The IBA specification refers to management "interfaces" and "agents".
The interface is the component that connects to the queue pair and
sends and receives MADs; the agent is the component that constructs
the reply to the MAD.

Rename the IB_{QPN,QKEY,QPT} constants as a first step towards making
this separation in gPXE.

In several places, we currently use size_t to represent a difference
between TCP sequence numbers.  This can cause compiler warnings
relating to printf format specifiers, since the result of
(uint32_t+size_t) may be an unsigned long on some compilers.

Fix by using uint32_t for all variables that represent a difference
between TCP sequence numbers.

Tested-by: Joshua Oreman <oremanj@xenon.get-linux.org>

This is required for all modern 802.11 devices, and allows drivers
to be written for them with minimally more effort than is required
for a wired NIC.

Signed-off-by: Michael Brown <mcb30@etherboot.org>
Modified-by: Michael Brown <mcb30@etherboot.org>

The Communication Manager is responsible for handling the setup and
teardown of RC connections.

Queue pairs are now assumed to be created in the INIT state, with a
call to ib_modify_qp() required to bring the queue pair to the RTS
state.

ib_modify_qp() no longer takes a modification list; callers should
modify the relevant queue pair parameters (e.g. qkey) directly and
then call ib_modify_qp() to synchronise the changes to the hardware.

The packet sequence number is now a property of the queue pair, rather
than of the device.

Each queue pair may have an associated address vector.  For RC queue
pairs, this is the address vector that will be programmed in to the
hardware as the remote address.  For UD queue pairs, it will be used
as the default address vector if none is supplied to ib_post_send().

Now that MAD handlers no longer return a status code, we can allow
them to return a pointer to a MAD structure if and only if they want
to send a response.  This provides a more natural and flexible
approach than using a "response method" field within the handler's
descriptor.

MAD handlers have to set the status fields within the MAD itself
anyway, in order to provide a meaningful response MAD; the additional
gPXE return status code is just noise.

Note that we probably don't need to ever explicitly set the status to
IB_MGMT_STATUS_OK, since it should already have this value from the
request.  (By not explicitly setting the status in this way, we can
safely have ib_sma_set_xxx() call ib_sma_get_xxx() in order to
generate the GetResponse MAD without worrying that ib_sma_get_xxx()
will clear any error status set by ib_sma_set_xxx().)

Most IB hardware seems not to allow allocation of the genuine QPNs 0
and 1, so allow for the externally-visible QPN (as constructed and
parsed by ib_packet, where used) to differ from the real
hardware-allocated QPN.

The send completion handler typically will just free the I/O buffer,
so allow this common case to be handled by the Infiniband core.

The GMA code was based upon the SMA code.  We can save space by making
the SMA simply an instance of the GMA.

Now that IPoIB has to deal with only one set of queues, the queue set
abstraction becomes merely an inconvenient wrapper.

Generalise out the multicast group membership record code from IPoIB.

The queue key is stored as a property of the queue pair, and so can
optionally be added by the Infiniband core at the time of calling
ib_post_send(), rather than always having to be specified by the
caller.

This allows IPoIB to avoid explicitly keeping track of the data queue
key.

Generalise out the path record lookup code from IPoIB.

Generalise the subnet management agent into a general management agent
capable of sending and responding to MADs, including support for
retransmissions as necessary.

Currently, all Infiniband users must create a process for polling
their completion queues (or rely on a regular hook such as
netdev_poll() in ipoib.c).

Move instead to a model whereby the Infiniband core maintains a single
process calling ib_poll_eq(), and polling the event queue triggers
polls of the applicable completion queues.  (At present, the
Infiniband core simply polls all of the device's completion queues.)
Polling a completion queue will now implicitly refill all attached
receive work queues; this is analogous to the way that netdev_poll()
implicitly refills the RX ring.

Infiniband users no longer need to create a process just to poll their
completion queues and refill their receive rings.

IPoIB and the SMA have separate constants for the packet size to be
used to I/O buffer allocations.  Merge these into the single
IB_MAX_PAYLOAD_SIZE constant.

(Various other points in the Infiniband stack have hard-coded
assumptions of a 2048-byte payload; we don't currently support
variable MTUs.)

IPoIB has a link-layer broadcast address that varies according to the
partition key.  We currently go through several contortions to pretend
that the link-layer address is a fixed constant; by making the
broadcast address a property of the network device rather than the
link-layer protocol it will be possible to simplify IPoIB's broadcast
handling.

Move the icky call to step() from aoe.c to ata.c; this takes it at
least one step further away from where it really doesn't belong.

Unfortunately, AoE has the ugly aoe_discover() mechanism which means
that we still have a step() loop in aoe.c for now; this needs to be
replaced at some future point.

Move the icky call to step() from iscsi.c to scsi.c; this takes it at
least one step further away from where it really doesn't belong.

Expand the NETDEV_LINK_UP bit into a link_rc status code field,
allowing specific reasons for link failure to be reported via
"ifstat".

Originally-authored-by: Joshua Oreman <oremanj@rwcr.net>

Commit 558c1a4 ("[tcp] Improve robustness in the presence of duplicated
received packets") introduced a regression in that an old duplicate
ACK received while in the ESTABLISHED state would pass through normal
ACK processing, including updating tcp->snd_seq.

Fix by ensuring that ACK processing ignores all duplicate ACKs.

All TCP errors or unusual events should now generate a debugging
message at DBGLVL_LOG, with enough information (SEQ and ACK numbers)
to be able to identify the corresponding packet (or missing packet) in
a network trace from the remote end.

