Check if a qeth device supports IPv6 TX checksum offload, and advertise
NETIF_F_IPV6_CSUM accordingly. Add support for setting the relevant bits
in IPv6 packet descriptors.

Currently this has only limited use (ie. UDP, or Jumbo Frames). For any
TCP traffic with a standard MSS, the TCP checksum gets calculated
as part of the linear GSO segmentation.
Signed-off-by: NKittipon Meesompop <kmeesomp@linux.vnet.ibm.com>
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add some wrappers to make the protocol-specific Assist code a little
more generic, and use them for sending protocol-agnostic commands in
the Checksum Offload Assist code.
Signed-off-by: NKittipon Meesompop <kmeesomp@linux.vnet.ibm.com>
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For new functionality, the L2 subdriver will start using IPv6 assists.
So move the query from the L3 subdriver into the common setup path.
Signed-off-by: NKittipon Meesompop <kmeesomp@linux.vnet.ibm.com>
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This matches the statistics we gather for the TX offload path.
Signed-off-by: NKittipon Meesompop <kmeesomp@linux.vnet.ibm.com>
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As the vid_list is only accessed from process context, there's no need to
protect it with a spinlock.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Both qeth sub drivers use the same QDIO queue handlers, there's no need
to expose them via the driver's discipline. No functional change.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For z/VM NICs, qeth needs to consider which of the three CCW devices in
an MPC group it uses for requesting a managed MAC address.

On the Base device, the hypervisor returns a default MAC which is
pre-assigned when creating the NIC (this MAC is also returned by the
READ MAC primitive). Querying any other device results in the allocation
of an additional MAC address.

For consistency with READ MAC and to avoid using up more addresses than
necessary, it is preferable to use the NIC's default MAC. So switch the
the diag26c over to using a NIC's Read device, which should always be
identical to the Base device.

Fixes: ec61bd2f ("s390/qeth: use diag26c to get MAC address on L2")
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Submitting a cmd IO request (usually on the WRITE device, but for IDX
also on the READ device) is currently done with ccw_device_start()
and a manual timeout in the caller.
On timeout, the caller cleans up the related resources (eg. IO buffer).
But 1) the IO might still be active and utilize those resources, and
    2) when the IO completes, qeth_irq() will attempt to clean up the
       same resources again.

Instead of introducing additional resource locking, switch to
ccw_device_start_timeout() to ensure IO termination after timeout, and
let the IRQ handler alone deal with cleaning up after a request.

This also removes a stray write->irq_pending reset from
clear_ipacmd_list(). The routine doesn't terminate any pending IO on
the WRITE device, so this should be handled properly via IO timeout
in the IRQ handler.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Creating the global workqueue during driver init may fail, deal with it.
Also, destroy the created workqueue on any subsequent error.

Fixes: 0f54761d ("qeth: Support VEPA mode")
Signed-off-by: NJulian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For control IO, qeth currently tracks the index of the buffer that it
expects to complete the next IO on each qeth_channel. If the channel
presents an IRQ while this buffer has not yet completed, no completion
processing for _any_ completed buffer takes place.
So if the 'next buffer' is skipped for any sort of reason* (eg. when it
is released due to error conditions, before the IO is started), the
buffer obviously won't switch to PROCESSED until it is eventually
allocated for a _different_ IO and completes.
Until this happens, all completion processing on that channel stalls
and pending requests possibly time out.

As a fix, remove the whole 'next buffer' logic and simply process any
IO buffer right when it completes. A channel will never have more than
one IO pending, so there's no risk of processing out-of-sequence.

*Note: currently just one location in the code really handles this problem,
       by advancing the 'next' index manually.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make sure to check both return code fields before(!) processing the
command response. Otherwise we risk operating on invalid data.

This matches an earlier fix for SETASSPARMS commands, see
commit ad3cbf61 ("s390/qeth: fix error handling in checksum cmd callback").
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename struct channel_path_desc to struct channel_path_desc_fmt0
to fit the scheme. Provide a macro for the function wrappers that
gather this and related data from firmware.
Signed-off-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

When the IRQ handler determines that one of the cmd IO channels has
failed and schedules recovery, block any further cmd requests from
being submitted. The request would inevitably stall, and prevent the
recovery from making progress until the request times out.

This sort of error was observed after Live Guest Relocation, where
the pending IO on the READ channel intentionally gets terminated to
kick-start recovery. Simultaneously the guest executed SIOCETHTOOL,
triggering qeth to issue a QUERY CARD INFO command. The command
then stalled in the inoperabel WRITE channel.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For calling ccw_device_start(), issue_next_read() needs to hold the
device's ccwlock.
This is satisfied for the IRQ handler path (where qeth_irq() gets called
under the ccwlock), but we need explicit locking for the initial call by
the MPC initialization.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

qeth_wait_for_threads() is potentially called by multiple users, make
sure to notify all of them after qeth_clear_thread_running_bit()
adjusted the thread_running_mask. With no timeout, callers would
otherwise stall.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

On removal, a qeth card's netdevice is currently not properly freed
because the call chain looks as follows:

qeth_core_remove_device(card)
	lx_remove_device(card)
		unregister_netdev(card->dev)
		card->dev = NULL			!!!
	qeth_core_free_card(card)
		if (card->dev)				!!!
			free_netdev(card->dev)

Fix it by free'ing the netdev straight after unregistering. This also
fixes the sysfs-driven layer switch case (qeth_dev_layer2_store()),
where the need to free the current netdevice was not considered at all.

