upstream u-boot with additional patches for our devices/boards: https://lists.denx.de/pipermail/u-boot/2017-March/282789.html (AXP crashes) ; Gbit ethernet patch for some LIME2 revisions ; with SPI flash support
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
u-boot/net/bootp.c

1115 lines
26 KiB

22 years ago
/*
* Based on LiMon - BOOTP.
*
* Copyright 1994, 1995, 2000 Neil Russell.
* (See License)
* Copyright 2000 Roland Borde
* Copyright 2000 Paolo Scaffardi
* Copyright 2000-2004 Wolfgang Denk, wd@denx.de
22 years ago
*/
#include <common.h>
#include <command.h>
#include <efi_loader.h>
22 years ago
#include <net.h>
#include <net/tftp.h>
22 years ago
#include "bootp.h"
#ifdef CONFIG_LED_STATUS
22 years ago
#include <status_led.h>
#endif
#ifdef CONFIG_BOOTP_RANDOM_DELAY
#include "net_rand.h"
#endif
22 years ago
#define BOOTP_VENDOR_MAGIC 0x63825363 /* RFC1048 Magic Cookie */
22 years ago
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
/*
* The timeout for the initial BOOTP/DHCP request used to be described by a
* counter of fixed-length timeout periods. TIMEOUT_COUNT represents
* that counter
*
* Now that the timeout periods are variable (exponential backoff and retry)
* we convert the timeout count to the absolute time it would have take to
* execute that many retries, and keep sending retry packets until that time
* is reached.
*/
#ifndef CONFIG_NET_RETRY_COUNT
# define TIMEOUT_COUNT 5 /* # of timeouts before giving up */
22 years ago
#else
# define TIMEOUT_COUNT (CONFIG_NET_RETRY_COUNT)
22 years ago
#endif
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
#define TIMEOUT_MS ((3 + (TIMEOUT_COUNT * 5)) * 1000)
22 years ago
#define PORT_BOOTPS 67 /* BOOTP server UDP port */
#define PORT_BOOTPC 68 /* BOOTP client UDP port */
22 years ago
#ifndef CONFIG_DHCP_MIN_EXT_LEN /* minimal length of extension list */
#define CONFIG_DHCP_MIN_EXT_LEN 64
22 years ago
#endif
net: More BOOTP retry timeout improvements It's not unusual for DHCP servers to take a couple hundred milliseconds to respond to DHCP discover messages. One possible reason for the delay can be that the server checks (typically using an ARP request) that the IP it's about to hand out isn't in use yet. To make matters worse, some servers may also queue up requests and process them sequentially, which can cause excessively long delays if clients retry too fast. Commit f59be6e850b3 ("net: BOOTP retry timeout improvements") shortened the retry timeouts significantly, but the BOOTP/DHCP implementation in U-Boot doesn't handle that well because it will ignore incoming replies to earlier requests. In one particular setup this increases the time it takes to obtain a DHCP lease from 630 ms to 8313 ms. This commit attempts to fix this in two ways. First it increases the initial retry timeout from 10 ms to 250 ms to give DHCP servers some more time to respond. At the same time a cache of outstanding DHCP request IDs is kept so that the implementation will know to continue transactions even after a retransmission of the DISCOVER message. The maximum retry timeout is also increased from 1 second to 2 seconds. An ID cache of size 4 will keep DHCP requests around for 8 seconds (once the maximum retry timeout has been reached) before dropping them. This should give servers plenty of time to respond. If it ever turns out that this isn't enough, the size of the cache can easily be increased. With this commit the DHCP lease on the above-mentioned setup still takes longer (1230 ms) than originally, but that's an acceptable compromise to improve DHCP lease acquisition time for a broader range of setups. To make it easier to benchmark DHCP in the future, this commit also adds the time it took to obtain a lease to the final "DHCP client bound to address x.x.x.x" message. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years ago
#ifndef CONFIG_BOOTP_ID_CACHE_SIZE
#define CONFIG_BOOTP_ID_CACHE_SIZE 4
#endif
u32 bootp_ids[CONFIG_BOOTP_ID_CACHE_SIZE];
net: More BOOTP retry timeout improvements It's not unusual for DHCP servers to take a couple hundred milliseconds to respond to DHCP discover messages. One possible reason for the delay can be that the server checks (typically using an ARP request) that the IP it's about to hand out isn't in use yet. To make matters worse, some servers may also queue up requests and process them sequentially, which can cause excessively long delays if clients retry too fast. Commit f59be6e850b3 ("net: BOOTP retry timeout improvements") shortened the retry timeouts significantly, but the BOOTP/DHCP implementation in U-Boot doesn't handle that well because it will ignore incoming replies to earlier requests. In one particular setup this increases the time it takes to obtain a DHCP lease from 630 ms to 8313 ms. This commit attempts to fix this in two ways. First it increases the initial retry timeout from 10 ms to 250 ms to give DHCP servers some more time to respond. At the same time a cache of outstanding DHCP request IDs is kept so that the implementation will know to continue transactions even after a retransmission of the DISCOVER message. The maximum retry timeout is also increased from 1 second to 2 seconds. An ID cache of size 4 will keep DHCP requests around for 8 seconds (once the maximum retry timeout has been reached) before dropping them. This should give servers plenty of time to respond. If it ever turns out that this isn't enough, the size of the cache can easily be increased. With this commit the DHCP lease on the above-mentioned setup still takes longer (1230 ms) than originally, but that's an acceptable compromise to improve DHCP lease acquisition time for a broader range of setups. To make it easier to benchmark DHCP in the future, this commit also adds the time it took to obtain a lease to the final "DHCP client bound to address x.x.x.x" message. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years ago
