Michael Brown [Thu, 31 Oct 2013 15:37:52 +0000 (15:37 +0000)]
[netdevice] Add generic concept of a network device configurator
iPXE supports multiple mechanisms for network device configuration:
DHCPv4 for IPv4, FIP for FCoE, and SLAAC for IPv6. At present, DHCPv4
requires an explicit action (e.g. a "dhcp" command), FIP is initiated
implicitly upon opening a network device, and SLAAC takes place
whenever a RA happens to be received.
Add a generic concept of a network device configurator, which provides
a common interface to triggering configuration and to reporting the
result of the configuration process.
Michael Brown [Tue, 5 Nov 2013 16:58:34 +0000 (16:58 +0000)]
[autoboot] Fix shell banner timeout
Commit 5e1fa5c ("[parseopt] Add parse_timeout()") introduced a
regression causing the shell banner timeout value (calculated in
milliseconds) to be treated as a timer tick count, resulting in a
timeout of approximately two minutes rather than the intended two
seconds.
Reported-by: Christian Hesse <list@eworm.de> Signed-off-by: Michael Brown <mcb30@ipxe.org>
Michael Brown [Fri, 1 Nov 2013 15:05:16 +0000 (15:05 +0000)]
[job] Allow job_progress() to return an ongoing job status code, if known
Some background jobs have a meaningful ongoing status code (e.g. the
current link status for a job waiting for a network link to come up).
Allow this to be exposed via the job_progress() method.
Michael Brown [Fri, 1 Nov 2013 15:52:31 +0000 (15:52 +0000)]
[interface] Default to calling intf_restart() in response to intf_close()
If an object interface does not provide an intf_close() method, then
default to calling intf_restart(). This allows static objects to
safely ignore intf_close(), without needing to add code solely to
ensure that the interface gets unplugged.
Michael Brown [Fri, 1 Nov 2013 02:22:12 +0000 (02:22 +0000)]
[parseopt] Add parse_timeout()
Parsing a timeout value (specified in milliseconds) into an internal
timeout value measured in timer ticks is a common operation. Provide
a parse_timeout() value to carry out this conversion automatically.
Michael Brown [Fri, 25 Oct 2013 16:24:09 +0000 (17:24 +0100)]
[pxe] Always retrieve cached DHCPACK and apply to relevant network device
When chainloading, always retrieve the cached DHCPACK packet from the
underlying PXE stack, and apply it as the original contents of the
"net<X>.dhcp" settings block. This allows cached DHCP settings to be
used for any chainloaded iPXE binary (not just undionly.kkpxe).
This change eliminates the undocumented "use-cached" setting. Issuing
the "dhcp" command will now always result in a fresh DHCP request.
Michael Brown [Fri, 25 Oct 2013 13:41:37 +0000 (14:41 +0100)]
[ipv6] Automatically choose source for link-local and multicast destinations
When transmitting to a link-local or multicast destination address,
use the network device's link-local address as the source address if
no explicit source address has been specified.
Michael Brown [Mon, 21 Oct 2013 13:17:47 +0000 (14:17 +0100)]
[ping] Add generic ping mechanism
Add generic pinger mechanism (analogous to the generic downloader
mechanism) which opens a ping socket, transmits ping requests, and
passes information about ping replies to a callback function.
Michael Brown [Mon, 21 Oct 2013 13:10:07 +0000 (14:10 +0100)]
[icmp] Add support for sending ICMP echo requests
Merge common functionality between IPv4 and IPv6 ICMP echo handling,
and add support for transmitting ICMP echo requests and delivering
ICMP echo replies to a (not yet implemented) ping_rx() function.
Michael Brown [Wed, 25 Sep 2013 11:55:46 +0000 (12:55 +0100)]
[build] Work around bug in gcc >= 4.8
Commit 238050d ("[build] Work around bug in gcc >= 4.8") works around
one instance of a bug in recent versions of gcc, in which "ebp" cannot
be specified within an asm clobber list.
Some versions of gcc seem to exhibit the same bug on other points in
the codebase. Fix by changing all instances of "ebp" in a clobber
list to use the push/pop %ebp workaround instead.
Originally-implemented-by: Víctor Román Archidona <contacto@victor-roman.es> Signed-off-by: Michael Brown <mcb30@ipxe.org>
Michael Brown [Wed, 4 Sep 2013 16:37:34 +0000 (17:37 +0100)]
[tcp] Reduce path MTU to 1280 bytes
The path MTU is currently hardcoded to 1460 bytes, which fails to
allow space for TCP options. Sending a maximum-sized datagram (which
is viable when using HTTP POST) will therefore fail since the Ethernet
MTU will be exceeded.
