[PATCH v3 0/3] BOOTP/DHCPv4 enhancements
From: Sean Edmond seanedmond@microsoft.com
In our datacenter application, a single DHCP server is servicing 36000+ clients. Improvements are required to the DHCPv4 retransmission behavior to align with RFC and ensure less pressure is exerted on the server: - retransmission backoff interval maximum is configurable (environment variable bootpretransmitperiodmax) - initial retransmission backoff interval is configurable (environment variable bootpretransmitperiodinit) - transaction ID is kept the same for each BOOTP/DHCPv4 request (not recreated on each retry)
For our application we'll use: - bootpretransmitperiodmax=16000 - bootpretransmitperiodinit=2000
A new configuration BOOTP_RANDOM_XID has been added to enable a randomized BOOTP/DHCPv4 transaction ID.
Add functionality for DHCPv4 sending/parsing option 209 (PXE config file). Enabled with Kconfig BOOTP_PXE_DHCP_OPTION. Note, this patch was submitted previously but this latest version has been enhanced to avoid a possible double free().
changes in v3: - add define for option 209 and rfc5071 reference - Set RETRANSMIT_PERIOD_MAX_MS=60000 - Add randomization factor to retransmission timeout - Add depends for BOOTP_RANDOM_XID
changes in v2: - use env_get_ulong() to get environment variables
Sean Edmond (3): net: Get pxe config file from dhcp option 209 net: bootp: BOOTP/DHCPv4 retransmission improvements net: bootp: add config option BOOTP_RANDOM_XID
Sean Edmond (3): net: Get pxe config file from dhcp option 209 net: bootp: BOOTP/DHCPv4 retransmission improvements net: bootp: add config option BOOTP_RANDOM_XID
cmd/Kconfig | 11 +++++++ cmd/pxe.c | 10 ++++++ net/bootp.c | 93 ++++++++++++++++++++++++++++++++++++++++++++--------- net/bootp.h | 2 ++ 4 files changed, 101 insertions(+), 15 deletions(-)
From: Sean Edmond seanedmond@microsoft.com
Allow dhcp server pass pxe config file full path by using option 209 as specified in RFC5071.
Signed-off-by: Sean Edmond seanedmond@microsoft.com --- cmd/Kconfig | 4 ++++ cmd/pxe.c | 10 ++++++++++ net/bootp.c | 21 +++++++++++++++++++++ net/bootp.h | 2 ++ 4 files changed, 37 insertions(+)
diff --git a/cmd/Kconfig b/cmd/Kconfig index fd16c3a48e5..df6d71c103f 100644 --- a/cmd/Kconfig +++ b/cmd/Kconfig @@ -1840,6 +1840,10 @@ config BOOTP_PXE_CLIENTARCH default 0x15 if ARM default 0x0 if X86
+config BOOTP_PXE_DHCP_OPTION + bool "Request & store 'pxe_configfile' from BOOTP/DHCP server" + depends on BOOTP_PXE + config BOOTP_VCI_STRING string depends on CMD_BOOTP diff --git a/cmd/pxe.c b/cmd/pxe.c index 704589702f2..9404f445187 100644 --- a/cmd/pxe.c +++ b/cmd/pxe.c @@ -65,6 +65,8 @@ static int pxe_dhcp_option_path(struct pxe_context *ctx, unsigned long pxefile_a int ret = get_pxe_file(ctx, pxelinux_configfile, pxefile_addr_r);
free(pxelinux_configfile); + /* set to NULL to avoid double-free if DHCP is tried again */ + pxelinux_configfile = NULL;
return ret; } @@ -141,6 +143,14 @@ int pxe_get(ulong pxefile_addr_r, char **bootdirp, ulong *sizep, bool use_ipv6) env_get("bootfile"), use_ipv6)) return -ENOMEM;
+ if (IS_ENABLED(CONFIG_BOOTP_PXE_DHCP_OPTION) && + pxelinux_configfile && !use_ipv6) { + if (pxe_dhcp_option_path(&ctx, pxefile_addr_r) > 0) + goto done; + + goto error_exit; + } + if (IS_ENABLED(CONFIG_DHCP6_PXE_DHCP_OPTION) && pxelinux_configfile && use_ipv6) { if (pxe_dhcp_option_path(&ctx, pxefile_addr_r) > 0) diff --git a/net/bootp.c b/net/bootp.c index 7b0f45e18a9..24547cf7bfd 100644 --- a/net/bootp.c +++ b/net/bootp.c @@ -26,6 +26,7 @@ #ifdef CONFIG_BOOTP_RANDOM_DELAY #include "net_rand.h" #endif +#include <malloc.h>
#define BOOTP_VENDOR_MAGIC 0x63825363 /* RFC1048 Magic Cookie */
