* kgdb 2.0.5
@ 2004-01-20 12:13 Amit S. Kale
2004-01-20 13:37 ` Pavel Machek
2004-01-20 20:35 ` Pavel Machek
0 siblings, 2 replies; 4+ messages in thread
From: Amit S. Kale @ 2004-01-20 12:13 UTC (permalink / raw)
To: Linux Kernel, Pavel Machek, George Anzinger
Cc: Steve Gonczi, Matt Mackall, Jeff Garzik
Hi,
kgdb 2.0.5 is available at
http://kgdb.sourceforge.net/kgdb-2/linux-2.6.1-kgdb-2.0.5.tar.bz2
ChangeLog
2004-01-20 Amit S. Kale <amitkale@emsyssoft.com>
* Created a ring buffer for kgdb ethernet packets. Several
fixes and changes to kgdb on ethernet.
2004-01-20 TimeSys Corporation
* Fixed a problem with not responding to Ctrl+C during priting of
console messages through gdb.
I have pasted below eth.patch for review. When using the ethernet interface,
gdb times out several times. It receives packets and junk instead of acks. I
see following type of messages out of 8139too.c on the console
"eth0:Out-of-sync dirty pointer, 15 vs. 20."
Any comments/suggestions/fixes on this patch are most welcome.
Thanks.
--
Amit Kale
EmSysSoft (http://www.emsyssoft.com)
KGDB: Linux Kernel Source Level Debugger (http://kgdb.sourceforge.net)
diff -Naurp linux-2.6.1/drivers/net/Kconfig
linux-2.6.1-kgdb-2.0.5-eth/drivers/net/Kconfig
--- linux-2.6.1/drivers/net/Kconfig 2003-11-24 07:02:50.000000000 +0530
+++ linux-2.6.1-kgdb-2.0.5-eth/drivers/net/Kconfig 2004-01-12
19:11:13.000000000 +0530
@@ -187,6 +187,12 @@ config NET_ETHERNET
Note that the answer to this question won't directly affect the
kernel: saying N will just cause the configurator to skip all
the questions about Ethernet network cards. If unsure, say N.
+
+config KGDB_ETH
+ bool "KGDB: On ethernet"
+ depends on KGDB
+ help
+ Uses ethernet interface for kgdb.
config MII
tristate "Generic Media Independent Interface device support"
diff -Naurp linux-2.6.1/drivers/net/kgdb_eth.c
linux-2.6.1-kgdb-2.0.5-eth/drivers/net/kgdb_eth.c
--- linux-2.6.1/drivers/net/kgdb_eth.c 1970-01-01 05:30:00.000000000 +0530
+++ linux-2.6.1-kgdb-2.0.5-eth/drivers/net/kgdb_eth.c 2004-01-20
16:14:22.000000000 +0530
@@ -0,0 +1,704 @@
+/*
+ * Network interface GDB stub
+ *
+ * Copyright (C), 2004 Amit S. Kale
+ *
+ * Written by San Mehat (nettwerk@biodome.org)
+ * Based upon 'gdbserial' by David Grothe (dave@gcom.com)
+ * and Scott Foehner (sfoehner@engr.sgi.com)
+ *
+ * Twiddled for 2.6 by Robert Walsh <rjwalsh@durables.org>
+ * and wangdi <wangdi@clusterfs.com>.
+ *
+ * Restructured for generic a gdb interface
+ * Reveral changes to make it free of device driver changes.
+ * Added internal buffers for this interface.
+ * by Amit S. Kale <amitkale@emsyssoft.com>
+ * Some cleanups by Pavel Machek <pavel@suse.cz>
+ */
+
+#include <linux/module.h>
+#include <linux/errno.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/timer.h>
+#include <linux/interrupt.h>
+#include <linux/config.h>
+#include <linux/major.h>
+#include <linux/string.h>
+#include <linux/fcntl.h>
+#include <linux/termios.h>
+#include <linux/kgdb.h>
+#include <linux/if_ether.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+#include <linux/delay.h>
+#include <linux/irq.h>
+#include <linux/inet.h>
+#include <linux/notifier.h>
+#include <net/tcp.h>
+#include <net/udp.h>
+
+#include <asm/system.h>
+#include <asm/io.h>
+#include <asm/segment.h>
+#include <asm/bitops.h>
+#include <asm/system.h>
+#include <asm/atomic.h>
+
+#define GDB_BUF_SIZE 512 /* power of 2, please */
+
+#define CHUNKSIZE 30
+#define MAXINCHUNK (CHUNKSIZE + 8)
+
+static char kgdb_buf[GDB_BUF_SIZE];
+static int kgdb_buf_in_inx;
+static atomic_t kgdb_buf_in_cnt;
+static int kgdb_buf_out_inx;
+
+static unsigned int kgdb_remoteip = 0;
+static unsigned short kgdb_listenport = 6443;
+static unsigned short kgdb_sendport= 6442;
+static int kgdb_eth = -1; /* Default tty mode */
+static unsigned char kgdb_remotemac[6] = {0xff,0xff,0xff,0xff,0xff,0xff};
+static unsigned char kgdb_localmac[6] = {0xff,0xff,0xff,0xff,0xff,0xff};
+
+static char kgdbeth_sendbuf[MAXINCHUNK];
+static int kgdbeth_sendbufchars;
+static irqreturn_t (*kgdbeth_irqhandler)(int, void *, struct pt_regs *) =
NULL;
+
+struct net_device *kgdb_netdevice = NULL;
+
+/* Indicates dept of recursion for xmitlock hold */
+static int xlockholdcount = 0;
+
+/* kgdb ethernet ring buffers. Increase the space if you get panics in
+ * kgdbeth_alloc_skb.
+ * Status of send_skbs can be known from the field users.
+ * If it's 1 the buffer is free.
+ * If the count 2 or more, the buffer is in use.
+ * Keeping 1 as the initial count prevents kfree_skb from freeing it. */
+#define SEND_BUFLEN 1024
+#define NUM_SENDBUF 128
+static char *send_bufs[NUM_SENDBUF];
+static struct sk_buff *send_skbs[NUM_SENDBUF];
+static struct sk_buff *send_skb;
+static int bufnum;
+
+static char kgdb_netdevname[16];
+
+#ifdef CONFIG_KGDB_CONSOLE
+#error kgdb over ethernet is not yet ready for console messages.
+#endif
+
+/*
+ * Returns next skb from kgdb skbs.
+ * Initializes users field of the skb to 2 so that kfree_skb doesn't attempt
+ * freeing it.
+ * Always call after holding xmitlock of the ethernet device.
+ */
+
+struct sk_buff *kgdbeth_alloc_skb(int size)
+{
+ struct sk_buff *skb;
+ u8 *data;
+ int i;
+
+ i = bufnum;
+
+ do {
+ skb = send_skbs[i];
+
+ if (atomic_read(&skb->users) == 1) {
+ bufnum = i;
+ i = -1;
+ break;
+ }
+ i = (i + 1) % NUM_SENDBUF;
+ } while (i != bufnum);
+ if (i == bufnum) {
+ panic("kgdb ethernet buffer overflow\n");
+ }
+ data = (u8 *)(send_bufs[bufnum]);
+
+ size = SKB_DATA_ALIGN(size);
+ if (size + sizeof(struct skb_shared_info) > SEND_BUFLEN)
+ panic("kgdb ethernet buffer too short for this request");
+
+ memset(skb, 0, offsetof(struct sk_buff, truesize));
+ skb->truesize = size + sizeof(struct sk_buff);
+ atomic_set(&skb->users, 2);
+ skb->head = data;
+ skb->data = data;
+ skb->tail = data;
+ skb->end = data + size;
+
+ atomic_set(&(skb_shinfo(skb)->dataref), 1);
+ skb_shinfo(skb)->nr_frags = 0;
+ skb_shinfo(skb)->tso_size = 0;
+ skb_shinfo(skb)->tso_segs = 0;
+ skb_shinfo(skb)->frag_list = NULL;
+
+ return skb;
+}
+
+/* Holds xmitlock of the ethernet device
+ * Recursive calls allowed */
+static void kgdbeth_holdxlock(void)
+{
+ if (spin_is_locked(&kgdb_netdevice->xmit_lock)) {
+ if (kgdb_netdevice->xmit_lock_owner == smp_processor_id()) {
+ goto gotit;
+ }
+ }
+ spin_lock(&kgdb_netdevice->xmit_lock);
+ kgdb_netdevice->xmit_lock_owner = smp_processor_id();
+
+gotit:
+ xlockholdcount++;
+}
+
+/* releases xmitlock of the ethernet device
+ * Recursive calls allowed */
+static void kgdbeth_relxlock(void)
+{
+ if (--xlockholdcount) {
+ kgdb_netdevice->xmit_lock_owner = -1;
+ spin_unlock(&kgdb_netdevice->xmit_lock);
+ }
+}
+
+/*
+ * Get a char if available, return -1 if nothing available.