This makes it possible to leave TCP debugging enabled in order to see
interesting TCP events, without flooding the console with at least one
message per packet.

In order to construct outgoing link-layer frames or parse incoming
ones properly, some protocols (such as 802.11) need more state than is
available in the existing variables passed to the link-layer protocol
handlers. To remedy this, add struct net_device *netdev as the first
argument to each of these functions, so that more information can be
fetched from the link layer-private part of the network device.

Updated all three call sites (netdevice.c, efi_snp.c, pxe_undi.c) and
both implementations (ethernet.c, ipoib.c) of ll_protocol to use the
new argument.

Signed-off-by: Michael Brown <mcb30@etherboot.org>

gPXE responds to duplicated ACKs with an immediate retransmission,
which can lead to a sorceror's apprentice syndrome.  It also responds
to out-of-range (or old duplicate) ACKs with a RST, which can cause
valid connections to be dropped.

Fix the sorceror's apprentice syndrome by leaving the retransmission
timer running (and so inhibiting the immediate retransmission) when we
receive a potential duplicate ACK.  This seems to match the behaviour
of Linux observed via wireshark traces.

Fix the RST issue by sending RST only on out-of-range ACKs that occur
before the connection is fully established, as per RFC 793.

These problems were exposed during development of the 802.11 wireless
link layer; the 802.11 protocol has a failure mode that can easily
cause duplicated packets.  The fixes were tested in a controlled way
by faking large numbers of duplicated packets in the rtl8139 driver.

Originally-fixed-by: Joshua Oreman <oremanj@rwcr.net>

Signed-off-by: Michael Brown <mcb30@etherboot.org>

This provides a mechanism for using arbitrarily-named variables within
gPXE, using the existing syntax for settings.

If the ProxyDHCPOFFER already includes PXE options (i.e. option 60 is
set to "PXEClient" and option 43 is present) then assume that the
ProxyDHCPREQUEST can be sent to port 67, rather than port 4011.  This
is a reasonable assumption, since in that case the ProxyDHCP server
has already demonstrated by responding to the DHCPDISCOVER that it is
listening on port 67.  (If the ProxyDHCP server were not listening on
port 67, then the standard DHCP server would have been configured to
respond with option 60 set to "PXEClient" but no option 43 present.)

The PXE specification is ambiguous on this point; the specified
behaviour covers only the cases in which option 43 is *not* present in
the ProxyDHCPOFFER.  In these cases, we will continue to send the
ProxyDHCPREQUEST to port 4011.

This change is required in order to allow us to interoperate with
dnsmasq, which listens only on port 67.  (dnsmasq relies on
unspecified behaviour of the Intel PXE stack, which it seems will
retain the ProxyDHCPOFFER as an options source and never issue a
ProxyDHCPREQUEST, thereby enabling dnsmasq to omit listening on port
4011.)

IBM Tivoli PXE Server 5.1.0.3 is reported to send trailing garbage
bytes at the end of the OACK packet, which causes gPXE to reject the
packet and abort the TFTP transfer.

Work around the problem by processing as much as possible of the OACK,
and treating name/value parsing errors as non-fatal.

Reported-by: Shao Miller <Shao.Miller@yrdsb.edu.on.ca>

We currently send all boot server discovery requests to port 4011.
Section 2.2.1 of the PXE spec states that boot server discovery
packets should be "sent broadcast (port 67), multicast (port 4011), or
unicast (port 4011)".  Adjust our behaviour so that any boot server
discovery packets that are sent to the broadcast address are directed
to port 67 rather than port 4011.

This is required for operation with dnsmasq as a PXE server, since
dnsmasq listens only on port 67, and relies upon this (specified)
behaviour.

This change may break some setups using the (itself very broken) Linux
PXE server from kano.org.uk.  This server will, in its default
configuration, listen only on port 4011.  It never constructs a boot
server list (PXE_BOOT_SERVERS, option 43.8), and uses the wrong
definitions for the discovery control bits (PXE_DISCOVERY_CONTROL,
option 43.6).  The upshot is that it will always instruct the client
to perform multicast and broadcast discovery only.  In setups lacking
a valid multicast route on the server side, this used to work because
gPXE would eventually give up on the (non-responsive) multicast
address and send a broadcast request to port 4011, which the Linux PXE
server would respond to.  Now that gPXE correctly sends this broadcast
request to port 67 instead, it is never seen by the Linux PXE server,
and the boot fails.  The fix is to either (a) set up a multicast route
correctly on the server side before starting the PXE server, or (b)
edit /etc/pxe.conf to contain the server's unicast address in the
"multicast_address" field (a hack that happens to work).

Suggested-by: Simon Kelley <simon@thekelleys.org.uk>

This prevents gPXE from wasting time attempting to contact a ProxyDHCP
server on port 4011 if the DHCP response already contains the relevant
PXE options.  This behaviour is hinted at (though not explicitly
specified) in the PXE spec, and seems to match what the Intel client
does.

Suggested-by: Simon Kelley <simon@thekelleys.org.uk>

Add FILE_LICENCE declarations to almost all files that make up the
various standard builds of gPXE.

Allow options with dedicated BOOTP fields to fall back to using the
equivalent DHCP option if the relevant field is empty.

When handling a redirection event, we need to close the existing
connection before opening the new connection.

Suggested-by: Bill Lortz <Bill.Lortz@premier.org>