unsigned int bootp_num_ids;
int bootp_try;
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
ulong bootp_start;
ulong bootp_timeout;
char net_nis_domain[32] = {0,}; /* Our NIS domain */
char net_hostname[32] = {0,}; /* Our hostname */
char net_root_path[64] = {0,}; /* Our bootpath */
22 years ago
static ulong time_taken_max;
#if defined(CONFIG_CMD_DHCP)
12 years ago
static dhcp_state_t dhcp_state = INIT;
static u32 dhcp_leasetime;
static struct in_addr dhcp_server_ip;
static u8 dhcp_option_overload;
#define OVERLOAD_FILE 1
#define OVERLOAD_SNAME 2
static void dhcp_handler(uchar *pkt, unsigned dest, struct in_addr sip,
unsigned src, unsigned len);
22 years ago
/* For Debug */
#if 0
static char *dhcpmsg2str(int type)
22 years ago
{
switch (type) {
case 1: return "DHCPDISCOVER"; break;
case 2: return "DHCPOFFER"; break;
case 3: return "DHCPREQUEST"; break;
case 4: return "DHCPDECLINE"; break;
case 5: return "DHCPACK"; break;
case 6: return "DHCPNACK"; break;
case 7: return "DHCPRELEASE"; break;
22 years ago
default: return "UNKNOWN/INVALID MSG TYPE"; break;
}
}
#endif
#endif
22 years ago
net: More BOOTP retry timeout improvements It's not unusual for DHCP servers to take a couple hundred milliseconds to respond to DHCP discover messages. One possible reason for the delay can be that the server checks (typically using an ARP request) that the IP it's about to hand out isn't in use yet. To make matters worse, some servers may also queue up requests and process them sequentially, which can cause excessively long delays if clients retry too fast. Commit f59be6e850b3 ("net: BOOTP retry timeout improvements") shortened the retry timeouts significantly, but the BOOTP/DHCP implementation in U-Boot doesn't handle that well because it will ignore incoming replies to earlier requests. In one particular setup this increases the time it takes to obtain a DHCP lease from 630 ms to 8313 ms. This commit attempts to fix this in two ways. First it increases the initial retry timeout from 10 ms to 250 ms to give DHCP servers some more time to respond. At the same time a cache of outstanding DHCP request IDs is kept so that the implementation will know to continue transactions even after a retransmission of the DISCOVER message. The maximum retry timeout is also increased from 1 second to 2 seconds. An ID cache of size 4 will keep DHCP requests around for 8 seconds (once the maximum retry timeout has been reached) before dropping them. This should give servers plenty of time to respond. If it ever turns out that this isn't enough, the size of the cache can easily be increased. With this commit the DHCP lease on the above-mentioned setup still takes longer (1230 ms) than originally, but that's an acceptable compromise to improve DHCP lease acquisition time for a broader range of setups. To make it easier to benchmark DHCP in the future, this commit also adds the time it took to obtain a lease to the final "DHCP client bound to address x.x.x.x" message. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years ago
static void bootp_add_id(ulong id)
{
if (bootp_num_ids >= ARRAY_SIZE(bootp_ids)) {
size_t size = sizeof(bootp_ids) - sizeof(id);
memmove(bootp_ids, &bootp_ids[1], size);
bootp_ids[bootp_num_ids - 1] = id;
} else {
bootp_ids[bootp_num_ids] = id;
bootp_num_ids++;
}
}
static bool bootp_match_id(ulong id)
{
unsigned int i;
for (i = 0; i < bootp_num_ids; i++)
if (bootp_ids[i] == id)
return true;
return false;
}
static int check_reply_packet(uchar *pkt, unsigned dest, unsigned src,
unsigned len)
22 years ago
{
struct bootp_hdr *bp = (struct bootp_hdr *)pkt;
22 years ago
int retval = 0;
if (dest != PORT_BOOTPC || src != PORT_BOOTPS)
retval = -1;
else if (len < sizeof(struct bootp_hdr) - OPT_FIELD_SIZE)
22 years ago
retval = -2;
else if (bp->bp_op != OP_BOOTREPLY)
22 years ago
retval = -3;
else if (bp->bp_htype != HWT_ETHER)
retval = -4;
else if (bp->bp_hlen != HWL_ETHER)
retval = -5;
else if (!bootp_match_id(net_read_u32(&bp->bp_id)))
22 years ago
retval = -6;
else if (memcmp(bp->bp_chaddr, net_ethaddr, HWL_ETHER) != 0)
retval = -7;
22 years ago
debug("Filtering pkt = %d\n", retval);
22 years ago
return retval;
}
/*
* Copy parameters of interest from BOOTP_REPLY/DHCP_OFFER packet
*/
static void store_net_params(struct bootp_hdr *bp)
22 years ago
{
#if !defined(CONFIG_BOOTP_SERVERIP)
struct in_addr tmp_ip;
bool overwrite_serverip = true;
#if defined(CONFIG_BOOTP_PREFER_SERVERIP)
overwrite_serverip = false;
#endif
net_copy_ip(&tmp_ip, &bp->bp_siaddr);
if (tmp_ip.s_addr != 0 && (overwrite_serverip || !net_server_ip.s_addr))
net_copy_ip(&net_server_ip, &bp->bp_siaddr);
memcpy(net_server_ethaddr,
((struct ethernet_hdr *)net_rx_packet)->et_src, 6);
if (
#if defined(CONFIG_CMD_DHCP)
!(dhcp_option_overload & OVERLOAD_FILE) &&
#endif
(strlen(bp->bp_file) > 0) &&
!net_boot_file_name_explicit) {
copy_filename(net_boot_file_name, bp->bp_file,
sizeof(net_boot_file_name));
}
22 years ago
debug("net_boot_file_name: %s\n", net_boot_file_name);
22 years ago
/* Propagate to environment:
* don't delete exising entry when BOOTP / DHCP reply does
22 years ago
* not contain a new value
*/
if (*net_boot_file_name)
env_set("bootfile", net_boot_file_name);
#endif
net_copy_ip(&net_ip, &bp->bp_yiaddr);
22 years ago
}
static int truncate_sz(const char *name, int maxlen, int curlen)