Note that free_netdev() takes care of the netif_napi_del() for us too.

Fixes: 4a71df50 ("qeth: new qeth device driver")
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Reviewed-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

init_qdio_queues() resets the Input Queue's overall QDIO state, and
positions the buffer cursor back to 0. So this is the obvious place to
also reset the queue's NAPI context (in contrast to doing it rather
randomly in the middle of the big set_online() path).
No functional change.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

napi_alloc_skb() doesn't need to disable IRQs during the allocation,
and thus may save us a few cycles.
Doing so requires a small fix-up in the HiperTransport path, which
currently assumes a fixed NET_SKB_PAD headroom padding. napi_alloc_skb()
adds an additional NET_IP_ALIGN padding, so use the proper helper for
setting up the mac_header offset.

Use this opportunity to convert the non-NAPI path to netdev_alloc_skb(),
which means that skb->dev is now always set-up during allocation and
doesn't need to be assigned manually.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The 'portname' attribute is deprecated and setting it has no effect.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

"s390/qeth: fix SETIP command handling" introduced a new helper, apply
it driver-wide.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If multiple IPA commands are build & sent out concurrently,
fill_ipacmd_header() may assign a seqno value to a command that's
different from what send_control_data() later assigns to this command's
reply.
This is due to other commands passing through send_control_data(),
and incrementing card->seqno.ipa along the way.

So one IPA command has no reply that's waiting for its seqno, while some
other IPA command has multiple reply objects waiting for it.
Only one of those waiting replies wins, and the other(s) times out and
triggers a recovery via send_ipa_cmd().

Fix this by making sure that the same seqno value is assigned to
a command and its reply object.
Do so immediately before submitting the command & while holding the
irq_pending "lock", to produce nicely ascending seqnos.

As a side effect, *all* IPA commands now use a reply object that's
waiting for its actual seqno. Previously, early IPA commands that were
submitted while the card was still DOWN used the "catch-all" IDX seqno.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

qeth_get_elements_for_range() doesn't know how to handle a 0-length
range (ie. start == end), and returns 1 when it should return 0.
Such ranges occur on TSO skbs, where the L2/L3/L4 headers (and thus all
of the skb's linear data) are skipped when mapping the skb into regular
buffer elements.

This overestimation may cause several performance-related issues:
1. sub-optimal IO buffer selection, where the next buffer gets selected
   even though the skb would actually still fit into the current buffer.
2. forced linearization, if the element count for a non-linear skb
   exceeds QETH_MAX_BUFFER_ELEMENTS.

Rather than modifying qeth_get_elements_for_range() and adding overhead
to every caller, fix up those callers that are in risk of passing a
0-length range.

Fixes: 2863c613 ("qeth: refactor calculation of SBALE count")
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

send_control_data() applies some special handling to SETIP v4 IPA
commands. But current code parses *all* command types for the SETIP
command code. Limit the command code check to IPA commands.

Fixes: 5b54e16f ("qeth: do not spin for SETIP ip assist command")
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The transport mode that a z/VM NIC is configured in, must match the
hypervisor-internal network which the NIC is coupled to.

To get this right automatically, have qeth issue a diag26c hypervisor call
that provides all sorts of information for a specific VNIC.
With z/VM update VM65918, this also includes the VNIC's required
transport mode.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Be a little more self-documenting, and get rid of OSA_ADDR_LEN.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

atomic_t variables are currently used to implement reference
counters with the following properties:
 - counter is initialized to 1 using atomic_set()
 - a resource is freed upon counter reaching zero
 - once counter reaches zero, its further
   increments aren't allowed
 - counter schema uses basic atomic operations
   (set, inc, inc_not_zero, dec_and_test, etc.)

Such atomic variables should be converted to a newly provided
refcount_t type and API that prevents accidental counter overflows
and underflows. This is important since overflows and underflows
can lead to use-after-free situation and be exploitable.

The variable qeth_reply.refcnt is used as pure reference counter.
Convert it to refcount_t and fix up the operations.
Suggested-by: NKees Cook <keescook@chromium.org>
Reviewed-by: NDavid Windsor <dwindsor@gmail.com>
Reviewed-by: NHans Liljestrand <ishkamiel@gmail.com>
Signed-off-by: NElena Reshetova <elena.reshetova@intel.com>
[jwi: removed the WARN_ONs. Use CONFIG_REFCOUNT_FULL if you care.]
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make sure to check both return code fields before processing the
response. Otherwise we risk operating on invalid data.