@@ -601,6 +602,10 @@ static int dhcp_extended(u8 *e, int message_type, struct in_addr server_ip, *e++ = 42; *cnt += 1; #endif + if (IS_ENABLED(CONFIG_BOOTP_PXE_DHCP_OPTION)) { + *e++ = DHCP_OPTION_PXE_CONFIG_FILE; /* PXELINUX Config File */ + *cnt += 1; + } /* no options, so back up to avoid sending an empty request list */ if (*cnt == 0) e -= 2; @@ -909,6 +914,22 @@ static void dhcp_process_options(uchar *popt, uchar *end) net_boot_file_name[size] = 0; } break; + case DHCP_OPTION_PXE_CONFIG_FILE: /* PXELINUX Config File */ + if (IS_ENABLED(CONFIG_BOOTP_PXE_DHCP_OPTION)) { + /* In case it has already been allocated when get DHCP Offer packet, + * free first to avoid memory leak. + */ + if (pxelinux_configfile) + free(pxelinux_configfile); + + pxelinux_configfile = (char *)malloc((oplen + 1) * sizeof(char)); + + if (pxelinux_configfile) + strlcpy(pxelinux_configfile, popt + 2, oplen + 1); + else + printf("Error: Failed to allocate pxelinux_configfile\n"); + } + break; default: #if defined(CONFIG_BOOTP_VENDOREX) if (dhcp_vendorex_proc(popt)) diff --git a/net/bootp.h b/net/bootp.h index 4e32b19d424..24b32c73f62 100644 --- a/net/bootp.h +++ b/net/bootp.h @@ -91,6 +91,8 @@ typedef enum { INIT, #define DHCP_NAK 6 #define DHCP_RELEASE 7
+#define DHCP_OPTION_PXE_CONFIG_FILE 209 /* "ConfigFile" option according to rfc5071 */ + /**********************************************************************/
#endif /* __BOOTP_H__ */
From: Sean Edmond seanedmond@microsoft.com
This patch introduces 3 improvements to align with RFC 951: - retransmission backoff interval maximum is configurable - initial retranmission backoff interval is configurable - transaction ID is kept the same for each BOOTP/DHCPv4 request
In applications where thousands of nodes are serviced by a single DHCP server, maximizing the retransmission backoff interval at 2 seconds (the current u-boot default) exerts high pressure on the DHCP server and network layer.
RFC 951 “7.2. Client Retransmission Strategy” states that the retransmission backoff interval should be limited to 60 seconds. This patch allows the interval to be configurable using the environment variable "bootpretransmitperiodmax"
The initial retranmission backoff period defaults to 250ms, which is also too small for these scenarios with many clients. This patch makes the initial retransmission interval to be configurable using the environment variable "bootpretransmitperiodinit".
Also, on a retransmission it is not expected for the transaction ID to change (only the 'secs' field should be updated). Let's save the transaction ID and use the same transaction ID for each BOOTP/DHCPv4 exchange.
Signed-off-by: Sean Edmond seanedmond@microsoft.com --- net/bootp.c | 71 +++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 55 insertions(+), 16 deletions(-)
diff --git a/net/bootp.c b/net/bootp.c index 24547cf7bfd..9e4f37239da 100644 --- a/net/bootp.c +++ b/net/bootp.c @@ -42,6 +42,22 @@ */ #define TIMEOUT_MS ((3 + (CONFIG_NET_RETRY_COUNT * 5)) * 1000)
+/* + * According to rfc951 : 7.2. Client Retransmission Strategy + * "After the 'average' backoff reaches about 60 seconds, it should be + * increased no further, but still randomized." + * + * U-Boot has saturated this backoff at 2 seconds for a long time. + * To modify, set the environment variable "bootpretransmitperiodmax" + */ +#define RETRANSMIT_PERIOD_MAX_MS 60000 + +/* Retransmission timeout for the initial packet (in milliseconds). + * This timeout will double on each retry. To modify, set the + * environment variable bootpretransmitperiodinit. + */ +#define RETRANSMIT_PERIOD_INIT_MS 250 + #ifndef CFG_DHCP_MIN_EXT_LEN /* minimal length of extension list */ #define CFG_DHCP_MIN_EXT_LEN 64 #endif @@ -53,6 +69,7 @@ u32 bootp_ids[CFG_BOOTP_ID_CACHE_SIZE]; unsigned int bootp_num_ids; int bootp_try; +u32 bootp_id; ulong bootp_start; ulong bootp_timeout; char net_nis_domain[32] = {0,}; /* Our NIS domain */ @@ -60,6 +77,7 @@ char net_hostname[32] = {0,}; /* Our hostname */ char net_root_path[CONFIG_BOOTP_MAX_ROOT_PATH_LEN] = {0,}; /* Our bootpath */
static ulong time_taken_max; +static u32 retransmit_period_max_ms;