22 years ago
{
if (curlen >= maxlen) {
printf("*** WARNING: %s is too long (%d - max: %d)"
" - truncated\n", name, curlen, maxlen);
22 years ago
curlen = maxlen - 1;
}
return curlen;
22 years ago
}
#if !defined(CONFIG_CMD_DHCP)
22 years ago
static void bootp_process_vendor_field(u8 *ext)
22 years ago
{
int size = *(ext + 1);
22 years ago
debug("[BOOTP] Processing extension %d... (%d bytes)\n", *ext,
*(ext + 1));
22 years ago
net_boot_file_expected_size_in_blocks = 0;
22 years ago
switch (*ext) {
/* Fixed length fields */
case 1: /* Subnet mask */
if (net_netmask.s_addr == 0)
net_copy_ip(&net_netmask, (struct in_addr *)(ext + 2));
22 years ago
break;
case 2: /* Time offset - Not yet supported */
22 years ago
break;
/* Variable length fields */
case 3: /* Gateways list */
if (net_gateway.s_addr == 0)
net_copy_ip(&net_gateway, (struct in_addr *)(ext + 2));
22 years ago
break;
case 4: /* Time server - Not yet supported */
22 years ago
break;
case 5: /* IEN-116 name server - Not yet supported */
22 years ago
break;
case 6:
if (net_dns_server.s_addr == 0)
net_copy_ip(&net_dns_server,
(struct in_addr *)(ext + 2));
#if defined(CONFIG_BOOTP_DNS2)
if ((net_dns_server2.s_addr == 0) && (size > 4))
net_copy_ip(&net_dns_server2,
(struct in_addr *)(ext + 2 + 4));
#endif
22 years ago
break;
case 7: /* Log server - Not yet supported */
22 years ago
break;
case 8: /* Cookie/Quote server - Not yet supported */
22 years ago
break;
case 9: /* LPR server - Not yet supported */
22 years ago
break;
case 10: /* Impress server - Not yet supported */
22 years ago
break;
case 11: /* RPL server - Not yet supported */
22 years ago
break;
case 12: /* Host name */
if (net_hostname[0] == 0) {
size = truncate_sz("Host Name",
sizeof(net_hostname), size);
memcpy(&net_hostname, ext + 2, size);
net_hostname[size] = 0;
22 years ago
}
break;
case 13: /* Boot file size */
22 years ago
if (size == 2)
net_boot_file_expected_size_in_blocks =
ntohs(*(ushort *)(ext + 2));
22 years ago
else if (size == 4)
net_boot_file_expected_size_in_blocks =
ntohl(*(ulong *)(ext + 2));
22 years ago
break;
case 14: /* Merit dump file - Not yet supported */
22 years ago
break;
case 15: /* Domain name - Not yet supported */
22 years ago
break;
case 16: /* Swap server - Not yet supported */
22 years ago
break;
case 17: /* Root path */
if (net_root_path[0] == 0) {
size = truncate_sz("Root Path",
sizeof(net_root_path), size);
memcpy(&net_root_path, ext + 2, size);
net_root_path[size] = 0;
22 years ago
}
break;
case 18: /* Extension path - Not yet supported */
22 years ago
/*
* This can be used to send the information of the
* vendor area in another file that the client can
* access via TFTP.
22 years ago
*/
break;
/* IP host layer fields */
case 40: /* NIS Domain name */
if (net_nis_domain[0] == 0) {
size = truncate_sz("NIS Domain Name",
sizeof(net_nis_domain), size);
memcpy(&net_nis_domain, ext + 2, size);
net_nis_domain[size] = 0;
22 years ago
}
break;
#if defined(CONFIG_CMD_SNTP) && defined(CONFIG_BOOTP_NTPSERVER)
case 42: /* NTP server IP */
net_copy_ip(&net_ntp_server, (struct in_addr *)(ext + 2));
break;
#endif
/* Application layer fields */
case 43: /* Vendor specific info - Not yet supported */
22 years ago
/*
* Binary information to exchange specific
* product information.
22 years ago
*/
break;
/* Reserved (custom) fields (128..254) */
}
22 years ago
}
static void bootp_process_vendor(u8 *ext, int size)
22 years ago
{
u8 *end = ext + size;
22 years ago
debug("[BOOTP] Checking extension (%d bytes)...\n", size);
22 years ago
while ((ext < end) && (*ext != 0xff)) {
if (*ext == 0) {
ext++;
} else {
u8 *opt = ext;
ext += ext[1] + 2;
if (ext <= end)
bootp_process_vendor_field(opt);
}
22 years ago
}
debug("[BOOTP] Received fields:\n");
if (net_netmask.s_addr)
debug("net_netmask : %pI4\n", &net_netmask);
if (net_gateway.s_addr)
debug("net_gateway : %pI4", &net_gateway);
if (net_boot_file_expected_size_in_blocks)
debug("net_boot_file_expected_size_in_blocks : %d\n",
net_boot_file_expected_size_in_blocks);
22 years ago
if (net_hostname[0])
debug("net_hostname : %s\n", net_hostname);
if (net_root_path[0])
debug("net_root_path : %s\n", net_root_path);
if (net_nis_domain[0])
debug("net_nis_domain : %s\n", net_nis_domain);
#if defined(CONFIG_CMD_SNTP) && defined(CONFIG_BOOTP_NTPSERVER)
if (net_ntp_server.s_addr)
debug("net_ntp_server : %pI4\n", &net_ntp_server);
#endif
}
22 years ago
/*
* Handle a BOOTP received packet.
*/
static void bootp_handler(uchar *pkt, unsigned dest, struct in_addr sip,
unsigned src, unsigned len)
22 years ago
{
struct bootp_hdr *bp;
22 years ago
debug("got BOOTP packet (src=%d, dst=%d, len=%d want_len=%zu)\n",
src, dest, len, sizeof(struct bootp_hdr));
22 years ago
bp = (struct bootp_hdr *)pkt;
22 years ago
/* Filter out pkts we don't want */
if (check_reply_packet(pkt, dest, src, len))
22 years ago
return;
/*
* Got a good BOOTP reply. Copy the data into our variables.
22 years ago
*/
#if defined(CONFIG_LED_STATUS) && defined(CONFIG_LED_STATUS_BOOT_ENABLE)
status_led_set(CONFIG_LED_STATUS_BOOT, CONFIG_LED_STATUS_OFF);
22 years ago
#endif
store_net_params(bp); /* Store net parameters from reply */
22 years ago
/* Retrieve extended information (we must parse the vendor area) */
if (net_read_u32((u32 *)&bp->bp_vend[0]) == htonl(BOOTP_VENDOR_MAGIC))
bootp_process_vendor((uchar *)&bp->bp_vend[4], len);
22 years ago
net_set_timeout_handler(0, (thand_f *)0);
bootstage_mark_name(BOOTSTAGE_ID_BOOTP_STOP, "bootp_stop");
22 years ago
debug("Got good BOOTP\n");
22 years ago
net_auto_load();
22 years ago
}
#endif
22 years ago
/*
* Timeout on BOOTP/DHCP request.