+ * Empty the receive buffer first, then look at the interface hardware.
+ */
+static int
+read_char(void)
+{
+ /* intr routine has queued chars */
+ if (atomic_read(&kgdb_buf_in_cnt) != 0) {
+ int chr;
+
+ chr = kgdb_buf[kgdb_buf_out_inx++];
+ kgdb_buf_out_inx &= (GDB_BUF_SIZE - 1);
+ atomic_dec(&kgdb_buf_in_cnt);
+ return chr;
+ }
+
+ return -1; /* no data */
+}
+
+/*
+ * Wait until the interface can accept a char, then write it.
+ */
+static void
+write_buffer(char *buf, int len)
+{
+ int total_len, eth_len, ip_len, udp_len;
+ struct in_device *in_dev;
+ struct sk_buff *skb;
+ struct udphdr *udph;
+ struct iphdr *iph;
+ struct ethhdr *eth;
+
+ if (!(in_dev = (struct in_device *) kgdb_netdevice->ip_ptr)) {
+ panic("No in_device available for interface!\n");
+ }
+
+ if (!(in_dev->ifa_list)) {
+ panic("No interface address set for interface!\n");
+ }
+ kgdbeth_holdxlock();
+
+ udp_len = len + sizeof(struct udphdr);
+ ip_len = eth_len = udp_len + sizeof(struct iphdr);
+ total_len = eth_len + ETH_HLEN;
+
+ skb = kgdbeth_alloc_skb(total_len);
+
+ skb_reserve(skb, total_len - 1);
+
+ memcpy(skb->data, (unsigned char *) buf, len);
+ skb->len += len;
+
+ udph = (struct udphdr *) skb_push(skb, sizeof(*udph));
+ udph->source = htons(kgdb_listenport);
+ udph->dest = htons(kgdb_sendport);
+ udph->len = htons(udp_len);
+ udph->check = 0;
+
+ iph = (struct iphdr *)skb_push(skb, sizeof(*iph));
+ iph->version = 4;
+ iph->ihl = 5;
+ iph->tos = 0;
+ iph->tot_len = htons(ip_len);
+ iph->id = 0;
+ iph->frag_off = 0;
+ iph->ttl = 64;
+ iph->protocol = IPPROTO_UDP;
+ iph->check = 0;
+ iph->saddr = in_dev->ifa_list->ifa_address;
+ iph->daddr = kgdb_remoteip;
+ iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
+
+ eth = (struct ethhdr *) skb_push(skb, ETH_HLEN);
+ eth->h_proto = htons(ETH_P_IP);
+ memcpy(eth->h_source, kgdb_localmac, kgdb_netdevice->addr_len);
+ memcpy(eth->h_dest, kgdb_remotemac, kgdb_netdevice->addr_len);
+
+ kgdb_netdevice->hard_start_xmit(skb, kgdb_netdevice);
+ if (atomic_read(&skb->users) != 1) {
+ BUG();
+ }
+ kgdbeth_relxlock();
+}
+
+static void kgdbeth_flush(void)
+{
+ if (!kgdbeth_sendbufchars) {
+ return;
+ }
+ write_buffer(kgdbeth_sendbuf, kgdbeth_sendbufchars);
+ kgdbeth_sendbufchars = 0;
+}
+
+static void kgdbeth_write_char(int chr)
+{
+ if (kgdbeth_sendbufchars == MAXINCHUNK) {
+ kgdbeth_flush();
+ }
+ kgdbeth_sendbuf[kgdbeth_sendbufchars++] = chr;
+}
+
+
+/*
+ * In the interrupt state the target machine will not respond to any
+ * arp requests, so handle them here.
+ */
+
+static void
+kgdb_eth_reply_arp(void)
+{
+ if (send_skb) {
+ kgdbeth_holdxlock();
+ kgdb_netdevice->hard_start_xmit(send_skb, kgdb_netdevice);
+ send_skb = NULL;
+ kgdbeth_relxlock();
+ }
+}
+
+static int
+make_arp_request(struct sk_buff *skb)
+{
+ struct arphdr *arp;
+ unsigned char *arp_ptr;
+ int type = ARPOP_REPLY;
+ int ptype = ETH_P_ARP;
+ u32 sip, tip;
+ unsigned char *sha, *tha;
+ struct in_device *in_dev = (struct in_device *) kgdb_netdevice->ip_ptr;
+
+ /* No arp on this interface */
+
+ if (kgdb_netdevice->flags & IFF_NOARP) {
+ return 0;
+ }
+
+ if (!pskb_may_pull(skb, (sizeof(struct arphdr) +
+ (2 * kgdb_netdevice->addr_len) +
+ (2 * sizeof(u32))))) {
+ return 0;
+ }
+
+ skb->h.raw = skb->nh.raw = skb->data;
+ arp = skb->nh.arph;
+
+ if ((arp->ar_hrd != htons(ARPHRD_ETHER) &&
+ arp->ar_hrd != htons(ARPHRD_IEEE802)) ||
+ arp->ar_pro != htons(ETH_P_IP)) {
+ return 0;
+ }
+
+ /* Understand only these message types */
+
+ if (arp->ar_op != htons(ARPOP_REQUEST)) {
+ return 0;
+ }
+
+ /* Extract fields */
+
+ arp_ptr= (unsigned char *)(arp+1);
+ sha = arp_ptr;
+ arp_ptr += kgdb_netdevice->addr_len;
+ memcpy(&sip, arp_ptr, 4);
+ arp_ptr += 4;
+ tha = arp_ptr;
+ arp_ptr += kgdb_netdevice->addr_len;
+ memcpy(&tip, arp_ptr, 4);
+
+ if (tip != in_dev->ifa_list->ifa_address) {
+ return 0;
+ }
+
+ if (kgdb_remoteip != sip) {
+ return 0;
+ }
+
+ /*
+ * Check for bad requests for 127.x.x.x and requests for multicast
+ * addresses. If this is one such, delete it.
+ */
+
+ if (LOOPBACK(tip) || MULTICAST(tip)) {
+ return 0;
+ }
+ /* reply to the ARP request */
+
+ kgdbeth_holdxlock();
+ if (send_skb) {
+ /* Get rid of any previous replies to ARP request. We hope
+ * that regular reply to ARP by network layers would have gone
+ * out by now. */
+ kfree_skb(send_skb);
+ }
+ send_skb = kgdbeth_alloc_skb(sizeof(struct arphdr) +
+ 2 * (kgdb_netdevice->addr_len + 4) +
+ LL_RESERVED_SPACE(kgdb_netdevice));
+ kgdbeth_relxlock();
+
+ skb_reserve(send_skb, LL_RESERVED_SPACE(kgdb_netdevice));
+ send_skb->nh.raw = send_skb->data;
+ arp = (struct arphdr *) skb_put(send_skb, sizeof(struct arphdr) + 2 *
(kgdb_netdevice->addr_len + 4));
+ send_skb->dev = kgdb_netdevice;
+ send_skb->protocol = htons(ETH_P_ARP);
+
+ /* Fill the device header for the ARP frame */
+
+ if (kgdb_netdevice->hard_header &&
+ kgdb_netdevice->hard_header(send_skb, kgdb_netdevice, ptype,
+ kgdb_remotemac, kgdb_localmac,
+ send_skb->len) < 0) {
+ kfree_skb(send_skb);
+ return 0;
+ }
+
+ /*
+ * Fill out the arp protocol part.
+ *
+ * we only support ethernet device type,
+ * which (according to RFC 1390) should always equal 1 (Ethernet).