Fixes: c9475369 ("s390/qeth: rework RX/TX checksum offload")
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Any modification to the takeover IP-ranges requires that we re-evaluate
which IP addresses are takeover-eligible. Otherwise we might do takeover
for some addresses when we no longer should, or vice-versa.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Just as for an explicit enable/disable, toggling the takeover mode also
requires that the IP addresses get updated. Otherwise all IPs that were
added to the table before the mode-toggle, get registered with the old
settings.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Using GSO with small MTUs currently results in a substantial throughput
regression - which is caused by how qeth needs to map non-linear skbs
into its IO buffer elements:
compared to a linear skb, each GSO-segmented skb effectively consumes
twice as many buffer elements (ie two instead of one) due to the
additional header-only part. This causes the Output Queue to be
congested with low-utilized IO buffers.

Fix this as follows:
If the MSS is low enough so that a non-SG GSO segmentation produces
order-0 skbs (currently ~3500 byte), opt out from NETIF_F_SG. This is
where we anticipate the biggest savings, since an SG-enabled
GSO segmentation produces skbs that always consume at least two
buffer elements.

Larger MSS values continue to get a SG-enabled GSO segmentation, since
1) the relative overhead of the additional header-only buffer element
becomes less noticeable, and
2) the linearization overhead increases.

With the throughput regression fixed, re-enable NETIF_F_SG by default to
reap the significant CPU savings of GSO.

Fixes: 5722963a ("qeth: do not turn on SG per default")
Reported-by: NNils Hoppmann <niho@de.ibm.com>
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It's good to have SPDX identifiers in all files to make it easier to
audit the kernel tree for correct licenses.

Update the drivers/s390/net/ files with the correct SPDX license
identifier based on the license text in the file itself.  The SPDX
identifier is a legally binding shorthand, which can be used instead of
the full boiler plate text.

This work is based on a script and data from Thomas Gleixner, Philippe
Ombredanne, and Kate Stewart.

Cc: Julian Wiedmann <jwi@linux.vnet.ibm.com>
Cc: Ursula Braun <ubraun@linux.vnet.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Kate Stewart <kstewart@linuxfoundation.org>
Cc: Philippe Ombredanne <pombredanne@nexb.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: NHeiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

A few lines down, qeth_prepare_control_data() makes further changes to
the control cmd buffer, and then also writes a trace entry for it.
So the first entry just pollutes the trace file with intermediate data,
drop it.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Reviewed-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Switch to napi_complete_done(), and thus enable delayed GRO flushing.
The timeout is configured via /sys/class/net/<if>/gro_flush_timeout.

Default timeout is 0, so no change in behaviour.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Reviewed-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Current code bails out when two subsequent buffer elements hold
insufficient data to contain a qeth_hdr packet descriptor.
This seems reasonable, but it would be legal for quirky hardware to
leave a few elements empty and then present packets in a subsequent
element. These packets would currently be dropped.

So make sure to check all buffer elements, until we hit the LAST_ENTRY
indication.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Move the allocation of SG skbs into the main path. This allows for
a little code sharing, and handling ENOMEM from within one place.

As side effect, L2 SG skbs now get the proper amount of additional
headroom (read: zero) instead of the hard-coded ETH_HLEN.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace the open-coded skb_add_rx_frag(), and use a fall-through
to remove some duplicated code.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

1. Drop the support for Token Ring,
2. use the ETH_DATA_LEN macro for the default L2 MTU,
3. handle OSM via the default case (as OSM is L2-only), and
4. document why the L3 MTU is reduced.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

With commit "s390/ccwgroup: tie a ccwgroup driver to its ccw driver",
the ccwgroup core now ensures that a qeth group device only consists of
ccw devices which are supported by qeth. Therefore remove qeth's
internal device matching, and use .driver_info to determine the card
type.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When recovering a device, qeth needs to re-run the IPA commands that
enable all previously active HW features.
Instead of duplicating qeth_set_features(), let netdev_update_features()
recover the missing HW features from dev->wanted_features.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When grouping devices, the ccwgroup core only checks whether all of the
devices are bound to the same ccw_driver. It has no means of checking
if the requesting ccwgroup driver actually supports this device type.
qeth implements its own device matching in qeth_core_probe_device(),
while ctcm and lcs currently have no sanity-checking at all.

Enable ccwgroup drivers to optionally defer the device type checking to
the ccwgroup core, by specifying their supported ccw_driver.
This allows us drop the device type matching from qeth, and improves
the robustness of ctcm and lcs.
Signed-off-by: NJulian Wiedmann <jwi@linux.vnet.ibm.com>
Acked-by: NSebastian Ott <sebott@linux.vnet.ibm.com>
Reviewed-by: NPeter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>