*/
static void bootp_timeout_handler(void)
22 years ago
{
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
ulong time_taken = get_timer(bootp_start);
if (time_taken >= time_taken_max) {
#ifdef CONFIG_BOOTP_MAY_FAIL
char *ethrotate;
ethrotate = env_get("ethrotate");
if ((ethrotate && strcmp(ethrotate, "no") == 0) ||
net_restart_wrap) {
puts("\nRetry time exceeded\n");
net_set_state(NETLOOP_FAIL);
} else
#endif
{
puts("\nRetry time exceeded; starting again\n");
net_start_again();
}
22 years ago
} else {
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
bootp_timeout *= 2;
net: More BOOTP retry timeout improvements It's not unusual for DHCP servers to take a couple hundred milliseconds to respond to DHCP discover messages. One possible reason for the delay can be that the server checks (typically using an ARP request) that the IP it's about to hand out isn't in use yet. To make matters worse, some servers may also queue up requests and process them sequentially, which can cause excessively long delays if clients retry too fast. Commit f59be6e850b3 ("net: BOOTP retry timeout improvements") shortened the retry timeouts significantly, but the BOOTP/DHCP implementation in U-Boot doesn't handle that well because it will ignore incoming replies to earlier requests. In one particular setup this increases the time it takes to obtain a DHCP lease from 630 ms to 8313 ms. This commit attempts to fix this in two ways. First it increases the initial retry timeout from 10 ms to 250 ms to give DHCP servers some more time to respond. At the same time a cache of outstanding DHCP request IDs is kept so that the implementation will know to continue transactions even after a retransmission of the DISCOVER message. The maximum retry timeout is also increased from 1 second to 2 seconds. An ID cache of size 4 will keep DHCP requests around for 8 seconds (once the maximum retry timeout has been reached) before dropping them. This should give servers plenty of time to respond. If it ever turns out that this isn't enough, the size of the cache can easily be increased. With this commit the DHCP lease on the above-mentioned setup still takes longer (1230 ms) than originally, but that's an acceptable compromise to improve DHCP lease acquisition time for a broader range of setups. To make it easier to benchmark DHCP in the future, this commit also adds the time it took to obtain a lease to the final "DHCP client bound to address x.x.x.x" message. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years ago
if (bootp_timeout > 2000)
bootp_timeout = 2000;
net_set_timeout_handler(bootp_timeout, bootp_timeout_handler);
bootp_request();
22 years ago
}
}
#define put_vci(e, str) \
do { \
size_t vci_strlen = strlen(str); \
*e++ = 60; /* Vendor Class Identifier */ \
*e++ = vci_strlen; \
memcpy(e, str, vci_strlen); \
e += vci_strlen; \
} while (0)
static u8 *add_vci(u8 *e)
{
char *vci = NULL;
char *env_vci = env_get("bootp_vci");
#if defined(CONFIG_SPL_BUILD) && defined(CONFIG_SPL_NET_VCI_STRING)
vci = CONFIG_SPL_NET_VCI_STRING;
#elif defined(CONFIG_BOOTP_VCI_STRING)
vci = CONFIG_BOOTP_VCI_STRING;
#endif
if (env_vci)
vci = env_vci;
if (vci)
put_vci(e, vci);
return e;
}
22 years ago
/*
* Initialize BOOTP extension fields in the request.
*/
#if defined(CONFIG_CMD_DHCP)
static int dhcp_extended(u8 *e, int message_type, struct in_addr server_ip,
struct in_addr requested_ip)
22 years ago
{
u8 *start = e;
u8 *cnt;
#ifdef CONFIG_LIB_UUID
char *uuid;
#endif
int clientarch = -1;
#if defined(CONFIG_BOOTP_VENDOREX)
u8 *x;
22 years ago
#endif
#if defined(CONFIG_BOOTP_SEND_HOSTNAME)
char *hostname;
#endif
22 years ago
*e++ = 99; /* RFC1048 Magic Cookie */
*e++ = 130;
*e++ = 83;
*e++ = 99;
22 years ago
*e++ = 53; /* DHCP Message Type */
*e++ = 1;
*e++ = message_type;
22 years ago
*e++ = 57; /* Maximum DHCP Message Size */
*e++ = 2;
*e++ = (576 - 312 + OPT_FIELD_SIZE) >> 8;
*e++ = (576 - 312 + OPT_FIELD_SIZE) & 0xff;
22 years ago
if (server_ip.s_addr) {
int tmp = ntohl(server_ip.s_addr);
22 years ago
*e++ = 54; /* ServerID */
*e++ = 4;
*e++ = tmp >> 24;
*e++ = tmp >> 16;
*e++ = tmp >> 8;
*e++ = tmp & 0xff;
}
22 years ago
if (requested_ip.s_addr) {
int tmp = ntohl(requested_ip.s_addr);
22 years ago
*e++ = 50; /* Requested IP */
*e++ = 4;
*e++ = tmp >> 24;
*e++ = tmp >> 16;
*e++ = tmp >> 8;
*e++ = tmp & 0xff;
}
#if defined(CONFIG_BOOTP_SEND_HOSTNAME)
hostname = env_get("hostname");
if (hostname) {
int hostnamelen = strlen(hostname);
*e++ = 12; /* Hostname */
*e++ = hostnamelen;
memcpy(e, hostname, hostnamelen);
e += hostnamelen;
}
#endif
#ifdef CONFIG_BOOTP_PXE_CLIENTARCH
clientarch = CONFIG_BOOTP_PXE_CLIENTARCH;
#endif
if (env_get("bootp_arch"))
clientarch = env_get_ulong("bootp_arch", 16, clientarch);
if (clientarch > 0) {
*e++ = 93; /* Client System Architecture */
*e++ = 2;
*e++ = (clientarch >> 8) & 0xff;
*e++ = clientarch & 0xff;
}
*e++ = 94; /* Client Network Interface Identifier */
*e++ = 3;
*e++ = 1; /* type field for UNDI */
*e++ = 0; /* major revision */
*e++ = 0; /* minor revision */
#ifdef CONFIG_LIB_UUID
uuid = env_get("pxeuuid");
if (uuid) {
if (uuid_str_valid(uuid)) {
*e++ = 97; /* Client Machine Identifier */
*e++ = 17;
*e++ = 0; /* type 0 - UUID */
uuid_str_to_bin(uuid, e, UUID_STR_FORMAT_STD);
e += 16;
} else {
printf("Invalid pxeuuid: %s\n", uuid);
}
}
#endif
e = add_vci(e);
#if defined(CONFIG_BOOTP_VENDOREX)
x = dhcp_vendorex_prep(e);
if (x)
return x - start;
22 years ago
#endif
*e++ = 55; /* Parameter Request List */
cnt = e++; /* Pointer to count of requested items */
*cnt = 0;
#if defined(CONFIG_BOOTP_SUBNETMASK)
*e++ = 1; /* Subnet Mask */
*cnt += 1;
22 years ago
#endif
#if defined(CONFIG_BOOTP_TIMEOFFSET)
*e++ = 2;
*cnt += 1;
#endif
#if defined(CONFIG_BOOTP_GATEWAY)
*e++ = 3; /* Router Option */
*cnt += 1;
22 years ago
#endif
#if defined(CONFIG_BOOTP_DNS)
*e++ = 6; /* DNS Server(s) */
*cnt += 1;
22 years ago
#endif
#if defined(CONFIG_BOOTP_HOSTNAME)
*e++ = 12; /* Hostname */
*cnt += 1;
22 years ago
#endif
#if defined(CONFIG_BOOTP_BOOTFILESIZE)
*e++ = 13; /* Boot File Size */
*cnt += 1;
22 years ago
#endif
#if defined(CONFIG_BOOTP_BOOTPATH)
*e++ = 17; /* Boot path */
*cnt += 1;
22 years ago
#endif
#if defined(CONFIG_BOOTP_NISDOMAIN)
*e++ = 40; /* NIS Domain name request */
*cnt += 1;
22 years ago
#endif
#if defined(CONFIG_BOOTP_NTPSERVER)
*e++ = 42;
*cnt += 1;
#endif
/* no options, so back up to avoid sending an empty request list */
if (*cnt == 0)
e -= 2;
*e++ = 255; /* End of the list */
22 years ago
/* Pad to minimal length */
22 years ago
#ifdef CONFIG_DHCP_MIN_EXT_LEN
while ((e - start) < CONFIG_DHCP_MIN_EXT_LEN)
*e++ = 0;
22 years ago
#endif
return e - start;
22 years ago
}
#else
22 years ago
/*
* Warning: no field size check - change CONFIG_BOOTP_* at your own risk!
22 years ago
*/
static int bootp_extended(u8 *e)
22 years ago
{
u8 *start = e;
22 years ago
*e++ = 99; /* RFC1048 Magic Cookie */
*e++ = 130;
*e++ = 83;
*e++ = 99;
22 years ago
#if defined(CONFIG_CMD_DHCP)
*e++ = 53; /* DHCP Message Type */
*e++ = 1;
*e++ = DHCP_DISCOVER;
*e++ = 57; /* Maximum DHCP Message Size */
*e++ = 2;
*e++ = (576 - 312 + OPT_FIELD_SIZE) >> 16;
*e++ = (576 - 312 + OPT_FIELD_SIZE) & 0xff;
#endif
22 years ago
add_vci(e);
#if defined(CONFIG_BOOTP_SUBNETMASK)
*e++ = 1; /* Subnet mask request */
*e++ = 4;
e += 4;
22 years ago
#endif
#if defined(CONFIG_BOOTP_GATEWAY)
*e++ = 3; /* Default gateway request */
*e++ = 4;
e += 4;
22 years ago
#endif
#if defined(CONFIG_BOOTP_DNS)
*e++ = 6; /* Domain Name Server */
*e++ = 4;
e += 4;
22 years ago
#endif
#if defined(CONFIG_BOOTP_HOSTNAME)
*e++ = 12; /* Host name request */
*e++ = 32;
e += 32;
22 years ago
#endif
#if defined(CONFIG_BOOTP_BOOTFILESIZE)
*e++ = 13; /* Boot file size */
*e++ = 2;
e += 2;
22 years ago
#endif
#if defined(CONFIG_BOOTP_BOOTPATH)
*e++ = 17; /* Boot path */
*e++ = 32;
e += 32;
22 years ago
#endif
#if defined(CONFIG_BOOTP_NISDOMAIN)
*e++ = 40; /* NIS Domain name request */
*e++ = 32;
e += 32;
22 years ago
#endif
#if defined(CONFIG_BOOTP_NTPSERVER)
*e++ = 42;
*e++ = 4;
e += 4;
#endif
22 years ago
*e++ = 255; /* End of the list */
22 years ago
/*
* If nothing in list, remove it altogether. Some DHCP servers get
* upset by this minor faux pas and do not respond at all.
*/
if (e == start + 3) {
printf("*** Warning: no DHCP options requested\n");
e -= 3;
}
return e - start;
22 years ago
}
#endif
22 years ago
void bootp_reset(void)
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
{
net: More BOOTP retry timeout improvements It's not unusual for DHCP servers to take a couple hundred milliseconds to respond to DHCP discover messages. One possible reason for the delay can be that the server checks (typically using an ARP request) that the IP it's about to hand out isn't in use yet. To make matters worse, some servers may also queue up requests and process them sequentially, which can cause excessively long delays if clients retry too fast. Commit f59be6e850b3 ("net: BOOTP retry timeout improvements") shortened the retry timeouts significantly, but the BOOTP/DHCP implementation in U-Boot doesn't handle that well because it will ignore incoming replies to earlier requests. In one particular setup this increases the time it takes to obtain a DHCP lease from 630 ms to 8313 ms. This commit attempts to fix this in two ways. First it increases the initial retry timeout from 10 ms to 250 ms to give DHCP servers some more time to respond. At the same time a cache of outstanding DHCP request IDs is kept so that the implementation will know to continue transactions even after a retransmission of the DISCOVER message. The maximum retry timeout is also increased from 1 second to 2 seconds. An ID cache of size 4 will keep DHCP requests around for 8 seconds (once the maximum retry timeout has been reached) before dropping them. This should give servers plenty of time to respond. If it ever turns out that this isn't enough, the size of the cache can easily be increased. With this commit the DHCP lease on the above-mentioned setup still takes longer (1230 ms) than originally, but that's an acceptable compromise to improve DHCP lease acquisition time for a broader range of setups. To make it easier to benchmark DHCP in the future, this commit also adds the time it took to obtain a lease to the final "DHCP client bound to address x.x.x.x" message. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years ago
bootp_num_ids = 0;
bootp_try = 0;
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
bootp_start = get_timer(0);
net: More BOOTP retry timeout improvements It's not unusual for DHCP servers to take a couple hundred milliseconds to respond to DHCP discover messages. One possible reason for the delay can be that the server checks (typically using an ARP request) that the IP it's about to hand out isn't in use yet. To make matters worse, some servers may also queue up requests and process them sequentially, which can cause excessively long delays if clients retry too fast. Commit f59be6e850b3 ("net: BOOTP retry timeout improvements") shortened the retry timeouts significantly, but the BOOTP/DHCP implementation in U-Boot doesn't handle that well because it will ignore incoming replies to earlier requests. In one particular setup this increases the time it takes to obtain a DHCP lease from 630 ms to 8313 ms. This commit attempts to fix this in two ways. First it increases the initial retry timeout from 10 ms to 250 ms to give DHCP servers some more time to respond. At the same time a cache of outstanding DHCP request IDs is kept so that the implementation will know to continue transactions even after a retransmission of the DISCOVER message. The maximum retry timeout is also increased from 1 second to 2 seconds. An ID cache of size 4 will keep DHCP requests around for 8 seconds (once the maximum retry timeout has been reached) before dropping them. This should give servers plenty of time to respond. If it ever turns out that this isn't enough, the size of the cache can easily be increased. With this commit the DHCP lease on the above-mentioned setup still takes longer (1230 ms) than originally, but that's an acceptable compromise to improve DHCP lease acquisition time for a broader range of setups. To make it easier to benchmark DHCP in the future, this commit also adds the time it took to obtain a lease to the final "DHCP client bound to address x.x.x.x" message. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years ago
bootp_timeout = 250;
net: BOOTP retry timeout improvements Currently, the BOOTP code sends out its initial request as soon as the Ethernet driver indicates "link up". If this packet is lost or not replied to for some reason, the code waits for a 1s timeout before retrying. For some reason, such early packets are often lost on my system, so this causes an annoying delay. To optimize this, modify the BOOTP code to have very short timeouts for the first packet transmitted, but gradually increase the timeout each time a timeout occurs. This way, if the first packet is lost, the second packet is transmitted quite quickly and hence the overall delay is low. However, if there's still no response, we don't keep spewing out packets at an insane speed. It's arguably more correct to try and find out why the first packet is lost. However, it seems to disappear inside my Ethenet chip; the TX chip indicates no error during TX (not that it has much in the way of reporting...), yet wireshark on the RX side doesn't see any packet. FWIW, I'm using an ASIX USB Ethernet adapter. Perhaps "link up" is reported too early or based on the wrong condition in HW, and we should add some fixed extra delay into the driver. However, this would slow down every link up event even if it ends up not being needed in some cases. Having BOOTP retry quickly applies the fix/WAR to every possible Ethernet device, and is quite simple to implement, so seems a better solution. Signed-off-by: Stephen Warren <swarren@nvidia.com> Acked-by: Joe Hershberger <joe.hershberger@ni.com>
10 years ago
}
void bootp_request(void)
22 years ago
{
uchar *pkt, *iphdr;
struct bootp_hdr *bp;
int extlen, pktlen, iplen;
int eth_hdr_size;
#ifdef CONFIG_BOOTP_RANDOM_DELAY
ulong rand_ms;
#endif
u32 bootp_id;
struct in_addr zero_ip;
struct in_addr bcast_ip;
char *ep; /* Environment pointer */
22 years ago
bootstage_mark_name(BOOTSTAGE_ID_BOOTP_START, "bootp_start");
#if defined(CONFIG_CMD_DHCP)
22 years ago
dhcp_state = INIT;
#endif
ep = env_get("bootpretryperiod");
if (ep != NULL)
time_taken_max = simple_strtoul(ep, NULL, 10);
else
time_taken_max = TIMEOUT_MS;
22 years ago
#ifdef CONFIG_BOOTP_RANDOM_DELAY /* Random BOOTP delay */
if (bootp_try == 0)
srand_mac();
22 years ago
if (bootp_try <= 2) /* Start with max 1024 * 1ms */
rand_ms = rand() >> (22 - bootp_try);
else /* After 3rd BOOTP request max 8192 * 1ms */
rand_ms = rand() >> 19;
22 years ago
printf("Random delay: %ld ms...\n", rand_ms);
mdelay(rand_ms);
22 years ago
#endif /* CONFIG_BOOTP_RANDOM_DELAY */
printf("BOOTP broadcast %d\n", ++bootp_try);
pkt = net_tx_packet;
memset((void *)pkt, 0, PKTSIZE);
22 years ago
eth_hdr_size = net_set_ether(pkt, net_bcast_ethaddr, PROT_IP);
pkt += eth_hdr_size;
22 years ago
/*
* Next line results in incorrect packet size being transmitted,
* resulting in errors in some DHCP servers, reporting missing bytes.
* Size must be set in packet header after extension length has been
* determined.
22 years ago
* C. Hallinan, DS4.COM, Inc.
*/
/* net_set_udp_header(pkt, 0xFFFFFFFFL, PORT_BOOTPS, PORT_BOOTPC,
sizeof (struct bootp_hdr)); */
iphdr = pkt; /* We need this later for net_set_udp_header() */
pkt += IP_UDP_HDR_SIZE;
22 years ago
bp = (struct bootp_hdr *)pkt;
22 years ago
bp->bp_op = OP_BOOTREQUEST;
bp->bp_htype = HWT_ETHER;
bp->bp_hlen = HWL_ETHER;
bp->bp_hops = 0;
/*
* according to RFC1542, should be 0 on first request, secs since
* first request otherwise
*/
bp->bp_secs = htons(get_timer(bootp_start) / 1000);
zero_ip.s_addr = 0;
net_write_ip(&bp->bp_ciaddr, zero_ip);
net_write_ip(&bp->bp_yiaddr, zero_ip);
net_write_ip(&bp->bp_siaddr, zero_ip);
net_write_ip(&bp->bp_giaddr, zero_ip);
memcpy(bp->bp_chaddr, net_ethaddr, 6);
copy_filename(bp->bp_file, net_boot_file_name, sizeof(bp->bp_file));
22 years ago
/* Request additional information from the BOOTP/DHCP server */
#if defined(CONFIG_CMD_DHCP)
extlen = dhcp_extended((u8 *)bp->bp_vend, DHCP_DISCOVER, zero_ip,
zero_ip);
22 years ago
#else
extlen = bootp_extended((u8 *)bp->bp_vend);
#endif
22 years ago
/*
* Bootp ID is the lower 4 bytes of our ethernet address
* plus the current time in ms.
22 years ago
*/
bootp_id = ((u32)net_ethaddr[2] << 24)
| ((u32)net_ethaddr[3] << 16)
| ((u32)net_ethaddr[4] << 8)
| (u32)net_ethaddr[5];
bootp_id += get_timer(0);
bootp_id = htonl(bootp_id);
bootp_add_id(bootp_id);
net_copy_u32(&bp->bp_id, &bootp_id);
22 years ago
/*
* Calculate proper packet lengths taking into account the
* variable size of the options field
*/
iplen = BOOTP_HDR_SIZE - OPT_FIELD_SIZE + extlen;
pktlen = eth_hdr_size + IP_UDP_HDR_SIZE + iplen;
bcast_ip.s_addr = 0xFFFFFFFFL;
net_set_udp_header(iphdr, bcast_ip, PORT_BOOTPS, PORT_BOOTPC, iplen);
net_set_timeout_handler(bootp_timeout, bootp_timeout_handler);
22 years ago
#if defined(CONFIG_CMD_DHCP)
22 years ago
dhcp_state = SELECTING;
net_set_udp_handler(dhcp_handler);
22 years ago
#else
net_set_udp_handler(bootp_handler);
#endif
net_send_packet(net_tx_packet, pktlen);
22 years ago
}
#if defined(CONFIG_CMD_DHCP)
static void dhcp_process_options(uchar *popt, uchar *end)
22 years ago
{
int oplen, size;
#if defined(CONFIG_CMD_SNTP) && defined(CONFIG_BOOTP_TIMEOFFSET)
int *to_ptr;
#endif
22 years ago
while (popt < end && *popt != 0xff) {
22 years ago
oplen = *(popt + 1);
switch (*popt) {
case 0:
oplen = -1; /* Pad omits len byte */
break;
case 1:
net_copy_ip(&net_netmask, (popt + 2));
break;
#if defined(CONFIG_CMD_SNTP) && defined(CONFIG_BOOTP_TIMEOFFSET)
case 2: /* Time offset */
to_ptr = &net_ntp_time_offset;
net_copy_u32((u32 *)to_ptr, (u32 *)(popt + 2));
net_ntp_time_offset = ntohl(net_ntp_time_offset);
break;
#endif
case 3:
net_copy_ip(&net_gateway, (popt + 2));
break;
case 6:
net_copy_ip(&net_dns_server, (popt + 2));
#if defined(CONFIG_BOOTP_DNS2)
if (*(popt + 1) > 4)
net_copy_ip(&net_dns_server2, (popt + 2 + 4));
#endif
break;
case 12:
size = truncate_sz("Host Name",
sizeof(net_hostname), oplen);
memcpy(&net_hostname, popt + 2, size);
net_hostname[size] = 0;
break;
case 15: /* Ignore Domain Name Option */
break;
case 17:
size = truncate_sz("Root Path",
sizeof(net_root_path), oplen);
memcpy(&net_root_path, popt + 2, size);
net_root_path[size] = 0;
break;
case 28: /* Ignore Broadcast Address Option */
break;
#if defined(CONFIG_CMD_SNTP) && defined(CONFIG_BOOTP_NTPSERVER)
case 42: /* NTP server IP */
net_copy_ip(&net_ntp_server, (popt + 2));
break;
#endif
case 51:
net_copy_u32(&dhcp_leasetime, (u32 *)(popt + 2));
break;
case 52:
dhcp_option_overload = popt[2];
break;
case 53: /* Ignore Message Type Option */
break;
case 54:
net_copy_ip(&dhcp_server_ip, (popt + 2));
break;
case 58: /* Ignore Renewal Time Option */
break;
case 59: /* Ignore Rebinding Time Option */
break;
case 66: /* Ignore TFTP server name */
break;
case 67: /* Bootfile option */
if (!net_boot_file_name_explicit) {
size = truncate_sz("Bootfile",
sizeof(net_boot_file_name),
oplen);
memcpy(&net_boot_file_name, popt + 2, size);
net_boot_file_name[size] = 0;
}
break;
default:
#if defined(CONFIG_BOOTP_VENDOREX)
if (dhcp_vendorex_proc(popt))
break;
22 years ago
#endif
printf("*** Unhandled DHCP Option in OFFER/ACK:"
" %d\n", *popt);
break;
22 years ago
}
popt += oplen + 2; /* Process next option */
}
}
static void dhcp_packet_process_options(struct bootp_hdr *bp)
{
uchar *popt = (uchar *)&bp->bp_vend[4];
uchar *end = popt + BOOTP_HDR_SIZE;
if (net_read_u32((u32 *)&bp->bp_vend[0]) != htonl(BOOTP_VENDOR_MAGIC))
return;
dhcp_option_overload = 0;
/*
* The 'options' field MUST be interpreted first, 'file' next,
* 'sname' last.
*/
dhcp_process_options(popt, end);
if (dhcp_option_overload & OVERLOAD_FILE) {
popt = (uchar *)bp->bp_file;
end = popt + sizeof(bp->bp_file);
dhcp_process_options(popt, end);
}
if (dhcp_option_overload & OVERLOAD_SNAME) {
popt = (uchar *)bp->bp_sname;
end = popt + sizeof(bp->bp_sname);
dhcp_process_options(popt, end);
}
}
static int dhcp_message_type(unsigned char *popt)
22 years ago
{
if (net_read_u32((u32 *)popt) != htonl(BOOTP_VENDOR_MAGIC))
22 years ago
return -1;
popt += 4;
while (*popt != 0xff) {
if (*popt == 53) /* DHCP Message Type */
22 years ago
return *(popt + 2);
if (*popt == 0) {
/* Pad */
popt += 1;
} else {
/* Scan through all options */
popt += *(popt + 1) + 2;
}
22 years ago
}
return -1;
}
static void dhcp_send_request_packet(struct bootp_hdr *bp_offer)
22 years ago
{
uchar *pkt, *iphdr;
struct bootp_hdr *bp;
22 years ago
int pktlen, iplen, extlen;
int eth_hdr_size;
struct in_addr offered_ip;
struct in_addr zero_ip;
struct in_addr bcast_ip;
22 years ago
debug("dhcp_send_request_packet: Sending DHCPREQUEST\n");
pkt = net_tx_packet;
memset((void *)pkt, 0, PKTSIZE);
22 years ago
eth_hdr_size = net_set_ether(pkt, net_bcast_ethaddr, PROT_IP);
pkt += eth_hdr_size;
22 years ago
iphdr = pkt; /* We'll need this later to set proper pkt size */
pkt += IP_UDP_HDR_SIZE;
22 years ago
bp = (struct bootp_hdr *)pkt;
22 years ago
bp->bp_op = OP_BOOTREQUEST;
bp->bp_htype = HWT_ETHER;
bp->bp_hlen = HWL_ETHER;
bp->bp_hops = 0;
bp->bp_secs = htons(get_timer(bootp_start) / 1000);
/* Do not set the client IP, your IP, or server IP yet, since it
* hasn't been ACK'ed by the server yet */
/*
* RFC3046 requires Relay Agents to discard packets with
* nonzero and offered giaddr
*/
zero_ip.s_addr = 0;
net_write_ip(&bp->bp_giaddr, zero_ip);
memcpy(bp->bp_chaddr, net_ethaddr, 6);
copy_filename(bp->bp_file, net_boot_file_name, sizeof(bp->bp_file));
22 years ago
/*
* ID is the id of the OFFER packet
*/
net_copy_u32(&bp->bp_id, &bp_offer->bp_id);
22 years ago
/*
* Copy options from OFFER packet if present
*/
/* Copy offered IP into the parameters request list */
net_copy_ip(&offered_ip, &bp_offer->bp_yiaddr);
extlen = dhcp_extended((u8 *)bp->bp_vend, DHCP_REQUEST,
dhcp_server_ip, offered_ip);
22 years ago
iplen = BOOTP_HDR_SIZE - OPT_FIELD_SIZE + extlen;
pktlen = eth_hdr_size + IP_UDP_HDR_SIZE + iplen;
bcast_ip.s_addr = 0xFFFFFFFFL;
net_set_udp_header(iphdr, bcast_ip, PORT_BOOTPS, PORT_BOOTPC, iplen);
22 years ago
#ifdef CONFIG_BOOTP_DHCP_REQUEST_DELAY
udelay(CONFIG_BOOTP_DHCP_REQUEST_DELAY);
#endif /* CONFIG_BOOTP_DHCP_REQUEST_DELAY */
debug("Transmitting DHCPREQUEST packet: len = %d\n", pktlen);
net_send_packet(net_tx_packet, pktlen);
22 years ago
}
/*
* Handle DHCP received packets.