+ */
+
+ arp->ar_hrd = htons(kgdb_netdevice->type);
+ arp->ar_pro = htons(ETH_P_IP);
+
+ arp->ar_hln = kgdb_netdevice->addr_len;
+ arp->ar_pln = 4;
+ arp->ar_op = htons(type);
+
+ arp_ptr=(unsigned char *)(arp + 1);
+
+ memcpy(arp_ptr, kgdb_netdevice->dev_addr, kgdb_netdevice->addr_len);
+ arp_ptr += kgdb_netdevice->addr_len;
+ memcpy(arp_ptr, &tip, 4);
+ arp_ptr += 4;
+ memcpy(arp_ptr, kgdb_localmac, kgdb_netdevice->addr_len);
+ arp_ptr += kgdb_netdevice->addr_len;
+ memcpy(arp_ptr, &sip, 4);
+ return 0;
+}
+
+
+/*
+ * Accept an skbuff from net_device layer and add the payload onto
+ * kgdb buffer
+ *
+ * When the kgdb stub routine read_char() is called it draws characters
+ * out of the buffer until it is empty and then polls the hardware.
+ *
+ * Return value of NET_RX_DROP means skb was used by this function.
+ * NET_RX_SUCCESS indicates this function didn't use it.
+
+ * Be prepared to respond to ARP requests. Let the caller also handle
+ * them if debugger is not active. We'll respond to the ARP request when
+ * a debugging session begins. It's necessary to respond to the request as
the
+ * debugging session may begin even before kernel has a chance finish the
+ * response.
+ */
+int
+kgdb_net_interrupt(struct sk_buff *skb)
+{
+ unsigned char chr;
+ struct iphdr *iph = (struct iphdr*)skb->data;
+ struct udphdr *udph= (struct udphdr*)(skb->data+(iph->ihl<<2));
+ unsigned char *data = (unsigned char *) udph + sizeof(struct udphdr);
+ int len;
+ int i;
+
+ if (!kgdb_initialized || !kgdb_netdevice) {
+ goto out;
+ }
+ if (skb->protocol == __constant_htons(ETH_P_ARP) && !send_skb) {
+ make_arp_request(skb);
+ goto out;
+ }
+ if (iph->protocol != IPPROTO_UDP ||
+ be16_to_cpu(udph->dest) != kgdb_listenport)
+ goto out;
+
+ kgdb_sendport = be16_to_cpu(udph->source);
+
+ len = (be16_to_cpu(iph->tot_len) -
+ (sizeof(struct udphdr) + sizeof(struct iphdr)));
+
+ for (i = 0; i < len; i++) {
+ chr = *data++;
+ if (chr == 3)
+ {
+ if (!atomic_read(&debugger_active)) {
+ breakpoint();
+ }
+ continue;
+ }
+ if (atomic_read(&kgdb_buf_in_cnt) >= GDB_BUF_SIZE) {
+ /* buffer overflow, clear it */
+ kgdb_buf_in_inx = 0;
+ atomic_set(&kgdb_buf_in_cnt, 0);
+ kgdb_buf_out_inx = 0;
+ break;
+ }
+ kgdb_buf[kgdb_buf_in_inx++] = chr;
+ kgdb_buf_in_inx &= (GDB_BUF_SIZE - 1);
+ atomic_inc(&kgdb_buf_in_cnt) ;
+ }
+ return NET_RX_DROP;
+
+out:
+ if (atomic_read(&debugger_active))
+ return NET_RX_DROP;
+ return NET_RX_SUCCESS;
+}
+
+/*
+ * Initializes ethernet interface to kgdb.
+ * Searches the device list and finds the device specified on kernel command
+ * line.
+ * Finds the irq handler for the device and saves its reference.
+ * Initializes kgdbeth data structures.
+ */
+
+int
+kgdbeth_hook(void)
+{
+ extern void kgdb_respond_ok(void);
+ struct irqaction *ia_ptr;
+ int i;
+
+ sprintf(kgdb_netdevname, "eth%d", kgdb_eth);
+
+ for (kgdb_netdevice = dev_base;
+ kgdb_netdevice != NULL;
+ kgdb_netdevice = kgdb_netdevice->next) {
+ if (strncmp(kgdb_netdevice->name, kgdb_netdevname, IFNAMSIZ) == 0) {
+ break;
+ }
+ }
+ if (!kgdb_netdevice) {
+ printk("kgdbeth: Unable to find interface %s\n",
+ kgdb_netdevname);
+ return -ENODEV;
+ }
+ if (!(kgdb_netdevice->flags & IFF_UP)) {
+ return -EINVAL;
+ }
+ ia_ptr = irq_desc[kgdb_netdevice->irq].action;
+ while (ia_ptr) {
+ if (!strncmp(kgdb_netdevname, ia_ptr->name, IFNAMSIZ)) {
+ kgdbeth_irqhandler = ia_ptr->handler;
+ break;
+ }
+ ia_ptr = ia_ptr->next;
+ }
+ if (!kgdbeth_irqhandler) {
+ printk("kgdbeth: Interface %s doesn't have an interrupt"
+ " handler cannot use it\n", kgdb_netdevname);
+ return -EINVAL;
+ }
+ for (i = 0; i < NUM_SENDBUF; i++) {
+ send_skbs[i] = kmalloc(sizeof (struct sk_buff), GFP_KERNEL);
+ send_bufs[i] = kmalloc(SEND_BUFLEN, GFP_KERNEL);
+ if (!send_skbs[i] || !send_bufs[i]) {
+ printk("kgdbeth: not enough memory\n");
+ return -ENOMEM;
+ }
+ atomic_set(&(send_skbs[i]->users), 1);
+ }
+
+
+ return 0;
+}
+
+/*
+ * kgdbeth_read_char
+ *
+ * This is a GDB stub routine. It waits for a character from the
+ * ethernet interface and then returns it.
+ */
+static int
+kgdbeth_read_char(void)
+{
+ int chr;
+
+ while ((chr = read_char()) < 0) {
+ if (send_skb) {
+ kgdb_eth_reply_arp();
+ }
+ (*kgdbeth_irqhandler)(kgdb_netdevice->irq,
+ (void *)kgdb_netdevice, 0);
+ }
+ return chr;
+}
+
+/*
+ * Hold onto the xmitlock and keep holding till the session ends.
+ * Disable device irq and keep it disabled till this session ends.
+ * Respond to last arp request we have received. It may not not have be
+ * responded yet.
+ */
+static void kgdbeth_begin_session(void) {
+ kgdbeth_holdxlock();
+ disable_irq(kgdb_netdevice->irq);
+ kgdb_eth_reply_arp();
+}
+
+static void kgdbeth_end_session(void)
+{
+ enable_irq(kgdb_netdevice->irq);
+ kgdbeth_relxlock();
+}
+
+struct kgdb_serial kgdbeth_serial = {
+ .chunksize = CHUNKSIZE,
+ .read_char = kgdbeth_read_char,
+ .write_char = kgdbeth_write_char,
+ .hook = kgdbeth_hook,
+ .flush = kgdbeth_flush,
+ .begin_session = kgdbeth_begin_session,
+ .end_session = kgdbeth_end_session,
+};
+
+static int __init parse_hw_addr(char **ptr, unsigned char *addr,
+ unsigned char delimiter)
+{
+ int i = 0;
+
+ while(**ptr != delimiter)
+ {
+ unsigned int c;
+
+ if (sscanf(*ptr, "%x", &c) != 1) {
+ return 1;
+ }
+ addr[i++] = c;
+ do
+ (*ptr)++;
+ while((**ptr != delimiter) && (*(*ptr -1) != ':'));
+ if (i > 6)
+ return -EINVAL;
+ }
+ if (i != 6)
+ return -EINVAL;
+ return 0;
+}
+
+int kgdbeth_thread(void *data)
+{
+ struct net_device *ndev = (struct net_device *)data;
+ daemonize("kgdbeth");
+ while (!ndev->ip_ptr) {
+ schedule();
+ }
+ debugger_entry();
+ return 0;
+}
+
+int kgdbeth_event(struct notifier_block * self, unsigned long val, void *
data)
+{
+ if (strcmp(((struct net_device *)data)->name, kgdb_netdevname)) {
+ goto out;
+ }
+ if (val!= NETDEV_UP)
+ goto out;
+ kernel_thread(kgdbeth_thread, data, CLONE_KERNEL);
+
+out:
+ return NOTIFY_DONE;
+}
+static struct notifier_block nb = {
+ .notifier_call = kgdbeth_event,
+};
+
+/*
+ * Syntax for this cmdline option is
+ * kgdbeth=interfacenum,localmac,listenport,remoteip,remotemac
+ */
+
+static int __init kgdbeth_opt(char *str)
+{
+ char ipaddrstr[16];
+ char *ipaddrptr = ipaddrstr;
+ extern int register_netdevice_notifier(struct notifier_block *nb);
+
+
+ if (register_netdevice_notifier(&nb)) {
+ printk("KGDB_ETH: couldn't register notifier\n");
+ return 0;
+ }
+
+ /* interfacenum */
+ if (*str < '0' || *str > '9')
