
process_t* process = get_current_thread()->process;
lock_acquire(&process->fds_lock);
int cur_fd = 0;
while(process->fds[cur_fd] != NULL && cur_fd < PROC_MAX_FDS) cur_fd++;
if (cur_fd == PROC_MAX_FDS)
{
lock_release(&process->fds_lock);
return -EMFILE;
}
socket_resource_t* sockresource = malloc(sizeof(socket_resource_t));
memset(sockresource, 0, sizeof(socket_resource_t));
// function pointers
sockresource->resource.grow = socket_grow ;
sockresource->resource.read = socket_read ;
sockresource->resource.write = socket_write ;
sockresource->resource.ioctl = socket_ioctl ;
sockresource->resource.unref = socket_unref ;
sockresource->resource.link = socket_link ;
sockresource->resource.unlink = socket_unlink;
sockresource->resource.mmap = socket_mmap ;
sockresource->domain = domain;
sockresource->type = type;
sockresource->protocol = protocol;
sockresource->state = SOCK_STATE_CLOSED;
sockresource->resource.stat.mode = STAT_IFSOCK;
udp_pcb_t* pcb = malloc(sizeof(udp_pcb_t));
memset(pcb, 0, sizeof(udp_pcb_t));
sem_init(&pcb->sem, 0);
sockresource->pcb = pcb;
// set up fd
vfs_node_t* sock_node = malloc(sizeof(vfs_node_t));
*sock_node = (vfs_node_t){
.mountpoint = NULL,
.redir = NULL,
.resource = (resource_t*)sockresource,
.filesystem = NULL,
.name = "", // TODO: what name for socket?
.parent = NULL,
.children = NULL,
// -1 means "not a directory", 0 is a directory with 0 children
.children_count = -1,
.symlink_target = NULL,
};
file_handle_t* sock_handle = malloc(sizeof(file_handle_t));
*sock_handle = (file_handle_t){
.resource = sock_node->resource,
.node = sock_node,
.refcount = 1
};
file_descriptor_t* sock_fd = malloc(sizeof(file_descriptor_t));
*sock_fd = (file_descriptor_t){
.handle = sock_handle,
};
process->fds[cur_fd] = sock_fd;
lock_release(&process->fds_lock);
// cast pointer to uint64_t for syscall signature match
return (uint64_t)cur_fd;
}
uint64_t sys_bind(uint64_t fd, uint64_t addr, uint64_t, uint64_t, uint64_t, uint64_t)
{
process_t* process = get_current_thread()->process;
if(fd >= PROC_MAX_FDS)
{
return -EINVAL;
}
file_descriptor_t* sockfd = process->fds[(int)fd];
if(sockfd == NULL)
{
return -EBADF;
}
if((sockfd->handle->resource->stat.mode & STAT_IFMT) != STAT_IFSOCK)
{
return -ENOTSOCK;
}
socket_resource_t* sockresource = (socket_resource_t*)sockfd->handle->resource;
if(sockresource->state == SOCK_STATE_BOUND)
{
return -EINVAL;
}
lock_acquire(&udp_port_table_lock);
uint16_t port = ntohs(((net_addr_t*)addr)->port);
if(udp_port_table[port] != NULL)
{
lock_release(&udp_port_table_lock);
return -EADDRINUSE;
}
sockresource->local_addr = *((net_addr_t*)addr);
udp_port_table[port] = sockresource;
sockresource->state = SOCK_STATE_BOUND;
lock_release(&udp_port_table_lock);
return 0;
}
uint64_t sys_sendto(uint64_t fd, uint64_t buf, uint64_t len,
uint64_t dest, uint64_t, uint64_t)
{
process_t* process = get_current_thread()->process;
if(fd >= PROC_MAX_FDS)
{
return -EINVAL;
}
file_descriptor_t* sockfd = process->fds[(int)fd];
if(sockfd == NULL)
{
return -EBADF;
}
if((sockfd->handle->resource->stat.mode & STAT_IFMT) != STAT_IFSOCK)
{
return -ENOTSOCK;
}
net_addr_t* dest_addr = (net_addr_t*)dest;
uint8_t* payload = (uint8_t*)buf;
// ethernet, ipv4, UDP headers, plus payload
size_t total_len = ETH_HEADER_LEN + IPV4_HEADER_LEN + UDP_HEADER_LEN + len;
route_result_t route;
if(!route_lookup(ntohl(dest_addr->ipv4), &route))
{
// TODO: error handling
return -ENOSYS;
}
uint8_t next_hop_bytes[4];
ip4_to_bytes(route.next_hop, next_hop_bytes);
uint8_t dest_mac[6];
if(!arp_lookup(route.device, next_hop_bytes, dest_mac))
{
// TODO: queue packet while arp request finishes
return -ENOSYS;
}
uint8_t* packet = malloc(total_len);
uint8_t dest_addr_bytes[4];
ip4_to_bytes(ntohl(dest_addr->ipv4), dest_addr_bytes);
eth_build_header(&packet[0], dest_mac, route.device->mac, ETHERTYPE_IPV4);
ipv4_build_header(&packet[ETH_HEADER_LEN], route.device->ip4, dest_addr_bytes,
IP_PROTO_UDP, UDP_HEADER_LEN + len, DEFAULT_TTL);
uint16_t source_port = ntohs(((socket_resource_t*)sockfd->handle->resource)->local_addr.port);
if (source_port == 0)
{
uint16_t next_ephemeral = get_next_ephemeral_port();
if(next_ephemeral == 0)
{
free(packet);
return -EADDRINUSE;
}
source_port = next_ephemeral;
socket_resource_t* sockresource = (socket_resource_t*)sockfd->handle->resource;
sockresource->local_addr.port = htons(source_port);
sockresource->state = SOCK_STATE_BOUND;
udp_port_table[source_port] = sockresource;
}
udp_build_header(&packet[ETH_HEADER_LEN+IPV4_HEADER_LEN], source_port, ntohs(dest_addr->port), len);
memcpy(&packet[ETH_HEADER_LEN+IPV4_HEADER_LEN+UDP_HEADER_LEN], payload, len);
// finally we have the complete packet!
route.device->transmit(packet, total_len);
free(packet);
return len;
}
uint64_t sys_recvfrom(uint64_t fd, uint64_t buf, uint64_t len,
uint64_t src, uint64_t, uint64_t)
{
(void)fd; (void)buf; (void)len; (void)src;
return (uint64_t)-ENOSYS;
}
uint64_t sys_close_socket(uint64_t fd, uint64_t, uint64_t, uint64_t, uint64_t, uint64_t)
{
(void)fd;
return (uint64_t)-ENOSYS;
}
bool socket_grow(resource_t* self, void* handle, uint64_t size)
{
return false; // sockets can't grow
}
int64_t socket_read(resource_t* self, void* handle, void* buf, uint64_t loc, uint64_t count)
{
// TODO: implement me
return -ENOSYS;
}
int64_t socket_write(resource_t* self, void* handle, void* buf, uint64_t loc, uint64_t count)
{
// TODO: implement me
return -ENOSYS;
}
int socket_ioctl(resource_t* self, void* handle, uint64_t request, void* argp)
{
// TODO: implement me
return -ENOSYS;
}
bool socket_unref(resource_t* self, void* handle)
{
// TODO: close the socket? I don't know if this call even makes sense
return false;
}
bool socket_link(resource_t* self, void* handle)
{
// TODO: AF_UNIX will need an implementation here probably
return false;
}
bool socket_unlink(resource_t* self, void* handle)
{
// TODO: AF_UNIX will need an implementation here probably
return false;
}
void* socket_mmap(resource_t* self, uint64_t page, int flags)
{
// cannot mmap a socket
return NULL;
}