Add the tap driver that can be used like:
$ ./ipxe.linux --net tap,if=tap0,mac=00:0c:29:c5:39:a1
The if setting is mandatory.
Signed-off-by: Piotr Jaroszyński <p.jaroszynski@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Support qemu-like arguments for network setup:
--net driver_name[,setting=value]*
and global settings:
--settings setting=value[,setting=value]*
Signed-off-by: Piotr Jaroszyński <p.jaroszynski@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Add the base to build linux drivers and the linux UI code on. UI
fills device requests, which are later walked over by the linux
root_driver and delegated to specific linux drivers.
Signed-off-by: Piotr Jaroszyński <p.jaroszynski@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
There exists an smbios userspace library so implementing this is
probably possible, but doesn't seem really important to have in
userspace. Hence provide a dummy implementation returning an error.
Signed-off-by: Piotr Jaroszyński <p.jaroszynski@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Add user access API for linux.
On linux userspace virtual == user == phys addresses. Physical
addresses also being the same is wrong, but there is no general way of
converting userspace addresses to physical as what appears to be
contiguous in userspace is physically fragmented. Currently only the
DMA memory is special-cased, but its conversion to bus addresses is
done in phys_to_bus. This is known to break virtio as it is passing
phys addresses to the virtual device.
Signed-off-by: Piotr Jaroszyński <p.jaroszynski@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Don't implement strtoul() on top of strtoull() as strtoull() is much
bigger and only used on linux currently. Instead refactor most of the
logic out of strtoul() into static inlines and reuse that. Also put it
in a separate object so it won't get linked in.
Signed-off-by: Piotr Jaroszyński <p.jaroszynski@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
pcbios specific get_memmap() is used by the b44 driver making
all-drivers builds fail on other platforms. Move it to the I/O API
group and provide a dummy implementation on EFI.
Signed-off-by: Piotr Jaroszyński <p.jaroszynski@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow packets in the receive queue to be discarded in order to free up
memory. This avoids a potential deadlock condition in which the
missing packet can never be received because the receive queue is
occupying all of the memory available for further RX buffers.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Add a facility allowing cached data to be discarded in order to
satisfy memory allocations that would otherwise fail.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Maintain a queue of received packets, so that lost packets need not
result in retransmission of the entire TCP window.
Increase the TCP window to 8kB, in order that we can potentially
transmit enough duplicate ACKs to trigger Fast Retransmission at the
sender.
Using a 10MB HTTP download in qemu-kvm with an artificial drop rate of
1 in 64 packets, this reduces the download time from around 26s to
around 4s.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
This patch adds a native iPXE virtio-net driver and removes the legacy
Etherboot virtio-net driver. The main reasons for doing this are:
1. Multiple virtio-net NICs are now supported by iPXE. The legacy
driver kept global state and caused issues in virtual machines with
more than one virtio-net device.
2. Faster downloads. The native iPXE driver downloads 100 MB over
HTTP in 12s, the legacy Etherboot driver in 37s. This simple
benchmark uses KVM with tap networking and the Python
SimpleHTTPServer both running on the same host.
Changes to core virtio code reduce vring descriptors to 256 (QEMU uses
128 for virtio-blk and 256 for virtio-net) and change the opaque token
from u16 to void*. Lowering the descriptor count reduces memory
consumption. The void* opaque token change makes driver code simpler.
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Weak functions whose visibility is hidden may be inlined due to a bug
in GCC. Explicitly mark weak functions noinline to work around the
problem.
This makes the PXE_MENU config option work again, the PXE boot menu
was never being called because the compiler inlined a weak stub
function.
The GCC bug was identified and fixed by Richard Sandiford
<rdsandiford@googlemail.com> but in the meantime iPXE needs to
implement a workaround.
Reported-by: Steve Jones <steve@squaregoldfish.co.uk>
Reported-by: Shao Miller <shao.miller@yrdsb.edu.on.ca>
Suggested-by: Joshua Oreman <oremanj@rwcr.net>
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Modified-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Continue calling step() while displaying the shell banner. This
potentially allows TCP connections to close gracefully after a failed
boot attempt.
Inspired-by: Guo-Fu Tseng <cooldavid@cooldavid.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Include the pause() and more() debugging functions within the general
iPXE debugging framework, by introducing DBGxxx_PAUSE() and
DBGxxx_MORE() macros.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
putline() was introduced back in 2007 for a feature that was never
committed. No console driver implements it and no code calls it, so
remove it from struct console_driver.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Add a PCI_CAP_ID_VNDR definition for the PCI standard "Vendor
Specific" capability ID.
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Since more reference-counted structures than embedded images might
want to mark themselves unfreeable, expose a dummy ref_no_free().
Signed-off-by: Michael Brown <mcb30@ipxe.org>
xfer_vredirect() should not be allowed to propagate to a pass-through
interface. For example, when an HTTPS connection is opened, the
redirect message should cause the TLS layer to reopen the TCP socket,
rather than causing the HTTP layer to disconnect from the TLS layer.
Fix by allowing for non-pass-through interface methods, and setting
xfer_vredirect() to be one such method.
This is slightly ugly, in that it complicates the notion of an
interface method call by adding a "pass-through" / "non-pass-through"
piece of metadata. However, the only current user of xfer_vredirect()
is iscsi.c, which uses it only because we don't yet have an
ioctl()-style call for retrieving the underlying socket address.
The new interface infrastructure allows for such a call to be created,
at which time this sole user of xfer_vredirect() can be removed,
xfer_vredirect() can cease to be an interface method and become simply
a wrapper around xfer_vreopen(), and the concept of a non-pass-through
interface method can be reverted.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Remove data-xfer as an interface type, and replace data-xfer
interfaces with generic interfaces supporting the data-xfer methods.
Filter interfaces (as used by the TLS layer) are handled using the
generic pass-through interface capability. A side-effect of this is
that deliver_raw() no longer exists as a data-xfer method. (In
practice this doesn't lose any efficiency, since there are no
instances within the current codebase where xfer_deliver_raw() is used
to pass data to an interface supporting the deliver_raw() method.)
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Remove name-resolution as an interface type, and replace
name-resolution interfaces with generic interfaces supporting the
resolv_done() method.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Remove job-control as an interface type, and replace job-control
interfaces with generic interfaces supporting the close() method.
(Both done() and kill() are absorbed into the function of close();
kill() is merely close(-ECANCELED).)
Signed-off-by: Michael Brown <mcb30@ipxe.org>
We have several types of object interface at present (data-xfer, job
control, name resolution), and there is some duplication of
functionality between them. For example, job_done(), job_kill() and
xfer_close() are almost isomorphic to each other.
This updated version of the object interface mechanism allows for each
interface to export an arbitrary list of supported operations.
Advantages include:
Operations methods now receive a pointer to the object, rather than
a pointer to the interface. This allows an object to, for example,
implement a single close() method that can handle close() operations
from any of its exposed interfaces.
The close() operation is implemented as a generic operation (rather
than having specific variants for data-xfer, job control, etc.).
This will allow functions such as monojob_wait() to be used to wait
for e.g. a name resolution to complete.
The amount of boilerplate code required in objects is reduced, not
least because it is no longer necessary to include per-interface
methods that simply use container_of() to derive a pointer to the
object and then tail-call to a common per-object method.
The cost of adding new operations is reduced; adding a new data-xfer
operation such as stat() no longer incurs the penalty of adding a
.stat member to the operations table of all existing data-xfer
interfaces.
The data-xfer, job control and name resolution interfaces have not yet
been updated to use the new interface mechanism, but the code will
still compile and run.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Standardise on using timer_init() to initialise an embedded retry
timer, to match the coding style used by other embedded objects.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Standardise on using ref_init() to initialise an embedded reference
count, to match the coding style used by other embedded objects.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
A new driver for JMicron Ethernet controller.
Reviewed-by: Joshua Oreman <oremanj@rwcr.net>
Reviewed-by: Michael Brown <mbrown@fensystems.co.uk>
Reviewed-by: Marty Connor <mdc@etherboot.org>
Signed-off-by: Guo-Fu Tseng <cooldavid@cooldavid.org>
Modified-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Add a new network driver that consumes the EFI Simple Network
Protocol. Also add a bus driver that can find the Simple Network
Protocol that iPXE was loaded from; the resulting behavior is similar
to the "undionly" driver for BIOS systems.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Autodetect the BSD licence statement in EFI header files, and add a
suitable FILE_LICENCE macro to the version imported into the iPXE
tree.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Now that the PACKED macro conflict is resolved, we can use an
unmodified import of the EFI header files (using
include/ipxe/efi/import.pl).
Synchronised to EDK2 SVN revision 10556.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Most of iPXE uses __attribute__((packed)) anyway, and PACKED conflicts
with an identically-named macro in the upstream EFI header files.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
This removes the need for inline safety wrappers, marginally reducing
the size penalty of weak functions, and works around an apparent
binutils bug that causes undefined weak symbols to not actually be
NULL when compiling with -fPIE (as EFI builds do).
A bug in versions of binutils prior to 2.16 (released in 2005) will
cause same-file weak definitions to not work with those
toolchains. Update the README to reflect our new dependency on
binutils >= 2.16.
Signed-off-by: Joshua Oreman <oremanj@rwcr.net>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Some switch configurations will refuse to enable our port unless we
can speak LACP to inform the switch that we are alive. Add a very
simple passive LACP implementation that is sufficient to convince at
least Linux's bonding driver (when tested using qemu attached to a tap
device enslaved to a bond device configured as "mode=802.3ad").
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Access to the gpxe.org and etherboot.org domains and associated
resources has been revoked by the registrant of the domain. Work
around this problem by renaming project from gPXE to iPXE, and
updating URLs to match.
Also update README, LOG and COPYRIGHTS to remove obsolete information.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
This commit adds an igb (Intel GigaBit) driver based on Intel source
code available at:
http://sourceforge.net/projects/e1000/
which is upstream source for the Linux kernel e1000 drivers, and
should support some PCIe e1000 variants.
Signed-off-by: Marty Connor <mdc@etherboot.org>
This commit adds an e1000e driver based on Intel source code
available at:
http://sourceforge.net/projects/e1000/
which is upstream source for the Linux kernel e1000 drivers, and
should support many PCIe e1000 variants.
Signed-off-by: Marty Connor <mdc@etherboot.org>
This commit replaces the current gPXE e1000 driver with one ported
from Intel source code available at
http://sourceforge.net/projects/e1000/
which is upstream source for the Linux kernel e1000 drivers, and
should support most if not all PCI e1000 variants.
Signed-off-by: Marty Connor <mdc@etherboot.org>
The NMB protocol code came from legacy Etherboot and was never updated
to work as a gPXE protocol. There has been no demand for this protocol,
so this patch removes it.
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Marty Connor <mdc@etherboot.org>
The IGMP code came from legacy Etherboot and was never updated to work
as a gPXE protocol. There has been no demand for this protocol, so this
patch removes it.
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Marty Connor <mdc@etherboot.org>
The NFS protocol code came from legacy Etherboot and was never updated
to work as a gPXE protocol. There has been no demand for this protocol,
so this patch removes it.
I have an unfinished NFSv3 over TCP implementation for gPXE that can be
used as the base for new work, should we want to resurrect this
protocol.
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Marty Connor <mdc@etherboot.org>
Currently, handling of URI escapes is ad-hoc; escaped strings are
stored as-is in the URI structure, and it is up to the individual
protocol to unescape as necessary. This is error-prone and expensive
in terms of code size. Modify this behavior by unescaping in
parse_uri() and escaping in unparse_uri() those fields that typically
handle URI escapes (hostname, user, password, path, query, fragment),
and allowing unparse_uri() to accept a subset of fields to print so
it can be easily used to generate e.g. the escaped HTTP path?query
request.
Signed-off-by: Joshua Oreman <oremanj@rwcr.net>
Signed-off-by: Marty Connor <mdc@etherboot.org>
For extremely tight space requirements and specific applications, it is
sometimes desirable to create gPXE images that cannot provide the PXE API
functionality to client programs. Add a configuration header option,
PXE_STACK, that can be removed to remove this stack. Also add PXE_MENU
to control the PXE boot menu, which most uses of gPXE do not need.
Signed-off-by: Marty Connor <mdc@etherboot.org>
When a DHCP session is started (using autoboot or a command-line `dhcp
net0'), check whether the new setting use-cached (DHCP option 175.178)
is TRUE; if so, skip DHCP and rely on currently registered
settings. This lets one combine a static IP with autoboot.
Before checking the use-cached setting, call a weak
get_cached_dhcpack() hook that can be implemented by particular builds
of gPXE supporting some fashion of retrieving a cached DHCPACK packet.
If one is available, it is registered as an options source, and then
either that packet's option 175.178 or the user's prior manual
use-cached setting can allow skipping duplicate DHCP.
Using cached packets is not the default because DHCP servers are often
configured to give gPXE different options than they give a vendor PXE
client; in order to break the infinite loop of PXE chaining, one would
need to load a gPXE with an embedded image that does something more
than autoboot.
Signed-off-by: Marty Connor <mdc@etherboot.org>
Weak symbols are a useful tool in eliminating unnecessary dependencies
between object files, but they are somewhat dangerous because one must
remember to test the weak symbol against NULL before using it. To
rectify that, add macros for declaring weak functions that will return
a default value inline if the file defining them is not available at
link time.
Signed-off-by: Marty Connor <mdc@etherboot.org>
This code is based on the linux skge driver. It supports Marvell Yukon
and SysKonnect Gigabit chipsets.
The code is based on code Michael Decker <mrd999@gmail.com> wrote for
Google Summer of Code 2008.
Support for dual-port cards is untested. The code, however, was left
in. In my opinion it's easier to fix the code if we need to, instead
of having to add support for it from scratch.
Signed-off-by: Marty Connor <mdc@etherboot.org>
This driver supports all current Myricom 10 gigabit Ethernet NICs.
It was written from scratch for gPXE by Glenn Brown <glenn@myri.com>,
referenencing Myricom's Linux and EFI drivers, with permission.
Signed-off-by: Glenn Brown <glenn@myri.com>
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Marty Connor <mdc@etherboot.org>
Taken from Linux /usr/include/linux/pci.h .
Signed-off-by: Glenn Brown <glenn@myri.com>
Signed-off-by: Stefan Hajnoczi <stefanha@gmail.com>
Signed-off-by: Marty Connor <mdc@etherboot.org>
EAPOL is a container protocol that can wrap either EAP packets or
802.11 EAPOL-Key frames. For cleanliness' sake, add a stub that strips
the framing and sends packets off to the appropriate handler if it
is compiled in.
Signed-off-by: Marty Connor <mdc@etherboot.org>
WEP is a highly flawed cryptosystem, barely better than no encryption at all,
but many people still use it. It does have the advantage of being very simple
and small in code size.
Signed-off-by: Marty Connor <mdc@etherboot.org>
Add commands `iwstat' (to list 802.11-specific status information for
802.11 devices) and `iwlist' (to scan for available networks and print
a list along with security information).
Signed-off-by: Marty Connor <mdc@etherboot.org>
This fixes an issue where passing a length as a compound expression
(e.g. using `hdrlen + datalen') would trigger compiler warnings and
potentially precedence-related errors.
Signed-off-by: Marty Connor <mdc@etherboot.org>
Both of these routines are used by 802.11 WPA, but they are generic
and could be needed by other protocols as well.
Signed-off-by: Marty Connor <mdc@etherboot.org>
It is often the case that some module of gPXE is only relevant if the
subsystem it depends on is already being included. For instance,
commands to manage wireless interfaces are quite useless if no
compiled-in driver has pulled in the wireless networking stack. There
may be a user-modifiable configuration options for these dependent
modules, but even if enabled, they should not be included when they
would be useless.
Solve this by allowing the creation of config_subsystem.c, for
configuration directives like those in the global config.c that should
only be considered when subsystem.c is included in the final gPXE
build.
For consistency, move core/config.c to the config/ directory, where
the other config_subsystem.c files will eventually reside.
Signed-off-by: Marty Connor <mdc@etherboot.org>
REQUIRE_SYMBOL() formerly used a formulation of symbol requirement
that would allow a link to succeed despite lacking a required symbol,
because it did not introduce any relocations. Fix by renaming it to
REQUEST_SYMBOL() (since the soft-requirement behavior can be useful)
and add a REQUIRE_SYMBOL() that truly requires.
Add EXPORT_SYMBOL() and IMPORT_SYMBOL() for REQUEST_SYMBOL()-like
behavior that allows one to make use of the symbol, by combining a
weak external on the symbol itself with a REQUEST_SYMBOL() of a second
symbol.
Signed-off-by: Marty Connor <mdc@etherboot.org>
The first byte of the IPoIB MAC address is used for flags indicating
support for "connected mode". Strip out the non-QPN bits of the first
dword when constructing the address vector for transmitted IPoIB
packets, so as not to end up passing an invalid QPN in the BTH.
The iBFT is Ethernet-centric in providing only six bytes for a MAC
address. This is most probably an indirect consequence of a similar
design flaw in the Windows NDIS stack. (The WinOF IPoIB stack
performs all sorts of contortions in order to pretend to the NDIS
layer that it is dealing with six-byte MAC addresses.)
There is no sensible way in which to extend the iBFT without breaking
compatibility with programs that expect to parse it. Add the notion
of an "Ethernet-compatible" MAC address to our link layer abstraction,
so that link layers can provide their own workarounds for this
limitation.
Some devices can only be reset via a mechanism that also resets the
card's PCI core, thus necessitating a backup and restore of all or
part of the PCI configuration space across a reset.
802.11 multicast hashing is the same as standard Ethernet hashing, so
just expose and use eth_mc_hash().
Signed-off-by: Joshua Oreman <oremanj@rwcr.net>
For IPoIB, the chaddr field is too small (16 bytes) to contain the
20-byte IPoIB link-layer address. RFC4390 mandates that we should
pass an empty chaddr field and rely on the DHCP client identifier
instead. This has many problems, not least of which is that a client
identifier containing an IPoIB link-layer address is not very useful
from the point of view of creating DHCP reservations, since the QPN
component is assigned at runtime and may vary between boots.
Leave the DHCP client identifier as-is, to avoid breaking existing
setups as far as possible, but expose the real hardware address (the
port GUID) via the DHCP chaddr field, using the broadcast flag to
instruct the DHCP server not to use this chaddr value as a link-layer
address.
This makes it possible (at least with ISC dhcpd) to create DHCP
reservations using host declarations such as:
host duckling {
fixed-address 10.252.252.99;
hardware unknown-32 00:02:c9:02:00:25:a1:b5;
}
IPoIB has a 20-byte link-layer address, of which only eight bytes
represent anything relating to a "hardware address".
The PXE and EFI SNP APIs expect the permanent address to be the same
size as the link-layer address, so fill in the "permanent address"
field with the initial link layer address (as generated by
register_netdev() based upon the real hardware address).
The hardware address is an intrinsic property of the hardware, while
the link-layer address can be changed at runtime. This separation is
exposed via APIs such as PXE and EFI, but is currently elided by gPXE.
Expose the hardware and link-layer addresses as separate properties
within a net device. Drivers should now fill in hw_addr, which will
be used to initialise ll_addr at the time of calling
register_netdev().
There is diagnostic value in being able to disambiguate between the
various reasons why an IB CM has rejected a connection attempt. In
particular, reason 8 "invalid service ID" can be used to identify an
incorrect SRP service_id root-path component, and reason 28 "consumer
reject" corresponds to a genuine SRP login rejection IU, which can be
passed up to the SRP layer.
For rejection reasons other than "consumer reject", we should not pass
through the private data, since it is most likely generated by the CM
without any protocol-specific knowledge.
With iSCSI, connection attempts are expensive; it may take many
seconds to determine that a connection will fail. SRP connection
attempts are much less expensive, so we may as well avoid the
"optimisation" of declaring a state of permanent failure after a
certain number of attempts. This allows a gPXE SRP initiator to
resume operations after an arbitrary amount of SRP target downtime.
SRP is the SCSI RDMA Protocol. It allows for a method of SAN booting
whereby the target is responsible for reading and writing data using
Remote DMA directly to the initiator's memory. The software initiator
merely sends and receives SCSI commands; it never has to touch the
actual data.
The minimal-surprise behaviour, when no explicit SRP initiator device
is specified, will probably be to use the most recently opened
Infiniband device. This matches our behaviour with using the most
recently opened net device for PXE, iSCSI, AoE, NBI, etc.
SRP over Infiniband uses a protocol whereby data is sent via a
combination of the CM private data fields and the RC queue pair
itself. This seems sufficiently generic that it's worth having
available as a separate protocol.
We will terminate our transaction as soon as we receive the first CM
REP, since that provides all the state that we need. However, the
peer may resend the REP if it didn't see our RTU, and if we don't
respond with another RTU we risk being disconnected. (This protocol
appears not to handle retries gracefully.)
Fix by adding a management agent that will listen for these duplicate
REPs and send back an RTU.
Previously the maximum packet length was computed using an erroneous
understanding of the role of the MIC field in TKIP-encrypted packets.
The field is actually considered to be part of the MSDU (encrypted and
fragmented data), not the MPDU (container for each encrypted
fragment). As such its size does not contribute to cryptographic
overhead outside the data field's size limitations. The net result is
that the previous maximum packet length value was 4 bytes too long;
fix it to the correct value of 2352.
Signed-off-by: Michael Brown <mcb30@etherboot.org>
The prior net80211 model of physical-layer behavior for drivers was
overly simplistic and limited the drivers that could be written. To
be more flexible, split the driver-provided list of supported rates by
band, and add a means for specifying a list of supported channels.
Allow drivers to specify a hardware channel value that will be tied to
uses of the channel.
Expose net80211_duration() to drivers, and make the rate it uses in
its computations configurable, so that it can be used in calculating
durations that must be set in hardware for ACK and CTS packets. Add
net80211_cts_duration() for the common case of calculating the
duration for a CTS packet.
Signed-off-by: Michael Brown <mcb30@etherboot.org>
A management interface is the component through which both local and
remote management agents are accessed.
This new implementation of a management interface allows for the user
to react to timed-out transactions, and also allows for cancellation
of in-progress transactions.
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.
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>
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.
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.
The PXE debugging messages have remained pretty much unaltered since
Etherboot 5.4, and are now difficult to read in comparison to most of
the rest of gPXE.
Bring the pxe_undi debug messages up to normal gPXE standards.
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>
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>