#if defined(CONFIG_CMD_DHCP) static dhcp_state_t dhcp_state = INIT; @@ -396,6 +414,7 @@ static void bootp_handler(uchar *pkt, unsigned dest, struct in_addr sip, static void bootp_timeout_handler(void) { ulong time_taken = get_timer(bootp_start); + int rand_minus_plus_100;
if (time_taken >= time_taken_max) { #ifdef CONFIG_BOOTP_MAY_FAIL @@ -414,8 +433,17 @@ static void bootp_timeout_handler(void) } } else { bootp_timeout *= 2; - if (bootp_timeout > 2000) - bootp_timeout = 2000; + if (bootp_timeout > retransmit_period_max_ms) + bootp_timeout = retransmit_period_max_ms; + + /* Randomize by adding bootp_timeout*RAND, where RAND + * is a randomization factor between -0.1..+0.1 + */ + srand(get_ticks() + rand()); + rand_minus_plus_100 = ((rand() % 200) - 100); + bootp_timeout = bootp_timeout + + ((bootp_timeout * rand_minus_plus_100) / 1000); + net_set_timeout_handler(bootp_timeout, bootp_timeout_handler); bootp_request(); } @@ -711,10 +739,14 @@ static int bootp_extended(u8 *e)
void bootp_reset(void) { + char *ep; /* Environment pointer */ + bootp_num_ids = 0; bootp_try = 0; bootp_start = get_timer(0); - bootp_timeout = 250; + + bootp_timeout = env_get_ulong("bootpretransmitperiodinit", 10, RETRANSMIT_PERIOD_INIT_MS); + }
void bootp_request(void) @@ -726,7 +758,6 @@ void bootp_request(void) #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 */ @@ -742,6 +773,9 @@ void bootp_request(void) else time_taken_max = TIMEOUT_MS;
+ retransmit_period_max_ms = env_get_ulong("bootpretransmitperiodmax", 10, + RETRANSMIT_PERIOD_MAX_MS); + #ifdef CONFIG_BOOTP_RANDOM_DELAY /* Random BOOTP delay */ if (bootp_try == 0) srand_mac(); @@ -801,18 +835,23 @@ void bootp_request(void) extlen = bootp_extended((u8 *)bp->bp_vend); #endif
- /* - * Bootp ID is the lower 4 bytes of our ethernet address - * plus the current time in ms. - */ - 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); + /* Only generate a new transaction ID for each new BOOTP request */ + if (bootp_try == 1) { + /* + * Bootp ID is the lower 4 bytes of our ethernet address + * plus the current time in ms. + */ + 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); + } else { + net_copy_u32(&bp->bp_id, &bootp_id); + }
/* * Calculate proper packet lengths taking into account the
From: Sean Edmond seanedmond@microsoft.com
The new config option BOOTP_RANDOM_XID will randomize the transaction ID for each new BOOT/DHCPv4 exchange.
Signed-off-by: Sean Edmond seanedmond@microsoft.com --- cmd/Kconfig | 7 +++++++ net/bootp.c | 31 +++++++++++++++++-------------- 2 files changed, 24 insertions(+), 14 deletions(-)
diff --git a/cmd/Kconfig b/cmd/Kconfig index df6d71c103f..ba3c7a34806 100644 --- a/cmd/Kconfig +++ b/cmd/Kconfig @@ -1852,6 +1852,13 @@ config BOOTP_VCI_STRING default "U-Boot.arm" if ARM default "U-Boot"
+config BOOTP_RANDOM_XID + bool "Send random transaction ID to BOOTP/DHCP server" + depends on CMD_BOOTP && (LIB_RAND || EXYNOS_ACE_SHA || RNG_NPCM) + help + Selecting this will allow for a random transaction ID to get + selected for each BOOTP/DHCPv4 exchange. + if CMD_DHCP6
config DHCP6_PXE_CLIENTARCH diff --git a/net/bootp.c b/net/bootp.c index 9e4f37239da..a8cddcb32e0 100644 --- a/net/bootp.c +++ b/net/bootp.c @@ -837,22 +837,25 @@ void bootp_request(void)
/* Only generate a new transaction ID for each new BOOTP request */ if (bootp_try == 1) { - /* - * Bootp ID is the lower 4 bytes of our ethernet address - * plus the current time in ms. - */ - 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); - } else { - net_copy_u32(&bp->bp_id, &bootp_id); + if (IS_ENABLED(CONFIG_BOOTP_RANDOM_XID)) { + srand(get_ticks() + rand()); + bootp_id = rand(); + } else { + /* + * Bootp ID is the lower 4 bytes of our ethernet address + * plus the current time in ms. + */ + 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); /* * Calculate proper packet lengths taking into account the * variable size of the options field
participants (1)
-
seanedmond＠linux.microsoft.com