+ goto errout;
+ kgdb_eth = *str - '0';
+ str++;
+ if (*str != ',')
+ goto errout;
+ str++;
+
+ /* localmac */
+ if (parse_hw_addr(&str, kgdb_localmac, ','))
+ goto errout;
+ str++;
+
+ /* port */
+ kgdb_listenport = simple_strtoul(str, &str, 10);
+
+ if (*str != ',')
+ goto errout;
+ str++;
+
+ /* remoteip */
+ while (*str != ',') {
+ if (!*str)
+ goto errout;
+ *ipaddrptr = *str;
+ str++;
+ ipaddrptr++;
+ }
+ str++;
+ *ipaddrptr = '\0';
+ kgdb_remoteip = in_aton(ipaddrstr);
+
+ /* remotemac */
+ if (parse_hw_addr(&str, kgdb_remotemac, '\0'))
+ goto errout;
+
+ kgdb_serial = &kgdbeth_serial;
+ return 1;
+
+errout:
+ printk("Invalid syntax for option kgdbeth=\n");
+ return 0;
+}
+
+__setup("kgdbeth=", kgdbeth_opt);
diff -Naurp linux-2.6.1/drivers/net/Makefile
linux-2.6.1-kgdb-2.0.5-eth/drivers/net/Makefile
--- linux-2.6.1/drivers/net/Makefile 2003-11-24 07:01:27.000000000 +0530
+++ linux-2.6.1-kgdb-2.0.5-eth/drivers/net/Makefile 2004-01-12
19:11:13.000000000 +0530
@@ -32,6 +32,8 @@ obj-$(CONFIG_BMAC) += bmac.o
obj-$(CONFIG_OAKNET) += oaknet.o 8390.o
+obj-$(CONFIG_KGDB_ETH) += kgdb_eth.o
+
obj-$(CONFIG_DGRS) += dgrs.o
obj-$(CONFIG_RCPCI) += rcpci.o
obj-$(CONFIG_VORTEX) += 3c59x.o
diff -Naurp linux-2.6.1/net/core/dev.c
linux-2.6.1-kgdb-2.0.5-eth/net/core/dev.c
--- linux-2.6.1/net/core/dev.c 2004-01-10 11:01:50.000000000 +0530
+++ linux-2.6.1-kgdb-2.0.5-eth/net/core/dev.c 2004-01-20 14:27:38.000000000
+0530
@@ -1380,7 +1380,6 @@ static void sample_queue(unsigned long d
}
#endif
-
/**
* netif_rx - post buffer to the network code
* @skb: buffer to post
@@ -1404,6 +1403,15 @@ int netif_rx(struct sk_buff *skb)
int this_cpu;
struct softnet_data *queue;
unsigned long flags;
+ int ret;
+ int kgdb_net_interrupt(struct sk_buff *skb);
+
+#ifdef CONFIG_KGDB_ETH
+ /* See if kgdb_eth wants this packet */
+ if ((ret = kgdb_net_interrupt(skb)) == NET_RX_DROP) {
+ return ret;
+ }
+#endif
if (!skb->stamp.tv_sec)
do_gettimeofday(&skb->stamp);
diff -Naurp linux-2.6.1/net/core/skbuff.c
linux-2.6.1-kgdb-2.0.5-eth/net/core/skbuff.c
--- linux-2.6.1/net/core/skbuff.c 2003-11-24 07:02:36.000000000 +0530
+++ linux-2.6.1-kgdb-2.0.5-eth/net/core/skbuff.c 2004-01-20 16:10:42.000000000
+0530
@@ -55,6 +55,7 @@
#include <linux/rtnetlink.h>
#include <linux/init.h>
#include <linux/highmem.h>
+#include <linux/debugger.h>
#include <net/protocol.h>
#include <net/dst.h>
@@ -126,6 +127,11 @@ struct sk_buff *alloc_skb(unsigned int s
{
struct sk_buff *skb;
u8 *data;
+ struct sk_buff *kgdbeth_alloc_skb(int size);
+
+ if (atomic_read(&debugger_active)) {
+ return kgdbeth_alloc_skb(size);
+ }
/* Get the HEAD */
skb = kmem_cache_alloc(skbuff_head_cache,
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: kgdb 2.0.5 2004-01-20 12:13 kgdb 2.0.5 Amit S. Kale @ 2004-01-20 13:37 ` Pavel Machek 2004-01-20 20:35 ` Pavel Machek 1 sibling, 0 replies; 4+ messages in thread From: Pavel Machek @ 2004-01-20 13:37 UTC (permalink / raw) To: Amit S. Kale Cc: Linux Kernel, George Anzinger, Steve Gonczi, Matt Mackall, Jeff Garzik Hi! > kgdb 2.0.5 is available at > http://kgdb.sourceforge.net/kgdb-2/linux-2.6.1-kgdb-2.0.5.tar.bz2 > > ChangeLog > 2004-01-20 Amit S. Kale <amitkale@emsyssoft.com> > * Created a ring buffer for kgdb ethernet packets. Several > fixes and changes to kgdb on ethernet. > > 2004-01-20 TimeSys Corporation > * Fixed a problem with not responding to Ctrl+C during priting of > console messages through gdb. > > I have pasted below eth.patch for review. When using the ethernet interface, > gdb times out several times. It receives packets and junk instead of acks. I > see following type of messages out of 8139too.c on the console > "eth0:Out-of-sync dirty pointer, 15 vs. 20." > > Any comments/suggestions/fixes on this patch are most welcome. BTW here's 2.0.4 to 2.0.5 [pseudo]diff. Index: patches/kgdb =================================================================== RCS file: /home2/machek/cvsroot/misc/kernel/patches/kgdb,v retrieving revision 1.8 diff -u -u -r1.8 kgdb --- patches/kgdb 18 Jan 2004 19:17:34 -0000 1.8 +++ patches/kgdb 20 Jan 2004 13:32:04 -0000 @@ -2031,12 +2031,14 @@ obj-$(CONFIG_VORTEX) += 3c59x.o Index: linux/drivers/net/kgdb_eth.c =================================================================== ---- linux.orig/drivers/net/kgdb_eth.c 2004-01-17 14:58:20.000000000 +0100 -+++ linux/drivers/net/kgdb_eth.c 2004-01-17 14:58:20.000000000 +0100 -@@ -0,0 +1,588 @@ +--- linux.orig/drivers/net/kgdb_eth.c 2004-01-20 14:29:19.000000000 +0100 ++++ linux/drivers/net/kgdb_eth.c 2004-01-20 14:29:19.000000000 +0100 +@@ -0,0 +1,704 @@ +/* + * Network interface GDB stub + * ++ * Copyright (C), 2004 Amit S. Kale ++ * + * Written by San Mehat (nettwerk@biodome.org) + * Based upon 'gdbserial' by David Grothe (dave@gcom.com) + * and Scott Foehner (sfoehner@engr.sgi.com) @@ -2045,6 +2047,8 @@ + * and wangdi <wangdi@clusterfs.com>. + * + * Restructured for generic a gdb interface ++ * Reveral changes to make it free of device driver changes. ++ * Added internal buffers for this interface. + * by Amit S. Kale <amitkale@emsyssoft.com> + * Some cleanups by Pavel Machek <pavel@suse.cz> + */ @@ -2100,9 +2104,107 @@ +static int kgdbeth_sendbufchars; +static irqreturn_t (*kgdbeth_irqhandler)(int, void *, struct pt_regs *) = NULL; + -+int kgdbeth_is_trapped; +struct net_device *kgdb_netdevice = NULL; + ++/* Indicates dept of recursion for xmitlock hold */ ++static int xlockholdcount = 0; ++ ++/* kgdb ethernet ring buffers. Increase the space if you get panics in ++ * kgdbeth_alloc_skb. ++ * Status of send_skbs can be known from the field users. ++ * If it's 1 the buffer is free. ++ * If the count 2 or more, the buffer is in use. ++ * Keeping 1 as the initial count prevents kfree_skb from freeing it. */ ++#define SEND_BUFLEN 1024 ++#define NUM_SENDBUF 128 ++static char *send_bufs[NUM_SENDBUF]; ++static struct sk_buff *send_skbs[NUM_SENDBUF]; ++static struct sk_buff *send_skb; ++static int bufnum; ++ ++static char kgdb_netdevname[16]; ++ ++#ifdef CONFIG_KGDB_CONSOLE ++#error kgdb over ethernet is not yet ready for console messages. ++#endif ++ ++/* ++ * Returns next skb from kgdb skbs. ++ * Initializes users field of the skb to 2 so that kfree_skb doesn't attempt ++ * freeing it. ++ * Always call after holding xmitlock of the ethernet device. ++ */ ++ ++struct sk_buff *kgdbeth_alloc_skb(int size) ++{ ++ struct sk_buff *skb; ++ u8 *data; ++ int i; ++ ++ i = bufnum; ++ ++ do { ++ skb = send_skbs[i]; ++ ++ if (atomic_read(&skb->users) == 1) { ++ bufnum = i; ++ i = -1; ++ break; ++ } ++ i = (i + 1) % NUM_SENDBUF; ++ } while (i != bufnum); ++ if (i == bufnum) { ++ panic("kgdb ethernet buffer overflow\n"); ++ } ++ data = (u8 *)(send_bufs[bufnum]); ++ ++ size = SKB_DATA_ALIGN(size); ++ if (size + sizeof(struct skb_shared_info) > SEND_BUFLEN) ++ panic("kgdb ethernet buffer too short for this request"); ++ ++ memset(skb, 0, offsetof(struct sk_buff, truesize)); ++ skb->truesize = size + sizeof(struct sk_buff); ++ atomic_set(&skb->users, 2); ++ skb->head = data; ++ skb->data = data; ++ skb->tail = data; ++ skb->end = data + size; ++ ++ atomic_set(&(skb_shinfo(skb)->dataref), 1); ++ skb_shinfo(skb)->nr_frags = 0; ++ skb_shinfo(skb)->tso_size = 0; ++ skb_shinfo(skb)->tso_segs = 0; ++ skb_shinfo(skb)->frag_list = NULL; ++ ++ return skb; ++} ++ ++/* Holds xmitlock of the ethernet device ++ * Recursive calls allowed */ ++static void kgdbeth_holdxlock(void) ++{ ++ if (spin_is_locked(&kgdb_netdevice->xmit_lock)) { ++ if (kgdb_netdevice->xmit_lock_owner == smp_processor_id()) { ++ goto gotit; ++ } ++ } ++ spin_lock(&kgdb_netdevice->xmit_lock); ++ kgdb_netdevice->xmit_lock_owner = smp_processor_id(); ++ ++gotit: ++ xlockholdcount++; ++} ++ ++/* releases xmitlock of the ethernet device ++ * Recursive calls allowed */ ++static void kgdbeth_relxlock(void) ++{ ++ if (--xlockholdcount) { ++ kgdb_netdevice->xmit_lock_owner = -1; ++ spin_unlock(&kgdb_netdevice->xmit_lock); ++ } ++} ++ +/* + * Get a char if available, return -1 if nothing available. + * Empty the receive buffer first, then look at the interface hardware. @@ -2143,16 +2245,14 @@ + if (!(in_dev->ifa_list)) { + panic("No interface address set for interface!\n"); + } ++ kgdbeth_holdxlock(); + + udp_len = len + sizeof(struct udphdr); + ip_len = eth_len = udp_len + sizeof(struct iphdr); + total_len = eth_len + ETH_HLEN; + -+ if (!(skb = alloc_skb(total_len, GFP_ATOMIC))) { -+ return; -+ } ++ skb = kgdbeth_alloc_skb(total_len); + -+ atomic_set(&skb->users, 1); + skb_reserve(skb, total_len - 1); + + memcpy(skb->data, (unsigned char *) buf, len); @@ -2183,11 +2283,11 @@ + memcpy(eth->h_source, kgdb_localmac, kgdb_netdevice->addr_len); + memcpy(eth->h_dest, kgdb_remotemac, kgdb_netdevice->addr_len); + -+ spin_lock(&kgdb_netdevice->xmit_lock); -+ kgdb_netdevice->xmit_lock_owner = smp_processor_id(); + kgdb_netdevice->hard_start_xmit(skb, kgdb_netdevice); -+ kgdb_netdevice->xmit_lock_owner = -1; -+ spin_unlock(&kgdb_netdevice->xmit_lock); ++ if (atomic_read(&skb->users) != 1) { ++ BUG(); ++ } ++ kgdbeth_relxlock(); +} + +static void kgdbeth_flush(void) @@ -2213,18 +2313,14 @@ + * arp requests, so handle them here. + */ + -+static struct sk_buff *send_skb = NULL; -+ +static void +kgdb_eth_reply_arp(void) +{ + if (send_skb) { -+ spin_lock(&kgdb_netdevice->xmit_lock); -+ kgdb_netdevice->xmit_lock_owner = smp_processor_id(); ++ kgdbeth_holdxlock(); + kgdb_netdevice->hard_start_xmit(send_skb, kgdb_netdevice); -+ kgdb_netdevice->xmit_lock_owner = -1; -+ spin_unlock(&kgdb_netdevice->xmit_lock); + send_skb = NULL; ++ kgdbeth_relxlock(); + } +} + @@ -2293,14 +2389,19 @@ + if (LOOPBACK(tip) || MULTICAST(tip)) { + return 0; + } -+ + /* reply to the ARP request */ + -+ send_skb = alloc_skb(sizeof(struct arphdr) + 2 * (kgdb_netdevice->addr_len + 4) + LL_RESERVED_SPACE(kgdb_netdevice), GFP_ATOMIC); -+ -+ if (send_skb == NULL) { -+ return 0; ++ kgdbeth_holdxlock(); ++ if (send_skb) { ++ /* Get rid of any previous replies to ARP request. We hope ++ * that regular reply to ARP by network layers would have gone ++ * out by now. */ ++ kfree_skb(send_skb); + } ++ send_skb = kgdbeth_alloc_skb(sizeof(struct arphdr) + ++ 2 * (kgdb_netdevice->addr_len + 4) + ++ LL_RESERVED_SPACE(kgdb_netdevice)); ++ kgdbeth_relxlock(); + + skb_reserve(send_skb, LL_RESERVED_SPACE(kgdb_netdevice)); + send_skb->nh.raw = send_skb->data; @@ -2350,18 +2451,16 @@ + * kgdb buffer + * + * When the kgdb stub routine read_char() is called it draws characters -+ * out of the buffer until it is empty and then reads directly from the -+ * serial port. -+ * -+ * We do not attempt to write chars from the interrupt routine since -+ * the stubs do all of that via write_char() which writes one byte -+ * after waiting for the interface to become ready. -+ * -+ * The debug stubs like to run with interrupts disabled since, after all, -+ * they run as a consequence of a breakpoint in the kernel. ++ * out of the buffer until it is empty and then polls the hardware. + * -+ * NOTE: Return value of 1 means it was for us and is an indication to -+ * the calling driver to destroy the sk_buff and not send it up the stack. ++ * Return value of NET_RX_DROP means skb was used by this function. ++ * NET_RX_SUCCESS indicates this function didn't use it. ++ ++ * Be prepared to respond to ARP requests. Let the caller also handle ++ * them if debugger is not active. We'll respond to the ARP request when ++ * a debugging session begins. It's necessary to respond to the request as the ++ * debugging session may begin even before kernel has a chance finish the ++ * response. + */ +int +kgdb_net_interrupt(struct sk_buff *skb) @@ -2374,18 +2473,16 @@ + int i; + + if (!kgdb_initialized || !kgdb_netdevice) { -+ return 0; ++ goto out; + } + if (skb->protocol == __constant_htons(ETH_P_ARP) && !send_skb) { + make_arp_request(skb); -+ return 0; -+ } -+ if (iph->protocol != IPPROTO_UDP) { -+ return 0; -+ } -+ if (be16_to_cpu(udph->dest) != kgdb_listenport) { -+ return 0; ++ goto out; + } ++ if (iph->protocol != IPPROTO_UDP || ++ be16_to_cpu(udph->dest) != kgdb_listenport) ++ goto out; ++ + kgdb_sendport = be16_to_cpu(udph->source); + + len = (be16_to_cpu(iph->tot_len) - @@ -2395,7 +2492,9 @@ + chr = *data++; + if (chr == 3) + { -+ breakpoint(); ++ if (!atomic_read(&debugger_active)) { ++ breakpoint(); ++ } + continue; + } + if (atomic_read(&kgdb_buf_in_cnt) >= GDB_BUF_SIZE) { @@ -2409,36 +2508,49 @@ + kgdb_buf_in_inx &= (GDB_BUF_SIZE - 1); + atomic_inc(&kgdb_buf_in_cnt) ; + } ++ return NET_RX_DROP; + -+ return 1; ++out: ++ if (atomic_read(&debugger_active)) ++ return NET_RX_DROP; ++ return NET_RX_SUCCESS; +} + ++/* ++ * Initializes ethernet interface to kgdb. ++ * Searches the device list and finds the device specified on kernel command ++ * line. ++ * Finds the irq handler for the device and saves its reference. ++ * Initializes kgdbeth data structures. ++ */ ++ +int +kgdbeth_hook(void) +{ -+ char kgdb_netdev[16]; + extern void kgdb_respond_ok(void); + struct irqaction *ia_ptr; ++ int i; + -+ sprintf(kgdb_netdev, "eth%d", kgdb_eth); ++ sprintf(kgdb_netdevname, "eth%d", kgdb_eth); + + for (kgdb_netdevice = dev_base; + kgdb_netdevice != NULL; + kgdb_netdevice = kgdb_netdevice->next) { -+ if (strncmp(kgdb_netdevice->name, kgdb_netdev, IFNAMSIZ) == 0) { ++ if (strncmp(kgdb_netdevice->name, kgdb_netdevname, IFNAMSIZ) == 0) { + break; + } + } + if (!kgdb_netdevice) { -+ printk("kgdbeth: Unable to find interface %s\n", kgdb_netdev); -+ return -1; ++ printk("kgdbeth: Unable to find interface %s\n", ++ kgdb_netdevname); ++ return -ENODEV; + } + if (!