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 f59be6e850 ("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>
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>
Changes in lib/uuid.c to:
- uuid_str_to_bin()
- uuid_bin_to_str()
New parameter is added to specify input/output string format in listed functions
This change allows easy recognize which UUID type is or should be stored in given
string array. Binary data of UUID and GUID is always stored in big endian, only
string representations are different as follows.
String byte: 0 36
String char: xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
string UUID: be be be be be
string GUID: le le le be be
This patch also updates functions calls and declarations in a whole code.
Signed-off-by: Przemyslaw Marczak <p.marczak@samsung.com>
Cc: Stephen Warren <swarren@nvidia.com>
Cc: Lukasz Majewski <l.majewski@samsung.com>
Cc: trini@ti.com
bootp.c:44:14: warning: symbol 'dhcp_state' was not declared. Should it be static?
bootp.c:45:15: warning: symbol 'dhcp_leasetime' was not declared. Should it be static?
bootp.c:46:10: warning: symbol 'NetDHCPServerIP' was not declared. Should it be static?
arp.c:30:17: warning: symbol 'NetArpWaitReplyIP' was not declared. Should it be static?
arp.c:37:16: warning: symbol 'NetArpTxPacket' was not declared. Should it be static?
arp.c:38:17: warning: symbol 'NetArpPacketBuf' was not declared. Should it be static?
atheros.c:33:19: warning: symbol 'AR8021_driver' was not declared. Should it be static?
net.c:183:7: warning: symbol 'PktBuf' was not declared. Should it be static?
net.c:159:21: warning: symbol 'net_state' was not declared. Should it be static?
ping.c:73:6: warning: symbol 'ping_start' was not declared. Should it be static?
ping.c:82:13: warning: symbol 'ping_receive' was not declared. Should it be static?
tftp.c:53:7: warning: symbol 'TftpRRQTimeoutMSecs' was not declared. Should it be static?
tftp.c:54:5: warning: symbol 'TftpRRQTimeoutCountMax' was not declared. Should it be static?
eth.c:125:19: warning: symbol 'eth_current' was not declared. Should it be static?
Note: in the ping.c fix, commit a36b12f95a
"net: Move PING out of net.c" mistakenly carried the ifdef CMD_PING
clause from when it was necessary to avoid warnings when it was embedded
in net.c.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
This patch adds support for networking in SPL. Some devices are
capable of loading SPL via network so it makes sense to load the
main U-Boot binary via network too. This patch tries to use
existing network code as much as possible. Unfortunately, it depends
on environment which in turn depends on other code so SPL size
is increased significantly. No effort was done to decouple network
code and environment so far.
Signed-off-by: Ilya Yanok <ilya.yanok@cogentembedded.com>
Acked-by: Joe Hershberger <joe.hershberger@ni.com>
Signed-off-by: Tom Rini <trini@ti.com>
Vendor Class Identifier option is common to BOOTP and DHCP and
can be useful without PXE. So send VCI in both BOOTP and DHCP
requests if CONFIG_BOOTP_VCI_STRING is defined.
Signed-off-by: Ilya Yanok <ilya.yanok@cogentembedded.com>
Signed-off-by: Tom Rini <trini@ti.com>
Some DHCP servers (notably dnsmasq) always transmit DHCP Option 28,
Broadcast Address as specified in RFC 2132. Without this patch u-boot
displays the warning:
*** Unhandled DHCP Option in OFFER/ACK: 28
The patch suppresses the warning and ignores DHCP Option 28. There is
no environment variable to set the broadcast address into and if for
some reason u-boot needs the broadcast it can be calculated from
ipaddr and netmask.
Signed-off-by: Brian Rzycki <bmr@freescale.com>
Fix NetSetState function name used with CONFIG_BOOTP_MAY_FAIL.
Signed-off-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com>
Cc: Joe Hershberger <joe.hershberger@gmail.com>
commit "net: use common rand()/srand() functions" introduced the following
build warning on the current u-boot-arm tree:
$ ./MAKEALL MPC8313ERDB_66
Configuring for MPC8313ERDB_66 - Board: MPC8313ERDB, Options: SYS_66MHZ
text data bss dec hex filename
271988 13976 41768 327732 50034 ./u-boot
In file included from bootp.c:15:0:
net_rand.h: In function 'srand_mac':
net_rand.h:40:2: warning: implicit declaration of function 'srand' [-Wimplicit-function-declaration]
adding this dependency fixes it.
Cc: Michael Walle <michael@walle.cc>
Cc: Joe Hershberger <joe.hershberger@ni.com>
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Acked-by: Michael Walle <michael@walle.cc>
This is useful if you want to look for a DHCP server, but try some
other settings if not available.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Acked-by: Simon Glass <sjg@chromium.org>
If CONFIG_BOOTP_SERVERIP is not defined, unused variable warning is
reported. This was fixed upstream using a compiler feature instead
of a simple reorder of the statements.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Acked-by: Simon Glass <sjg@chromium.org>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Call a built-in dummy if none is registered... don't require
protocols to register a handler (eliminating dummies)
NetConsole now uses the ARP handler when waiting on arp
(instead of needing a #define hack in arp.c)
Clear handlers at the end of net loop
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Acked-by: Simon Glass <sjg@chromium.org>
ICMP (ping) was reimplementing IP header code... it now shares code.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Acked-by: Simon Glass <sjg@chromium.org>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Clearer constant name.
Also remove related BOOTP_SIZE which was unused and doesn't take
into account VLAN packets.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Rename IP header related things to IP_UDP. The existing definition
of IP_t includes UDP header, so name it to accurately describe the
structure.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Make the MAC-seeded random number generator available to /net in
general. MAC-seeded rand will be needed by link-local as well, so
give it an interface.
Signed-off-by: Joe Hershberger <joe.hershberger@ni.com>
Acked-by: Simon Glass <sjg@chromium.org>
This inserts bootstage calls into tftp, usb start and bootm. We
could go further, but this is a reasonable start to illustrate
the concept.
Signed-off-by: Simon Glass <sjg@chromium.org>
The rarp code includes another instance of the auto_load logic, so call
what is now net_auto_load() instead.
This also fixes an incorrect call to TftpStart() which was never seen
since apparently no boards enable rarp.
Signed-off-by: Simon Glass <sjg@chromium.org>
Acked-by: Mike Frysinger <vapier@gentoo.org>
These options are required to be present in RFC 4578 compliant DHCP
requests. They give more information to DHCP servers to allow serving
different DHCP responses to different systems based on client
architecture, client capabilities, UUID, or vendor.
Signed-off-by: Jason Hobbs <jason.hobbs@calxeda.com>
Commit 093498669 (Put common autoload code into auto_load() function)
broke handling of autoload environment variable not being set.
The bootp/dhcp code will just keep on requesting IP address forever
and never start TFTP download.
Fix it by moving TftpStart() outside the conditional like it was before.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Tested-by: Fabio Estevam <fabio.estevam@freescale.com>
Acked-by: Simon Glass <sjg@chromium.org>
This is needed for the upcoming TFTP server implementation.
This also simplifies PingHandler() and fixes rxhand_f documentation.
Signed-off-by: Luca Ceresoli <luca.ceresoli@comelit.it>
Cc: Wolfgang Denk <wd@denx.de>
Acked-by: Detlev Zundel <dzu@denx.de>
Can't get IP address with dhcp due to the dhcp server not
allow the empty param list request under some network env
This patch is based on Gray Remlin's initial patch.
Signed-off-by: Jason Liu <r64343@freescale.com>
Signed-off-by: Gray Remlin <g_remlin@rocketmail.com>
Fix warning: bootp.c:695: warning: dereferencing type-punned pointer
will break strict-aliasing rules
Signed-off-by: Wolfgang Denk <wd@denx.de>
Cc: Ben Warren <biggerbadderben@gmail.com>
Minor ./net cleanups - no functional changes
- change #ifdef DEBUG printf(); #endif to just debug()
- changed __FUNCTION__ to __func__
- got rid of extra whitespace between function and opening brace
- removed unnecessary braces on if statements
gcc dead code elimination should make this functionally/size equivalent
when DEBUG is not defined. (confirmed on Blackfin, with gcc 4.3.3).
Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
The net code is mostly consistent in using 'Packet' rather than 'Pkt', so
rename the minor detractor to follow suite.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
The problem is that BOOTP_SIZE uses ETHER_HDR_SIZE which is 14 bytes.
If sending a VLAN tagged frame (when env variable vlan is set) this
should be VLAN_ETHER_HDR_SIZE=18 which is what NetSetEther returns.
Signed-off-by: Norbert van Bolhuis <nvbolhuis@aimvalley.nl>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
Now that our printf functions support the %pI4 modifier like the kernel,
let's drop the inflexible print_IPaddr() function and covert over to the
%pI4 modifier.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
CC: Ben Warren <biggerbadderben@gmail.com>
Enforce millisecond semantics of the first argument to NetSetTimeout() --
the change is transparent for well-behaving boards (CFG_HZ == 1000 and
get_timer() countiing in milliseconds).
Rationale for this patch is to enable millisecond granularity for
network-related timeouts, which is needed for the upcoming automatic
software update feature.
Summary of changes:
- do not scale the first argument to NetSetTimeout() by CFG_HZ
- change timeout values used in the networking code to milliseconds
Signed-off-by: Rafal Czubak <rcz@semihalf.com>
Signed-off-by: Bartlomiej Sieka <tur@semihalf.com>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
The DHCP handler has 1 state that is not listed in this case, causing a
failure message when there is actually no failure.
Signed-off-by: Remy Bohmer <linux@bohmer.net>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
on the network with it's offered IP number; it should not reply until
after it has received a DHCP ACK message. Also ensures that U-Boot
does it's DHCPREQUEST as broadcast (per RFC 2131).
Signed-off-by: Robin Getz <rgetz@blackfin.uclinux.org>
Acked-by: Ben Warren <biggerbadderben@gmail.com>
Signed-off-by: Wolfgang Denk <wd@denx.de>
Added option CONFIG_BOOTP_DHCP_REQUEST_DELAY. This provides an optional
delay before sending "DHCP Request" in net/bootp.c. Required to overcome
interoperability problems with Windows Server 200x DHCP server when U-Boot
client responds too fast for server to handle.
Signed-off-by: Aras Vaichas <arasv@magtech.com.au>
Signed-off-by: Ben Warren <biggerbadderben@gmail.com>
This is a multi-part message in MIME format.
commit e6e505eae94ed721e123e177489291fc4544b7b8
Author: Justin Flammia <jflammia@savantav.com>
Date: Mon Oct 29 17:19:03 2007 -0400
Found a bug in the way the DHCP Request packet is built, where the IP address
that is offered by the server is bound to prematurely. This patch is a fix of
that bug where the IP address offered by the DHCP server is not used until
after the DHCP ACK from the server is received.
Signed-off-by: Justin Flammia <jflammia@savantav.com>
Signed-off-by: Ben Warren <bwarren@qstreams.com>
