[GDB] Remote debugging over UDP

This commit implements GDB over UDP.  Using UDP is more complex than
serial and has required some restructuring.

The GDB stub is now built using one or both of GDBSERIAL and GDBUDP
config.h options.

To enter the debugger, execute the gPXE shell command:
gdbstub <transport> [<options>...]

Where <transport> is "serial" or "udp".  For "udp", the name of a
configured network device is required:
gdbstub udp net0

The GDB stub listens on UDP port 43770 by default.

1 files changed, 228 insertions, 0 deletions

diff --git a/src/core/gdbudp.c b/src/core/gdbudp.c
new file mode 100644
index 00000000..fb381644
--- /dev/null
+++ b/src/core/gdbudp.c
@@ -0,0 +1,228 @@
+/*
+ * Copyright (C) 2008 Stefan Hajnoczi <stefanha@gmail.com>.
+ *
+ * This program 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 2 of the
+ * License, or any later version.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#include <stdio.h>
+#include <string.h>
+#include <byteswap.h>
+#include <gpxe/iobuf.h>
+#include <gpxe/in.h>
+#include <gpxe/if_arp.h>
+#include <gpxe/if_ether.h>
+#include <gpxe/ip.h>
+#include <gpxe/udp.h>
+#include <gpxe/netdevice.h>
+#include <gpxe/gdbstub.h>
+
+struct gdb_transport udp_gdb_transport __gdb_transport;
+
+static struct net_device *netdev;
+static uint8_t dest_eth[ETH_ALEN];
+static uint8_t source_eth[ETH_ALEN];
+static struct sockaddr_in dest_addr;
+static struct sockaddr_in source_addr;
+
+static void gdbudp_ensure_netdev_open ( struct net_device *netdev ) {
+	if ( ( netdev->state & NETDEV_OPEN) == 0 ) {
+		netdev_open ( netdev );
+	}
+	/* TODO forcing the netdev to be open is useful when
+	 * gPXE closes the netdev between breakpoints.  Should
+	 * we restore the state of the netdev, i.e. closed,
+	 * before leaving the interrupt handler? */
+}
+
+static size_t gdbudp_recv ( char *buf, size_t len ) {
+	struct io_buffer *iob;
+	struct ethhdr *ethhdr;
+	struct arphdr *arphdr;
+	struct iphdr *iphdr;
+	struct udp_header *udphdr;
+	size_t payload_len;
+
+	assert ( netdev );
+	gdbudp_ensure_netdev_open ( netdev );
+
+	for ( ; ; ) {
+		while ( ( iob = netdev_rx_dequeue ( netdev ) ) != NULL ) {
+			if ( iob_len ( iob ) > sizeof ( *ethhdr ) + sizeof ( *iphdr ) + sizeof ( *udphdr ) + len ) {
+				goto bad_packet;
+			}
+
+			/* Ethernet header */
+			ethhdr = iob->data;
+			iob_pull ( iob, sizeof ( *ethhdr ) );
+			if ( ethhdr->h_protocol == htons ( ETH_P_ARP ) ) {
+				/* Handle ARP requests so the client can connect to us */
+				arphdr = iob->data;
+				if ( arphdr->ar_hrd != htons ( ARPHRD_ETHER ) ||
+						arphdr->ar_pro != htons ( ETH_P_IP ) ||
+						arphdr->ar_hln != ETH_ALEN ||
+						arphdr->ar_pln != sizeof ( struct in_addr ) ||
+						arphdr->ar_op != htons ( ARPOP_REQUEST ) ||
+						memcmp ( arp_target_pa ( arphdr ), &source_addr.sin_addr.s_addr, sizeof ( struct in_addr ) ) ) {
+					goto bad_packet;
+				}
+
+				/* Generate an ARP reply */
+				arphdr->ar_op = htons ( ARPOP_REPLY );
+				memswap ( arp_sender_pa ( arphdr ), arp_target_pa ( arphdr ), sizeof ( struct in_addr ) );
+				memcpy ( arp_target_ha ( arphdr ), arp_sender_ha ( arphdr ), ETH_ALEN );
+				memcpy ( arp_sender_ha ( arphdr ), source_eth, ETH_ALEN );
+
+				/* Fix up ethernet header */
+				ethhdr = iob_push ( iob, sizeof ( *ethhdr ) );
+				memswap ( ethhdr->h_source, ethhdr->h_dest, ETH_ALEN );
+
+				netdev_tx ( netdev, iob );
+				continue; /* no need to free iob */
+			}
+			if ( ethhdr->h_protocol != htons ( ETH_P_IP ) ) {
+				goto bad_packet;
+			}
+
+			/* IP header */
+			iphdr = iob->data;
+			iob_pull ( iob, sizeof ( *iphdr ) );
+			if ( iphdr->protocol != IP_UDP || iphdr->dest.s_addr != source_addr.sin_addr.s_addr ) {
+				goto bad_packet;
+			}
+
+			/* UDP header */
+			udphdr = iob->data;
+			if ( udphdr->dest != source_addr.sin_port ) {
+				goto bad_packet;
+			}
+
+			/* Learn the remote connection details */
+			memcpy ( dest_eth, ethhdr->h_source, ETH_ALEN );
+			dest_addr.sin_addr.s_addr = iphdr->src.s_addr;
+			dest_addr.sin_port = udphdr->src;
+
+			/* Payload */
+			payload_len = ntohs ( udphdr->len );
+			if ( payload_len < sizeof ( *udphdr ) ||
+					payload_len > iob_len ( iob ) ) {
+				goto bad_packet;
+			}
+			payload_len -= sizeof ( *udphdr );
+			iob_pull ( iob, sizeof ( *udphdr ) );
+			memcpy ( buf, iob->data, payload_len );
+
+			free_iob ( iob );
+			return payload_len;
+
+bad_packet:
+			free_iob ( iob );
+		}
+		netdev_poll ( netdev );
+	}
+}
+
+static void gdbudp_send ( const char *buf, size_t len ) {
+	struct io_buffer *iob;
+	struct ethhdr *ethhdr;
+	struct iphdr *iphdr;
+	struct udp_header *udphdr;
+
+	/* Check that we are connected */
+	if ( dest_addr.sin_port == 0 ) {
+		return;
+	}
+
+	assert ( netdev );
+	gdbudp_ensure_netdev_open ( netdev );
+
+	iob = alloc_iob ( sizeof ( *ethhdr ) + sizeof ( *iphdr ) + sizeof ( *udphdr ) + len );
+	if ( !iob ) {
+		return;
+	}
+
+	/* Payload */
+	iob_reserve ( iob, sizeof ( *ethhdr ) + sizeof ( *iphdr ) + sizeof ( *udphdr ) );
+	memcpy ( iob_put ( iob, len ), buf, len );
+
+	/* UDP header */
+	udphdr = iob_push ( iob, sizeof ( *udphdr ) );
+	udphdr->src = source_addr.sin_port;
+	udphdr->dest = dest_addr.sin_port;
+	udphdr->len = htons ( iob_len ( iob ) );
+	udphdr->chksum = 0; /* optional and we are not using it */
+
+	/* IP header */
+	iphdr = iob_push ( iob, sizeof ( *iphdr ) );
+	memset ( iphdr, 0, sizeof ( *iphdr ) );
+	iphdr->verhdrlen = ( IP_VER | ( sizeof ( *iphdr ) / 4 ) );
+	iphdr->service = IP_TOS;
+	iphdr->len = htons ( iob_len ( iob ) );	
+	iphdr->ttl = IP_TTL;
+	iphdr->protocol = IP_UDP;
+	iphdr->dest.s_addr = dest_addr.sin_addr.s_addr;
+	iphdr->src.s_addr = source_addr.sin_addr.s_addr;
+	iphdr->chksum = tcpip_chksum ( iphdr, sizeof ( *iphdr ) );
+
+	/* Ethernet header */
+	ethhdr = iob_push ( iob, sizeof ( *ethhdr ) );
+	memcpy ( ethhdr->h_dest, dest_eth, ETH_ALEN );
+	memcpy ( ethhdr->h_source, source_eth, ETH_ALEN );
+	ethhdr->h_protocol = htons ( ETH_P_IP );
+
+	netdev_tx ( netdev, iob );
+}
+
+static int gdbudp_init ( int argc, char **argv ) {
+	struct settings *settings;
+
+	if ( argc != 1 ) {
+		printf ( "udp: missing <interface> argument\n" );
+		return 1;
+	}
+
+	netdev = find_netdev ( argv[0] );
+	if ( !netdev ) {
+		printf ( "%s: no such interface\n", argv[0] );
+		return 1;
+	}
+
+	if ( !netdev_link_ok ( netdev ) ) {
+		printf ( "%s: link not up\n", argv[0] );
+		return 1;
+	}
+
+	/* Load network settings from device.  We keep the MAC address,
+	 * IP address, and UDP port.  The MAC and IP could be fetched
+	 * from the network device each time they are used in rx/tx.
+	 * Storing a separate copy makes it possible to use different
+	 * MAC/IP settings than the network stack. */
+	memcpy ( source_eth, netdev->ll_addr, ETH_ALEN );
+	source_addr.sin_port = htons ( 43770 ); /* TODO default port */
+	settings = netdev_settings ( netdev );
+	fetch_ipv4_setting ( settings, &ip_setting, &source_addr.sin_addr );
+	if ( source_addr.sin_addr.s_addr == 0 ) {
+		printf ( "%s: no IP address configured\n", argv[0] );
+		return 1;
+	}
+
+	return 0;
+}
+
+struct gdb_transport udp_gdb_transport __gdb_transport = {
+	.name = "udp",
+	.init = gdbudp_init,
+	.send = gdbudp_send,
+	.recv = gdbudp_recv,
+};