(kgdb_netdevice->flags & IFF_UP)) { -+ return -1; ++ return -EINVAL; + } + ia_ptr = irq_desc[kgdb_netdevice->irq].action; + while (ia_ptr) { -+ if (!strncmp(kgdb_netdev, ia_ptr->name, IFNAMSIZ)) { ++ if (!strncmp(kgdb_netdevname, ia_ptr->name, IFNAMSIZ)) { + kgdbeth_irqhandler = ia_ptr->handler; + break; + } @@ -2446,32 +2558,28 @@ + } + if (!kgdbeth_irqhandler) { + printk("kgdbeth: Interface %s doesn't have an interrupt" -+ " handler cannot use it\n", kgdb_netdev); -+ return -1; ++ " handler cannot use it\n", kgdb_netdevname); ++ return -EINVAL; ++ } ++ for (i = 0; i < NUM_SENDBUF; i++) { ++ send_skbs[i] = kmalloc(sizeof (struct sk_buff), GFP_KERNEL); ++ send_bufs[i] = kmalloc(SEND_BUFLEN, GFP_KERNEL); ++ if (!send_skbs[i] || !send_bufs[i]) { ++ printk("kgdbeth: not enough memory\n"); ++ return -ENOMEM; ++ } ++ atomic_set(&(send_skbs[i]->users), 1); + } + -+ return 0; -+} + -+/* -+ * Poll an ethernet interface for incoming characters -+ */ -+static void kgdbeth_rx_poll(void) -+{ -+ if (!kgdbeth_is_trapped) -+ disable_irq(kgdb_netdevice->irq); -+ kgdbeth_irqhandler(kgdb_netdevice->irq, (void *)kgdb_netdevice, 0); -+ if (!kgdbeth_is_trapped) -+ enable_irq(kgdb_netdevice->irq); ++ return 0; +} + +/* + * kgdbeth_read_char + * + * This is a GDB stub routine. It waits for a character from the -+ * serial interface and then returns it. If there is no serial -+ * interface connection then it returns a bogus value which will -+ * almost certainly cause the system to hang. ++ * ethernet interface and then returns it. + */ +static int +kgdbeth_read_char(void) @@ -2482,20 +2590,28 @@ + if (send_skb) { + kgdb_eth_reply_arp(); + } -+ kgdbeth_rx_poll(); ++ (*kgdbeth_irqhandler)(kgdb_netdevice->irq, ++ (void *)kgdb_netdevice, 0); + } + return chr; +} + -+static void kgdbeth_begin_session(void) -+{ -+ kgdbeth_is_trapped = 1; ++/* ++ * Hold onto the xmitlock and keep holding till the session ends. ++ * Disable device irq and keep it disabled till this session ends. ++ * Respond to last arp request we have received. It may not not have be ++ * responded yet. ++ */ ++static void kgdbeth_begin_session(void) { ++ kgdbeth_holdxlock(); ++ disable_irq(kgdb_netdevice->irq); + kgdb_eth_reply_arp(); +} + +static void kgdbeth_end_session(void) +{ -+ kgdbeth_is_trapped = 0; ++ enable_irq(kgdb_netdevice->irq); ++ kgdbeth_relxlock(); +} + +struct kgdb_serial kgdbeth_serial = { @@ -2545,7 +2661,7 @@ + +int kgdbeth_event(struct notifier_block * self, unsigned long val, void * data) +{ -+ if (strcmp(((struct net_device *)data)->name, "eth0")) { ++ if (strcmp(((struct net_device *)data)->name, kgdb_netdevname)) { + goto out; + } + if (val!= NETDEV_UP) @@ -2624,8 +2740,8 @@ +__setup("kgdbeth=", kgdbeth_opt); Index: linux/include/asm-i386/kgdb.h =================================================================== ---- linux.orig/include/asm-i386/kgdb.h 2004-01-17 14:58:20.000000000 +0100 -+++ linux/include/asm-i386/kgdb.h 2004-01-17 14:58:20.000000000 +0100 +--- linux.orig/include/asm-i386/kgdb.h 2004-01-20 14:29:19.000000000 +0100 ++++ linux/include/asm-i386/kgdb.h 2004-01-20 14:29:19.000000000 +0100 @@ -0,0 +1,49 @@ +#ifndef _ASM_KGDB_H_ +#define _ASM_KGDB_H_ @@ -3131,49 +3247,10 @@ +char *kgdb_mem2hex(char *mem, char *buf, int count, int can_fault); + +#endif /* _KGDB_H_ */ -Index: linux/include/linux/netdevice.h -=================================================================== ---- linux.orig/include/linux/netdevice.h 2004-01-17 14:56:40.000000000 +0100 -+++ linux/include/linux/netdevice.h 2004-01-17 14:58:20.000000000 +0100 -@@ -533,6 +533,11 @@ - extern struct net_device *dev_get_by_index(int ifindex); - extern struct net_device *__dev_get_by_index(int ifindex); - extern int dev_restart(struct net_device *dev); -+#ifdef CONFIG_KGDB_ETH -+extern int kgdb_net_interrupt(struct sk_buff *skb); -+extern void kgdb_send_arp_request(void); -+extern int kgdbeth_is_trapped; -+#endif - - typedef int gifconf_func_t(struct net_device * dev, char * bufptr, int len); - extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf); -@@ -591,12 +596,22 @@ - - static inline void netif_wake_queue(struct net_device *dev) - { -+#ifdef CONFIG_KGDB_ETH -+ if (kgdbeth_is_trapped) { -+ return; -+ } -+#endif - if (test_and_clear_bit(__LINK_STATE_XOFF, &dev->state)) - __netif_schedule(dev); - } - - static inline void netif_stop_queue(struct net_device *dev) - { -+#ifdef CONFIG_KGDB_ETH -+ if (kgdbeth_is_trapped) { -+ return; -+ } -+#endif - set_bit(__LINK_STATE_XOFF, &dev->state); - } - Index: linux/include/linux/sched.h =================================================================== ---- linux.orig/include/linux/sched.h 2004-01-17 14:56:40.000000000 +0100 -+++ linux/include/linux/sched.h 2004-01-17 14:58:20.000000000 +0100 +--- linux.orig/include/linux/sched.h 2004-01-20 14:28:45.000000000 +0100 ++++ linux/include/linux/sched.h 2004-01-20 14:29:19.000000000 +0100 @@ -173,7 +173,9 @@ #define MAX_SCHEDULE_TIMEOUT LONG_MAX @@ -3235,9 +3312,9 @@ # According to Alan Modra <alan@linuxcare.com.au>, the -fno-omit-frame-pointer is Index: linux/kernel/kgdbstub.c =================================================================== ---- linux.orig/kernel/kgdbstub.c 2004-01-17 14:58:20.000000000 +0100 -+++ linux/kernel/kgdbstub.c 2004-01-17 14:58:20.000000000 +0100 -@@ -0,0 +1,1227 @@ +--- linux.orig/kernel/kgdbstub.c 2004-01-20 14:29:19.000000000 +0100 ++++ linux/kernel/kgdbstub.c 2004-01-20 14:29:19.000000000 +0100 +@@ -0,0 +1,1236 @@ +/* + * 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 @@ -3453,12 +3530,6 @@ + i = 0; + while ((ch = kgdb_serial->read_char()) == gdbseq[i++] && + checkconnect) { -+ if (ch == 3) { -+ kgdb_serial->write_char('+'); -+ if (kgdb_serial->flush) -+ kgdb_serial->flush(); -+ breakpoint(); -+ } + if (!gdbseq[i]) { + kgdb_serial->write_char('+'); + if (kgdb_serial->flush) @@ -3472,7 +3543,12 @@ + break; + } + } -+ ++ if (checkconnect && ch == 3) { ++ kgdb_serial->write_char('+'); ++ if (kgdb_serial->flush) ++ kgdb_serial->flush(); ++ breakpoint(); ++ } + } while (( ch & 0x7f) != '+'); + +} @@ -3793,6 +3869,11 @@ + +/* + * This function does all command procesing for interfacing to gdb. ++ * ++ * Locking hierarchy: ++ * interface locks, if any (begin_session) ++ * kgdb lock (debugger_active) ++ * + */ +int kgdb_handle_exception(int exVector, int signo, int err_code, + struct pt_regs *linux_regs) @@ -3811,6 +3892,15 @@ + int numshadowth = num_online_cpus() + kgdb_ops->shadowth; + long kgdb_usethreadid = 0; + ++ /* Panic on recursive debugger calls. */ ++ if (atomic_read(&debugger_active) == smp_processor_id() + 1) { ++ return 0; ++ } ++ ++ /* Grab interface locks first */ ++ if (kgdb_serial->begin_session) ++ kgdb_serial->begin_session(); ++ + /* + * Interrupts will be restored by the 'trap return' code, except when + * single stepping. @@ -3828,9 +3918,6 @@ + + debugger_step = 0; + -+ -+ local_irq_save(flags); -+ + current->thread.debuggerinfo = linux_regs; + + if (kgdb_ops->disable_hw_debug) @@ -3848,9 +3935,6 @@ + /* Master processor is completely in the debugger */ + if (kgdb_ops->post_master_code) + kgdb_ops->post_master_code(linux_regs, exVector, err_code); -+ if (kgdb_serial->begin_session) -+ kgdb_serial->begin_session(); -+ + if (atomic_read(&kgdb_killed_or_detached) && + atomic_read(&kgdb_might_be_resumed)) { + getpacket(remcomInBuffer); @@ -4221,9 +4305,6 @@ + + if(kgdb_ops->handler_exit) + kgdb_ops->handler_exit(); -+ if (kgdb_serial->end_session) -+ kgdb_serial->end_session(); -+ + procindebug[smp_processor_id()] = 0; + + for (i = 0; i < num_online_cpus(); i++) { @@ -4244,9 +4325,14 @@ + } + + /* Free debugger_active */ -+ atomic_set(&debugger_active, 0); + atomic_set(&kgdb_killed_or_detached, 1); ++ atomic_set(&debugger_active, 0); + local_irq_restore(flags); ++ ++ /* Release interface locks */ ++ if (kgdb_serial->end_session) ++ kgdb_serial->end_session(); ++ + return ret; +} + @@ -4550,8 +4636,8 @@ return; Index: linux/net/core/dev.c =================================================================== ---- linux.orig/net/core/dev.c 2004-01-17 14:56:40.000000000 +0100 -+++ linux/net/core/dev.c 2004-01-17 14:58:20.000000000 +0100 +--- linux.orig/net/core/dev.c 2004-01-20 14:28:45.000000000 +0100 ++++ linux/net/core/dev.c 2004-01-20 14:29:19.000000000 +0100 @@ -1380,7 +1380,6 @@ } #endif @@ -4560,25 +4646,43 @@ /** * netif_rx - post buffer to the network code * @skb: buffer to post -@@ -1405,6 +1404,21 @@ +@@ -1404,6 +1403,15 @@ + int this_cpu; struct softnet_data *queue; unsigned long flags; - ++ int ret; ++ int kgdb_net_interrupt(struct sk_buff *skb); ++ +#ifdef CONFIG_KGDB_ETH + /* See if kgdb_eth wants this packet */ -+ if (!kgdb_net_interrupt(skb)) { -+ /* No.. if we're 'trapped' then junk it */ -+ if (kgdbeth_is_trapped) { -+ kfree_skb(skb); -+ return NET_RX_DROP; -+ } -+ } else { -+ /* kgdb_eth ate the packet... drop it silently */ -+ kfree_skb(skb); -+ return NET_RX_DROP; ++ if ((ret = kgdb_net_interrupt(skb)) == NET_RX_DROP) { ++ return ret; + } +#endif -+ + if (!skb->stamp.tv_sec) do_gettimeofday(&skb->stamp); +Index: linux/net/core/skbuff.c +=================================================================== +--- linux.orig/net/core/skbuff.c 2003-11-24 17:08:33.000000000 +0100 ++++ linux/net/core/skbuff.c 2004-01-20 14:29:19.000000000 +0100 +@@ -55,6 +55,7 @@ + #include <linux/rtnetlink.h> + #include <linux/init.h> + #include <linux/highmem.h> ++#include <linux/debugger.h> + + #include <net/protocol.h> + #include <net/dst.h> +@@ -126,6 +127,11 @@ + { + struct sk_buff *skb; + u8 *data; ++ struct sk_buff *kgdbeth_alloc_skb(int size); ++ ++ if (atomic_read(&debugger_active)) { ++ return kgdbeth_alloc_skb(size); ++ } + /* Get the HEAD */ + skb = kmem_cache_alloc(skbuff_head_cache, -- When do you have a heart between your knees? [Johanka's followup: and *two* hearts?] ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: kgdb 2.0.5 2004-01-20 12:13 kgdb 2.0.5 Amit S. Kale 2004-01-20 13:37 ` Pavel Machek @ 2004-01-20 20:35 ` Pavel Machek 2004-01-21 16:04 ` Amit S. Kale 1 sibling, 1 reply; 4+ messages in thread From: Pavel Machek @ 2004-01-20 20:35 UTC (permalink / raw) To: Amit S. Kale Cc: Linux Kernel, George Anzinger, Steve Gonczi, Matt Mackall, Jeff Garzik Hi! > kgdb 2.0.5 is available at > http://kgdb.sourceforge.net/kgdb-2/linux-2.6.1-kgdb-2.0.5.tar.bz2 > > ChangeLog > 2004-01-20 Amit S. Kale <amitkale@emsyssoft.com> > * Created a ring buffer for kgdb ethernet packets. Several > fixes and changes to kgdb on ethernet. > > 2004-01-20 TimeSys Corporation > * Fixed a problem with not responding to Ctrl+C during priting of > console messages through gdb. > > I have pasted below eth.patch for review. When using the ethernet interface, > gdb times out several times. It receives packets and junk instead of acks. I > see following type of messages out of 8139too.c on the console > "eth0:Out-of-sync dirty pointer, 15 vs. 20." > > Any comments/suggestions/fixes on this patch are most welcome. Okay, so you wanted comments :-) @@ -2031,12 +2031,14 @@ obj-$(CONFIG_VORTEX) += 3c59x.o Index: linux/drivers/net/kgdb_eth.c =================================================================== ---- linux.orig/drivers/net/kgdb_eth.c 2004-01-17 14:58:20.000000000 +0100 -+++ linux/drivers/net/kgdb_eth.c 2004-01-17 14:58:20.000000000 +0100 -@@ -0,0 +1,588 @@ +--- linux.orig/drivers/net/kgdb_eth.c 2004-01-20 14:29:19.000000000 +0100 ++++ linux/drivers/net/kgdb_eth.c 2004-01-20 14:29:19.000000000 +0100 +@@ -0,0 +1,704 @@ +/* + * Network interface GDB stub + * ++ * Copyright (C), 2004 Amit S. Kale ++ * + * Written by San Mehat (nettwerk@biodome.org) + * Based upon 'gdbserial' by David Grothe (dave@gcom.com) + * and Scott Foehner (sfoehner@engr.sgi.com) @@ -2045,6 +2047,8 @@ + * and wangdi <wangdi@clusterfs.com>. + * + * Restructured for generic a gdb interface ++ * Reveral changes to make it free of device driver changes. ++ * Added internal buffers for this interface. + * by Amit S. Kale <amitkale@emsyssoft.com> + * Some cleanups by Pavel Machek <pavel@suse.cz> + */ @@ -2100,9 +2104,107 @@ +static int