Reduce the hardcoded path MTU to produce a maximum datagram of 1280
bytes, which is the size required of data link layers by IPv6. It is
a reasonable assumption that all intermediary data link layers will be
able to convey this packet without fragmentation, even for IPv4.
Note that this reduction has a minimal impact upon download
throughput, since it affects only the transmit data path.
Originally-fixed-by: Suresh Sundriyal <ssundriy@vmware.com> Signed-off-by: Michael Brown <mcb30@ipxe.org>
Michael Brown [Mon, 26 Aug 2013 13:23:54 +0000 (14:23 +0100)]
[ipv6] Replace IPv6 stack
Replace the existing partially-implemented IPv6 stack with a fresh
implementation.
This implementation is not yet complete. The IPv6 transmit and
receive datapaths are functional (including fragment reassembly and
parsing of arbitrary extension headers). NDP neighbour solicitations
and advertisements are supported. ICMPv6 echo is supported.
At present, only link-local addresses may be used, and there is no way
to specify an IPv6 address as part of a URI (either directly or via
a DNS lookup).
Michael Brown [Sun, 1 Sep 2013 15:13:58 +0000 (16:13 +0100)]
[neighbour] Generalise concept of neighbour discovery
Split the protocol-independent portions of arp.c into a separate file
neighbour.c, to allow for sharing of functionality between IPv4+ARP
and IPv6+NDP.
Michael Brown [Thu, 29 Aug 2013 13:10:16 +0000 (14:10 +0100)]
[netdevice] Add concept of a network device index
IPv6 link-local socket addresses require some way to specify a local
network device. We cannot simply use a pointer to the network device,
since a struct sockaddr_in6 may be long-lived and has no way to hold a
reference to the network device.
Using a network device index allows a socket address to cleanly refer
to a network device without worrying about whether or not that device
continues to exist.
Michael Brown [Tue, 27 Aug 2013 15:28:54 +0000 (16:28 +0100)]
[linux] Give tap devices a name and bus type
Give tap devices a meaningful name, and avoid segmentation faults when
attempting to retrieve ${net0/bustype} by assigning a new bus type for
tap devices.
Michael Brown [Wed, 14 Aug 2013 14:07:52 +0000 (15:07 +0100)]
[uri] Allow URIs to incorporate a parameter list
HTTP POST requires the ability to associate a parameter list with a
URI. There is no standardised syntax for this. Use a non-standard
URI syntax to incorporate the specification of a parameter list within
a URI:
URI = [ absoluteURI | relativeURI ]
[ "#" fragment ] [ "##params" [ "=" paramsName ] ]
Michael Brown [Mon, 12 Aug 2013 12:42:12 +0000 (13:42 +0100)]
[settings] Expose memory map via settings mechanism
Allow memory map entries to be read using the syntax
${memmap/<region>.<properties>.<scale>}
where <region> is the index of the memory region, <properties> is a
bitmask where bit 0 represents the start address and bit 1 represents
the length (allowing the end address to be encoded by having both bits
0 and 1 set), and <scale> is the number of bits by which to shift the
result.
This allows for several values of interest to be encoded. For
example:
${memmap/<region>.1.0:hexraw} # 64-bit start address of <region>
${memmap/<region>.2.0:hexraw} # 64-bit length of <region>, in bytes
${memmap/<region>.3.0:hexraw} # 64-bit end address of <region>
${memmap/<region>.2.10:int32} # Length of <region>, in kB
${memmap/<region>.2.20:int32} # Length of <region>, in MB
The numeric encoding is slightly more sophisticated than described
here, allowing a single encoding to cover multiple regions. (See the
source code for details.) The primary use case for this feature is to
provide the total system memory size (in MB) via the "memsize"
predefined setting.
Michael Brown [Tue, 6 Aug 2013 18:16:30 +0000 (19:16 +0100)]
[settings] Expose CPUID instruction via settings mechanism
Allow CPUID values to be read using the syntax
${cpuid/<register>.<function>}
For example, ${cpuid/2.0x80000001} will give the value of %ecx after
calling CPUID with %eax=0x80000001. Values for <register> are encoded
as %eax=0, %ebx=1, %ecx=2, %edx=3.
The numeric encoding is more sophisticated than described above,
allowing for settings such as the CPU model (obtained by calling CPUID
with %eax=0x80000002-0x80000004 inclusive and concatenating the values
returned in %eax:%ebx:%ecx:%edx). See the source code for details.
The "cpuvendor" and "cpumodel" settings provide easy access to these
more complex CPUID settings.
This functionality is intended to complement the "cpuid" command,
which allows for testing individual CPUID feature bits.
Michael Brown [Tue, 6 Aug 2013 14:52:31 +0000 (15:52 +0100)]
[udp] Move high-frequency debug messages to DBGLVL_EXTRA
This makes it possible to leave UDP debugging enabled in order to see
interesting UDP events, without flooding the console with at least one
message per packet.
Michael Brown [Tue, 23 Jul 2013 15:49:12 +0000 (16:49 +0100)]
[script] Allow commands following a script label
Allow commands to be placed on the same line as a label. This allows
for improved legibility of loop constructions by incorporating the
loop check condition into the same line as the loop label. For
example, to iterate over network devices using the forthcoming "inc"
command:
set idx:int16 0
:loop isset ${net${idx}/mac} || goto loop_done
echo net${idx} is a ${net${idx}/chip} with MAC ${net${idx}/mac}
inc idx && goto loop
:loop_done
Michael Brown [Tue, 23 Jul 2013 13:46:45 +0000 (14:46 +0100)]
[script] Allow initial whitespace on lines containing labels
Initial whitespace is already accepted on lines containing commands,
since it gets ignored by the system() call. Minimise surprise and
allow for neater indentation of scripts by also allowing whitespace on
lines containing labels.
Michael Brown [Fri, 19 Jul 2013 13:07:18 +0000 (14:07 +0100)]
[settings] Eliminate calls to {fetch,store}f_named_setting() in NVO commands
A deliberate side effect of this commit is that the "read" command
will now preserve the type of the setting, if the setting name
contains no type information. For example:
iPXE> set foo:ipv4 192.168.0.1
iPXE> read foo
192.168.0.100
iPXE> show foo
foo:ipv4 = 192.168.0.100
rather than the arguably unexpected behaviour of resetting the type to
"string".
Michael Brown [Fri, 19 Jul 2013 13:53:38 +0000 (14:53 +0100)]
[settings] Change "not-found" semantics of fetch_setting_copy()
fetch_settings_copy() currently returns success and a NULL data
pointer to indicate a non-existent setting. This is intended to allow
the caller to differentiate between a non-existent setting and an
error in allocating memory for the copy of the setting.
The underlying settings blocks' fetch() methods provide no way to
perform an existence check separate from an attempt to fetch the
setting. A "non-existent setting" therefore means simply a setting
for which an error was encountered when attempting to fetch from every
settings block within the subtree.
Since any underlying error within a settings block (e.g. a GuestRPC
failure when attempting to retrieve a VMware GuestInfo setting) will
produce the effect of a "non-existent setting", it seems somewhat
meaningless to give special treatment to memory allocation errors
within fetch_setting_copy().
Remove the special treatment and simplify the semantics of
fetch_setting_copy() by directly passing through any underlying error
(including non-existence) encountered while fetching the setting.
Michael Brown [Thu, 18 Jul 2013 12:35:19 +0000 (13:35 +0100)]
[settings] Clarify usage of the term "named setting"
There are currently two conflicting usages of the term "named setting"
within iPXE: one refers to predefined settings (such as show up in the
"config" UI), the other refers to settings identified by a name (such
as "net0.dhcp/ip").
Split these usages into the term "predefined setting" and "named
setting" to avoid ambiguity.
Michael Brown [Mon, 15 Jul 2013 22:08:50 +0000 (00:08 +0200)]
[build] Fix %.licence build target
Our use of --gc-sections causes the linker to discard the symbols
defined by FILE_LICENCE(), meaning that the resulting licence
determination is incomplete.
We must use the KEEP() directive in the linker script to force the
linker to not discard the licence symbols. Using KEEP(*(COMMON))
would be undesirable, since there are some symbols in COMMON which we
may wish to discard.
Fix by placing symbols defined by PROVIDE_SYMBOL() (which is used by
FILE_LICENCE()) into a special ".provided" section, which we then mark
with KEEP(). All such symbols are zero-length, so there is no cost in
terms of the final binary size.
Since the symbols are no longer in COMMON, the linker will reject
symbols with the same name coming from multiple objects. We therefore
append the object name to the licence symbol, to ensure that it is
unique.
Reported-by: Marin Hannache <git@mareo.fr> Signed-off-by: Michael Brown <mcb30@ipxe.org>
Michael Brown [Mon, 15 Jul 2013 15:30:39 +0000 (17:30 +0200)]
[ifmgmt] Avoid relying on global variable within ifcommon_exec()
The getopt API defines optind as a global variable. When used by the
"autoboot" command, the payload function passed to ifcommon_exec() may
result in a new iPXE script being executed; the commands therein would
then overwrite the value of optind. On returning, ifcommon_exec()
would continue processing the list of interfaces from an undefined
point.
Fix by using a local variable to hold the index within the list of
interfaces.
Reported-by: Robin Smidsrød <robin@smidsrod.no> Signed-off-by: Michael Brown <mcb30@ipxe.org>
projects / thirdparty / ipxe.git / log