*/
static void dhcp_handler(uchar *pkt, unsigned dest, struct in_addr sip,
unsigned src, unsigned len)
22 years ago
{
struct bootp_hdr *bp = (struct bootp_hdr *)pkt;
22 years ago
debug("DHCPHandler: got packet: (src=%d, dst=%d, len=%d) state: %d\n",
src, dest, len, dhcp_state);
22 years ago
/* Filter out pkts we don't want */
if (check_reply_packet(pkt, dest, src, len))
22 years ago
return;
debug("DHCPHandler: got DHCP packet: (src=%d, dst=%d, len=%d) state: "
"%d\n", src, dest, len, dhcp_state);
22 years ago
if (net_read_ip(&bp->bp_yiaddr).s_addr == 0)
return;
22 years ago
switch (dhcp_state) {
case SELECTING:
/*
* Wait an appropriate time for any potential DHCPOFFER packets
* to arrive. Then select one, and generate DHCPREQUEST
* response. If filename is in format we recognize, assume it
* is a valid OFFER from a server we want.
22 years ago
*/
debug("DHCP: state=SELECTING bp_file: \"%s\"\n", bp->bp_file);
#ifdef CONFIG_SYS_BOOTFILE_PREFIX
22 years ago
if (strncmp(bp->bp_file,
CONFIG_SYS_BOOTFILE_PREFIX,
strlen(CONFIG_SYS_BOOTFILE_PREFIX)) == 0) {
#endif /* CONFIG_SYS_BOOTFILE_PREFIX */
dhcp_packet_process_options(bp);
efi_net_set_dhcp_ack(pkt, len);
22 years ago
debug("TRANSITIONING TO REQUESTING STATE\n");
22 years ago
dhcp_state = REQUESTING;
net_set_timeout_handler(5000, bootp_timeout_handler);
dhcp_send_request_packet(bp);
#ifdef CONFIG_SYS_BOOTFILE_PREFIX
22 years ago
}
#endif /* CONFIG_SYS_BOOTFILE_PREFIX */
22 years ago
return;
break;
case REQUESTING:
debug("DHCP State: REQUESTING\n");
22 years ago
if (dhcp_message_type((u8 *)bp->bp_vend) == DHCP_ACK) {
dhcp_packet_process_options(bp);
/* Store net params from reply */
store_net_params(bp);
22 years ago
dhcp_state = BOUND;
net: More BOOTP retry timeout improvements It's not unusual for DHCP servers to take a couple hundred milliseconds to respond to DHCP discover messages. One possible reason for the delay can be that the server checks (typically using an ARP request) that the IP it's about to hand out isn't in use yet. To make matters worse, some servers may also queue up requests and process them sequentially, which can cause excessively long delays if clients retry too fast. Commit f59be6e850b3 ("net: BOOTP retry timeout improvements") shortened the retry timeouts significantly, but the BOOTP/DHCP implementation in U-Boot doesn't handle that well because it will ignore incoming replies to earlier requests. In one particular setup this increases the time it takes to obtain a DHCP lease from 630 ms to 8313 ms. This commit attempts to fix this in two ways. First it increases the initial retry timeout from 10 ms to 250 ms to give DHCP servers some more time to respond. At the same time a cache of outstanding DHCP request IDs is kept so that the implementation will know to continue transactions even after a retransmission of the DISCOVER message. The maximum retry timeout is also increased from 1 second to 2 seconds. An ID cache of size 4 will keep DHCP requests around for 8 seconds (once the maximum retry timeout has been reached) before dropping them. This should give servers plenty of time to respond. If it ever turns out that this isn't enough, the size of the cache can easily be increased. With this commit the DHCP lease on the above-mentioned setup still takes longer (1230 ms) than originally, but that's an acceptable compromise to improve DHCP lease acquisition time for a broader range of setups. To make it easier to benchmark DHCP in the future, this commit also adds the time it took to obtain a lease to the final "DHCP client bound to address x.x.x.x" message. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years ago
printf("DHCP client bound to address %pI4 (%lu ms)\n",
&net_ip, get_timer(bootp_start));
net_set_timeout_handler(0, (thand_f *)0);
bootstage_mark_name(BOOTSTAGE_ID_BOOTP_STOP,
"bootp_stop");
22 years ago
net_auto_load();
22 years ago
return;
}
break;
case BOUND:
/* DHCP client bound to address */
break;
22 years ago
default:
puts("DHCP: INVALID STATE\n");
22 years ago
break;
}
}
void dhcp_request(void)
22 years ago
{
bootp_request();
22 years ago
}
#endif /* CONFIG_CMD_DHCP */