kgdbeth_sendbufchars; +static irqreturn_t (*kgdbeth_irqhandler)(int, void *, struct pt_regs *) = NULL; + -+int kgdbeth_is_trapped; +struct net_device *kgdb_netdevice = NULL; + ++/* Indicates dept of recursion for xmitlock hold */ ++static int xlockholdcount = 0; This should be xlock_hold_count according to CodingStyle. ++/* Holds xmitlock of the ethernet device ++ * Recursive calls allowed */ ++static void kgdbeth_holdxlock(void) ++{ Why not calling it simply kgdbeth_lock()? ++ if (spin_is_locked(&kgdb_netdevice->xmit_lock)) { ++ if (kgdb_netdevice->xmit_lock_owner == smp_processor_id()) { ++ goto gotit; ++ } ++ } ++ spin_lock(&kgdb_netdevice->xmit_lock); ++ kgdb_netdevice->xmit_lock_owner = smp_processor_id(); ++ ++gotit: ++ xlockholdcount++; ++} ++ ++/* releases xmitlock of the ethernet device ++ * Recursive calls allowed */ ++static void kgdbeth_relxlock(void) kgdbeth_unlock()? +int +kgdbeth_hook(void) +{ -+ char kgdb_netdev[16]; + extern void kgdb_respond_ok(void); + struct irqaction *ia_ptr; ++ int i; + -+ sprintf(kgdb_netdev, "eth%d", kgdb_eth); ++ sprintf(kgdb_netdevname, "eth%d", kgdb_eth); kgdb_netdev_name? @@ -2183,11 +2283,11 @@ + memcpy(eth->h_source, kgdb_localmac, kgdb_netdevice->addr_len); + memcpy(eth->h_dest, kgdb_remotemac, kgdb_netdevice->addr_len); + -+ spin_lock(&kgdb_netdevice->xmit_lock); -+ kgdb_netdevice->xmit_lock_owner = smp_processor_id(); + kgdb_netdevice->hard_start_xmit(skb, kgdb_netdevice); -+ kgdb_netdevice->xmit_lock_owner = -1; -+ spin_unlock(&kgdb_netdevice->xmit_lock); ++ if (atomic_read(&skb->users) != 1) { ++ BUG(); ++ } ++ kgdbeth_relxlock(); +} + +static void kgdbeth_flush(void) BUG_ON(atomic_read() != 1)? Pavel -- When do you have a heart between your knees? [Johanka's followup: and *two* hearts?] ^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: kgdb 2.0.5 2004-01-20 20:35 ` Pavel Machek @ 2004-01-21 16:04 ` Amit S. Kale 0 siblings, 0 replies; 4+ messages in thread From: Amit S. Kale @ 2004-01-21 16:04 UTC (permalink / raw) To: Pavel Machek Cc: Linux Kernel, George Anzinger, Steve Gonczi, Matt Mackall, Jeff Garzik On Wednesday 21 Jan 2004 2:05 am, Pavel Machek wrote: > Hi! > > > kgdb 2.0.5 is available at > > http://kgdb.sourceforge.net/kgdb-2/linux-2.6.1-kgdb-2.0.5.tar.bz2 > > > > ChangeLog > > 2004-01-20 Amit S. Kale <amitkale@emsyssoft.com> > > * Created a ring buffer for kgdb ethernet packets. Several > > fixes and changes to kgdb on ethernet. > > > > 2004-01-20 TimeSys Corporation > > * Fixed a problem with not responding to Ctrl+C during priting of > > console messages through gdb. > > > > I have pasted below eth.patch for review. When using the ethernet > > interface, gdb times out several times. It receives packets and junk > > instead of acks. I see following type of messages out of 8139too.c on the > > console > > "eth0:Out-of-sync dirty pointer, 15 vs. 20." > > > > Any comments/suggestions/fixes on this patch are most welcome. > > Okay, so you wanted comments :-) Done. These will appear in 2.0.6 > > > @@ -2031,12 +2031,14 @@ > obj-$(CONFIG_VORTEX) += 3c59x.o > Index: linux/drivers/net/kgdb_eth.c > =================================================================== > ---- linux.orig/drivers/net/kgdb_eth.c 2004-01-17 14:58:20.000000000 +0100 > -+++ linux/drivers/net/kgdb_eth.c 2004-01-17 14:58:20.000000000 +0100 > -@@ -0,0 +1,588 @@ > +--- linux.orig/drivers/net/kgdb_eth.c 2004-01-20 14:29:19.000000000 +0100 > ++++ linux/drivers/net/kgdb_eth.c 2004-01-20 14:29:19.000000000 +0100 > +@@ -0,0 +1,704 @@ > +/* > + * Network interface GDB stub > + * > ++ * Copyright (C), 2004 Amit S. Kale > ++ * > + * Written by San Mehat (nettwerk@biodome.org) > + * Based upon 'gdbserial' by David Grothe (dave@gcom.com) > + * and Scott Foehner (sfoehner@engr.sgi.com) > @@ -2045,6 +2047,8 @@ > + * and wangdi <wangdi@clusterfs.com>. > + * > + * Restructured for generic a gdb interface > ++ * Reveral changes to make it free of device driver changes. > ++ * Added internal buffers for this interface. > + * by Amit S. Kale <amitkale@emsyssoft.com> > + * Some cleanups by Pavel Machek <pavel@suse.cz> > + */ > @@ -2100,9 +2104,107 @@ > +static int kgdbeth_sendbufchars; > +static irqreturn_t (*kgdbeth_irqhandler)(int, void *, struct pt_regs *) = > NULL; + > -+int kgdbeth_is_trapped; > +struct net_device *kgdb_netdevice = NULL; > + > ++/* Indicates dept of recursion for xmitlock hold */ > ++static int xlockholdcount = 0; > > This should be xlock_hold_count according to CodingStyle. > > ++/* Holds xmitlock of the ethernet device > ++ * Recursive calls allowed */ > ++static void kgdbeth_holdxlock(void) > ++{ > > Why not calling it simply kgdbeth_lock()? > > ++ if (spin_is_locked(&kgdb_netdevice->xmit_lock)) { > ++ if (kgdb_netdevice->xmit_lock_owner == smp_processor_id()) { > ++ goto gotit; > ++ } > ++ } > ++ spin_lock(&kgdb_netdevice->xmit_lock); > ++ kgdb_netdevice->xmit_lock_owner = smp_processor_id(); > ++ > ++gotit: > ++ xlockholdcount++; > ++} > ++ > ++/* releases xmitlock of the ethernet device > ++ * Recursive calls allowed */ > ++static void kgdbeth_relxlock(void) > > kgdbeth_unlock()? > > +int > +kgdbeth_hook(void) > +{ > -+ char kgdb_netdev[16]; > + extern void kgdb_respond_ok(void); > + struct irqaction *ia_ptr; > ++ int i; > + > -+ sprintf(kgdb_netdev, "eth%d", kgdb_eth); > ++ sprintf(kgdb_netdevname, "eth%d", kgdb_eth); > > kgdb_netdev_name? > > @@ -2183,11 +2283,11 @@ > + memcpy(eth->h_source, kgdb_localmac, kgdb_netdevice->addr_len); > + memcpy(eth->h_dest, kgdb_remotemac, kgdb_netdevice->addr_len); > + > -+ spin_lock(&kgdb_netdevice->xmit_lock); > -+ kgdb_netdevice->xmit_lock_owner = smp_processor_id(); > + kgdb_netdevice->hard_start_xmit(skb, kgdb_netdevice); > -+ kgdb_netdevice->xmit_lock_owner = -1; > -+ spin_unlock(&kgdb_netdevice->xmit_lock); > ++ if (atomic_read(&skb->users) != 1) { > ++ BUG(); > ++ } > ++ kgdbeth_relxlock(); > +} > + > +static void kgdbeth_flush(void) > > BUG_ON(atomic_read() != 1)? > > Pavel -- Amit Kale EmSysSoft (http://www.emsyssoft.com) KGDB: Linux Kernel Source Level Debugger (http://kgdb.sourceforge.net) ^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2004-01-21 16:04 UTC | newest] Thread overview: 4+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2004-01-20 12:13 kgdb 2.0.5 Amit S. Kale 2004-01-20 13:37 ` Pavel Machek 2004-01-20 20:35 ` Pavel Machek 2004-01-21 16:04 ` Amit S. Kale
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox