Allow network upper-layer drivers (such as LLDP, which attaches to
each network device in order to provide a corresponding LLDP settings
block) to specify a size for private data, which will be allocated as
part of the network device structure (as with the existing private
data allocated for the underlying device driver).
This will allow network upper-layer drivers to be simplified by
omitting memory allocation and freeing code. If the upper-layer
driver requires a reference counter (e.g. for interface
initialisation), then it may use the network device's existing
reference counter, since this is now the reference counter for the
containing block of memory.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
A running link block timer holds a reference to the network device and
will prevent it from being freed until the timer expires. It is
impossible for free_netdev() to be called while the timer is still
running: the call to stop_timer() therein is therefore a no-op.
Stop the link block timer when the device is closed, to allow a
link-blocked device to be freed immediately upon unregistration of the
device. (Since link block state is updated in response to received
packets, the state is effectively undefined for a closed device: there
is therefore no reason to leave the timer running.)
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Many laptops now include the ability to specify a "system-specific MAC
address" (also known as "pass-through MAC"), which is supposed to be
used for both the onboard NIC and for any attached docking station or
other USB NIC. This is intended to simplify interoperability with
software or hardware that relies on a MAC address to recognise an
individual machine: for example, a deployment server may associate the
MAC address with a particular operating system image to be deployed.
This therefore creates legitimate situations in which duplicate MAC
addresses may exist within the same system.
As described in commit 98d09a1 ("[netdevice] Avoid registering
duplicate network devices"), the Xen netfront driver relies on the
rejection of duplicate MAC addresses in order to inhibit registration
of the emulated PCI devices that a Xen PV-HVM guest will create to
shadow each of the paravirtual network devices.
Move the code that rejects duplicate MAC addresses from the network
device core to the Xen netfront driver, to allow for the existence of
duplicate MAC addresses in non-Xen setups.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The network device index currently serves two purposes: acting as a
sequential index for network device names ("net0", "net1", etc), and
acting as an opaque unique integer identifier used in socket address
scope IDs.
There is no particular need for these usages to be linked, and it can
lead to situations in which devices are named unexpectedly. For
example: if a system has two network devices "net0" and "net1", a VLAN
is created as "net1-42", and then a USB NIC is connected, then the USB
NIC will be named "net3" rather than the expected "net2" since the
VLAN device "net1-42" will have consumed an index.
Separate the usages: rename the "index" field to "scope_id" (matching
its one and only use case), and assign the name without reference to
the scope ID by finding the first unused name. For consistency,
assign the scope ID by similarly finding the first unused scope ID.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
When CONSOLE_SYSLOG is used, a DBG() from within a network device
driver may cause its transmit() or poll() methods to be unexpectedly
re-entered. Since these methods are not intended to be re-entrant,
this can lead to undefined behaviour.
Add an explicit re-entrancy guard to both methods. Note that this
must operate at a per-netdevice level, since there are legitimate
circumstances under which the netdev_tx() or netdev_poll() functions
may be re-entered (e.g. when using VLAN devices).
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Unlike netdev_rx_err(), there is no valid circumstance under which
netdev_rx() may be called with a null I/O buffer, since a call to
netdev_rx() represents the successful reception of a packet. Fix the
code comment to reflect this.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
netdev_tx_err() may be called with a null I/O buffer (e.g. to record a
transmit error with no associated buffer). Avoid a potential null
pointer dereference in the DMA unmapping code path.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Include a potential DMA mapping within the definition of an I/O
buffer, and move all I/O buffer DMA mapping functions from dma.h to
iobuf.h. This avoids the need for drivers to maintain a separate list
of DMA mappings for each I/O buffer that they may handle.
Network device drivers typically do not keep track of transmit I/O
buffers, since the network device core already maintains a transmit
queue. Drivers will typically call netdev_tx_complete_next() to
complete a transmission without first obtaining the relevant I/O
buffer pointer (and will rely on the network device core automatically
cancelling any pending transmissions when the device is closed).
To allow this driver design approach to be retained, update the
netdev_tx_complete() family of functions to automatically perform the
DMA unmapping operation if required. For symmetry, also update the
netdev_rx() family of functions to behave the same way.
As a further convenience for drivers, allow the network device core to
automatically perform DMA mapping on the transmit datapath before
calling the driver's transmit() method. This avoids the need to
introduce a mapping error handling code path into the typically
error-free transmit methods.
With these changes, the modifications required to update a typical
network device driver to use the new DMA API are fairly minimal:
- Allocate and free descriptor rings and similar coherent structures
using dma_alloc()/dma_free() rather than malloc_phys()/free_phys()
- Allocate and free receive buffers using alloc_rx_iob()/free_rx_iob()
rather than alloc_iob()/free_iob()
- Calculate DMA addresses using dma() or iob_dma() rather than
virt_to_bus()
- Set a 64-bit DMA mask if needed using dma_set_mask_64bit() and
thereafter eliminate checks on DMA address ranges
- Either record the DMA device in netdev->dma, or call iob_map_tx() as
part of the transmit() method
- Ensure that debug messages use virt_to_phys() when displaying
"hardware" addresses
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The Hermon driver uses vlan_find() to identify the appropriate VLAN
device for packets that are received with the VLAN tag already
stripped out by the hardware. Generalise this capability and expose
it for use by other network card drivers.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The UNDI layer uses the NETDEV_IRQ_ENABLED flag to choose whether to
return PXENV_UNDI_ISR_OUT_OURS or PXENV_UNDI_ISR_OUT_NOT_OURS for a
given interrupt. For a network device that does not support
interrupts, the flag will never be set and so pxenv_undi_isr() will
always return PXENV_UNDI_ISR_OUT_NOT_OURS. This causes some NBPs
(such as lpxelinux.0) to hang.
Redefine NETDEV_IRQ_ENABLED as a simple administrative flag which can
be set even on network devices that do not support interrupts. This
allows pxenv_undi_isr() (which is the sole user of NETDEV_IRQ_ENABLED)
to function as expected by lpxelinux.0.
Signed-off-by: Martin Habets <mhabets@solarflare.com>
Modified-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Some external code (such as the UEFI UNDI driver for the Realtek USB
NIC on a Microsoft Surface Book) will block during transmission
attempts and can take several seconds to report a transmit error. If
there is a large queue of pending transmissions, then the accumulated
time from a series of such failures can easily exceed the EFI watchdog
timeout, resulting in what appears to be a system lockup followed by a
reboot.
Work around this problem by immediately cancelling any pending
transmissions as soon as any transmit error occurs.
The only expected transmit error under normal operation is ENOBUFS
arising when the hardware transmit queue is full. By definition, this
can happen only for drivers that do not utilise deferred
transmissions, and so this new behaviour will not affect these
drivers.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Separate out the concept of "hardware maximum supported frame length"
and "configured link MTU", and limit the latter according to the
former.
In networks where the DHCP-supplied link MTU is inconsistent with the
hardware or driver capabilities (e.g. a network using jumbo frames),
this will result in iPXE advertising a TCP MSS consistent with a size
that can actually be received.
Note that the term "MTU" is typically used to refer to the maximum
length excluding the link-layer headers; we adopt this usage.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Avoid using zero as a network device index, so that a zero
sin6_scope_id can be used to mean "unspecified" (rather than
unintentionally meaning "net0").
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Provide a generic inject_fault() function that can be used to inject
random faults with configurable probabilities.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
When Spanning Tree Protocol (STP) is used, there may be a substantial
delay (tens of seconds) from the time that the link goes up to the
time that the port starts forwarding packets.
Add a generic concept of a "blocked link" (i.e. a link which is up but
which is not expected to communicate successfully), and allow "ifstat"
to indicate when a link is blocked.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
These files cannot be automatically relicensed by util/relicense.pl
since they either contain unusual but trivial contributions (such as
the addition of __nonnull function attributes), or contain lines
dating back to the initial git revision (and so require manual
knowledge of the code's origin).
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Reject network devices which appear to be duplicates of those already
available via a different underlying hardware device. On a Xen PV-HVM
system, this allows us to filter out the emulated PCI NICs (which
would otherwise appear alongside the netfront NICs).
Note that we cannot use the Xen facility to "unplug" the emulated PCI
NICs, since there is no guarantee that the OS we subsequently load
will have a native netfront driver.
We permit devices with the same MAC address if they are attached to
the same underlying hardware device (e.g. VLAN devices).
Inspired-by: Marin Hannache <git@mareo.fr>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
When functioning as an EFI driver, drivers can be disconnected and
reconnected multiple times (e.g. via the EFI shell "connect" command,
or by running an executable such as ipxe.efi which will temporarily
disconnect existing drivers).
Minimise surprise by resetting the network device index to zero
whenever the last device is unregistered. This is not foolproof, but
it does handle the common case of having all devices unregistered and
then reregistered in the original order.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
When opening a VLAN device, vlan_open() will call netdev_open() on the
trunk device. This will result in a call to netdev_notify(), which
will cause vlan_notify() to call vlan_sync() on the original VLAN
device, which will see that the trunk device is now open but the VLAN
device apparently isn't (since it has not yet been flagged as open by
netdev_open()). The upshot is a second attempt to open the VLAN
device, which will result in an erroneous second call to vlan_open().
This convoluted chain of events then terminates harmlessly since
vlan_open() calls netdev_open() on the trunk device, which just
returns immediately since the trunk device is by now flagged as being
already open.
Prevent this from happening by having netdev_open() flag the device as
open prior to calling the device's open() method, and reflagging it as
closed if the open() method fails.
Originally-fixed-by: Wissam Shoukair <wissams@mellanox.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
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.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Most network upper-layer drivers do not implement all three methods
(probe, notify, and remove). Save code by making all methods
optional.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
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.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow any iPXE command expecting a network device name to accept
"netX" as a synonym for "most recently opened network device".
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Devices with small transmit descriptor rings may temporarily run out
of space. Provide netdev_tx_defer() to allow drivers to defer packets
for retransmission as soon as a descriptor becomes available.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
iPXE currently seeds the random number generator using the system
timer tick count. When large numbers of machines are booted
simultaneously, multiple machines may end up choosing the same DHCP
transaction ID (XID) value; this can cause problems.
Fix by using the least significant (and hence most variable) bits of
each network device's link-layer address to perturb the random number
generator. This introduces some per-machine unique data into the
random number generator's seed, and so reduces the chances of DHCP XID
collisions.
This does not affect the ANS X9.82-compatible random bit generator
used by TLS and other cryptography code, which uses an entirely
separate source of entropy.
Originally-implemented-by: Bernhard Kohl <bernhard.kohl@nsn.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Some NICs (e.g. Hermon) provide hardware support for stripping the
VLAN tag, but do not provide any way for this support to be disabled.
Drivers for this hardware must therefore call vlan_find() to identify
a suitable receiving network device.
Provide a weak version of vlan_find() which will always return NULL if
VLAN support has not been enabled (either directly, or by enabling
a feature such as FCoE which requires VLAN support). This allows the
VLAN code to be omitted from builds where the user has not requested
support for VLANs.
Inspired-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The iBFT has a VLAN field that should be filled in. Add the
vlan_tag() function to extract the VLAN tag of a network device.
Since VLAN support is optional, define a weak function that returns 0
when iPXE is built without VLAN support.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Modified-by: Michael Brown <mcb30@ipxe.org>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Avoid memory leaks by clearing any (non-child) settings immediately
before unregistering the network device settings block.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Including a netdev_poll() within net_tx() can cause the net_step()
loop to end up processing hundreds or thousands of packets within a
single step, since each received packet being processed may trigger a
response which, in turn causes a poll for further received packets.
Network devices must now ensure that the TX ring is at least as large
as the RX ring, in order to avoid running out of TX descriptors. This
should not cause any problems; unlike the RX ring, there is no
substantial memory cost incurred by increasing the TX ring size.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
The current logic is to process at most one received packet per call
to net_poll(), on the basis that refilling the hardware descriptor
ring should be delayed as little as possible. However, this limits
the rate at which packets can be processed and ultimately ends up
adding latency which, in turn, limits the achievable throughput.
With temporary modifications in place to essentially remove all
resource constraints (heap size increased to 16MB, RX descriptor ring
increased to 64 descriptors) and a TCP window size of 1MB, the
throughput on a gigabit (i.e. 119MBps) network can be observed to fall
off exponentially from around 115MBps to around 75MBps. Changing
net_poll() to process all received packets results in a steady
119MBps throughput.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Drivers are currently expected to initialise only the hardware
address, with the link-layer protocol code taking care of converting
this into a valid link-layer address. Some drivers (e.g. undinet) can
legitimately determine both the hardware and link-layer addresses,
which may differ.
Allow for this situation by checking to see if the link-layer address
is empty before initialising it from the hardware address.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow the link layer to directly report whether or not a packet is
multicast or broadcast at the time of calling pull(), rather than
relying on heuristics to determine this at a later stage.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Give the step() method a pointer to the containing object, rather than
a pointer to the process. This is consistent with the operation of
interface methods, and allows a single function to serve as both an
interface method and a process step() method.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Allow TX errors to be recorded against a network device even when the
packet didn't make it as far as netdev_tx().
Inspired-by: Dominik Russenberger <dominik.russenberger@terreactive.ch>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
For devices that start in a link-down state, the user will see a
message such as:
[Link status: The socket is not connected (http://ipxe.org/38086001)]
Waiting for link-up on net0...
This is potentially misleading, since it suggests that there is a
genuine problem. Add a dedicated error message for "link down",
giving instead:
[Link status: Down (http://ipxe.org/38086101)]
Waiting for link-up on net0...
Reported-by: Tal Aloni <tal.aloni.il@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>
netdev_close() assumes that devices that are open are on the
open_list, which wasn't true if device specific opening failed.
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Pass the settings block name as a parameter to register_settings(),
rather than defining it with settings_init() (and then possibly
changing it by directly manipulating settings->name).
Signed-off-by: Michael Brown <mcb30@ipxe.org>
Michael Brown
Martin Habets
Stefan Hajnoczi
Piotr Jaroszyński