The existing adaptive ITR algorithm is overly restrictive. It throttles
incorrectly for various traffic rates, and does not produce good
performance. The algorithm now allows for more interrupts per second,
and does some calculation to help improve for smaller packet loads. In
addition, take into account the new itr_scale from the hardware which
indicates how much to scale due to PCIe link speed.
Reported-by: NMatthew Vick <matthew.vick@intel.com>
Reported-by: NAlex Duyck <alexander.duyck@gmail.com>
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Define a macro for identifying when the itr value is dynamic or
adaptive. The concept was taken from i40e. This helps make clear what
the check is, and reduces the line length to something more reasonable
in a few places.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The Intel Ethernet Switch FM10000 Host Interface interrupt throttle
timers are based on the PCIe link speed. Because of this, the value
being programmed into the ITR registers must be scaled accordingly.

For the PF, this is as simple as reading the PCIe link speed and storing
the result. However, in the case of SR-IOV, the VF's interrupt throttle
timers are based on the link speed of the PF. However, the VF is unable
to get the link speed information from its configuration space, so the
PF must inform it of what scale to use.

Rather than pass this scale via mailbox message, take advantage of
unused bits in the TDLEN register to pass the scale. It is the
responsibility of the PF to program this for the VF while setting up the
VF queues and the responsibility of the VF to get the information
accordingly. This is preferable because it allows the VF to set up the
interrupts properly during initialization and matches how the MAC
address is passed in the TDBAL/TDBAH registers.

Since we're modifying fm10k_type.h, we may as well also update the
copyright year.
Reported-by: NMatthew Vick <matthew.vick@intel.com>
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Originally this statistic was renamed because the method of dropping was
called "drop_oversized_messages", but this logic has changed much, and
this counter does actually represent messages which we failed to
transmit for a number of reasons. Rename the counter back to tx_dropped
since this is when it will increment, and it is less confusing.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

A previous bug was uncovered by addition of a debug stat to indicate the
actual number of DWORDS we pulled from the mbmem. It turned out this was
not the same as the tx_dwords counter. While the previous bug fix should
have corrected this in all cases, add some debug stats that count the
number of DWORDs pushed or pulled from the mbmem. A future debugger may
take advantage of this statistic for debugging purposes. Since we're
modifying fm10k_mbx.h, update the copyright year as well.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Since the resultant data type of the mac_update.mac_upper field is u16,
it does not make sense to typecast u8 variables to u32 first. Since
we're modifying fm10k_pf.c, also update the copyright year.
Reported-by: NMatthew Vick <matthew.vick@intel.com>
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The init_hw function may fail, and in the case of VFs, it might change
the number of maximum queues available. Thus, for every flow which
checks init_hw, we need to ensure that we clear the queue scheme before,
and initialize it after. The fm10k_io_slot_reset path will end up
triggering a reset so fm10k_reinit needs this change. The
fm10k_io_error_detected and fm10k_io_resume also need to properly clear
and reinitialize the queue scheme.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

A recent change modified init_hw in some flows the function may fail on
VF devices. For example, if a VF doesn't yet own its own queues.
However, many callers of init_hw didn't bother to check the error code.
Other callers checked but only displayed diagnostic messages without
actually handling the consequences.

Fix this by (a) always returning and preventing the netdevice from going
up, and (b) printing the diagnostic in every flow for consistency. This
should resolve an issue where VF drivers would attempt to come up
before the PF has finished assigning queues.

In addition, change the dmesg output to explicitly show the actual
function that failed, instead of combining reset_hw and init_hw into a
single check, to help for future debugging.

Fixes: 1d568b0f6424 ("fm10k: do not assume VF always has 1 queue")
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

VF drivers must detect how many queues are available. Previously, the
driver assumed that each VF has at minimum 1 queue. This assumption is
incorrect, since it is possible that the PF has not yet assigned the
queues to the VF by the time the VF checks. To resolve this, we added a
check first to ensure that the first queue is infact owned by the VF at
init_hw_vf time. However, the code flow did not reset hw->mac.max_queues
to 0. In some cases, such as during reinit flows, we call init_hw_vf
without clearing the previous value of hw->mac.max_queues. Due to this,
when init_hw_vf errors out, if its error code is not properly handled
the VF driver may still believe it has queues which no longer belong to
it. Fix this by clearing the hw->mac.max_queues on exit due to errors.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Don't change netdev hw_features later in fm10k_probe, instead set all
values inside fm10k_alloc_netdev. To do so, we need to know the MAC type
(whether it is PF or VF) in order to determine what to do. This helps
ensure that all logic regarding features is co-located.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The fm10k_msix_clean_rings function runs from hard interrupt context or
with interrupts already disabled in netpoll.

It can use napi_schedule_irqoff() instead of napi_schedule()
Signed-off-by: NAlexander Duyck <aduyck@mirantis.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch corrects an issue in which the polling routine would increase
the budget for Rx to at least 1 per queue if multiple queues were present.
This would result in Rx packets being processed when the budget was 0 which
is meant to indicate that no Rx can be handled.
Signed-off-by: NAlexander Duyck <aduyck@mirantis.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Based on hardware testing, the host interface supports up to 15368 bytes
as the maximum frame size. To determine the correct MTU, we subtract 8
for the internal switch tag, 14 for the L2 header, and 4 for the
appended FCS header, resulting in 15342 bytes of payload for our maximum
MTU on jumbo frames.
Signed-off-by: NMatthew Vick <matthew.vick@intel.com>
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Acked-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

It is possible that the PF has not yet assigned resources to the VF.
Although rare, this could result in the VF attempting to read queues it
does not own and result in FUM or THI faults in the PF. To prevent this,
check queue 0 before we continue in init_hw_vf.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <Krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This was detected by Coverity.
The function skb_cow_head leaves skb alone on failure, so caller needs
to free.
Signed-off-by: NStephen Hemminger <stephen@networkplumber.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As per Eric Dumazet's previous patches:
(see commit (24d2e4a5) - tg3: use napi_complete_done())

Quoting verbatim:
Using napi_complete_done() instead of napi_complete() allows
us to use /sys/class/net/ethX/gro_flush_timeout

GRO layer can aggregate more packets if the flush is delayed a bit,
without having to set too big coalescing parameters that impact
latencies.
</end quote>

Tested
configuration: low latency via ethtool -C ethx adaptive-rx off
				rx-usecs 10 adaptive-tx off tx-usecs 15
workload: streaming rx using netperf TCP_MAERTS

igb:
MIGRATED TCP MAERTS TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.0.0.1 () port 0 AF_INET : demo
...
Interim result:  941.48 10^6bits/s over 1.000 seconds ending at 1440193171.589

Alignment      Offset         Bytes    Bytes       Recvs   Bytes    Sends
Local  Remote  Local  Remote  Xfered   Per                 Per
Recv   Send    Recv   Send             Recv (avg)          Send (avg)
    8       8      0       0 1176930056  1475.36    797726   16384.00  71905

MIGRATED TCP MAERTS TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.0.0.1 () port 0 AF_INET : demo
...
Interim result:  941.49 10^6bits/s over 0.997 seconds ending at 1440193142.763

Alignment      Offset         Bytes    Bytes       Recvs   Bytes    Sends
Local  Remote  Local  Remote  Xfered   Per                 Per
Recv   Send    Recv   Send             Recv (avg)          Send (avg)
    8       8      0       0 1175182320  50476.00     23282   16384.00  71816

i40e:
Hard to test because the traffic is incoming so fast (24Gb/s) that GRO
always receives 87kB, even at the highest interrupt rate.

Other drivers were only compile tested.
Signed-off-by: NJesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Many drivers initialize uselessly n_priv_flags, n_stats, testinfo_len,
eedump_len & regdump_len fields in their .get_drvinfo() ethtool op.
It's not necessary as these fields is filled in ethtool_get_drvinfo().

v2: removed unused variable
v3: removed another unused variable
Signed-off-by: NIvan Vecera <ivecera@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Check for actual value NETREG_UNINITIALIZED in case it ever changes from
the current value of zero.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Signed-off-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Signed-off-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Add a private ethtool flag to enable display of these statistics, which
are generally less useful. However, sometimes it can be useful for
debugging purposes. The most useful portion is the ability to see what
the PF thinks the VF mailboxes look like.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

When we connect to the mailbox, we insert a fake disconnect header so
that the code does not see an invalid header and thus instantly error
every time we bring up the mailbox. However, we incorrectly record the
tail and head from the local perspective. Since the remote end shouldn't
have anything for us, add a "create_fake_disconnect_hdr" function which
inverts the TAIL and HEAD fields. This enables us to connect without any
errors of either TAIL or HEAD incorrectness, and prevents creating
extraneous error messages. This is necessary now since mbx_reset_work
does not actually reset the Tx FIFO head and tail pointers, thus head
and tail might not be equivalent on a reconnect.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch fixes a corner case issue with the PF/VF mailbox code.
Currently, fm10k_mbx_reset_work clears various state about the mailbox.
However, it does not clear the Tx FIFO head/tail pointers. We can't
simply clear these pointers as we unintentionally drop untransmitted
messages without error.

Doing nothing results in a possible phantom re-transmission of messages,
since we leave tx.head and tx.tail intact, but clear the tx_pulled and
tail_len values. This means that the PF could continuously re-send a
message which triggers a reset in the VF. Upon reset, the VF will
re-receive the same message after a reconnect.

If we reset the tx.head and tx.tail pointers completely, we end up
dropping some messages that were pending before connect. This results in
missing LPORT_MSG_READY bits, and VFs will end up reporting no link.

However, we can resolve both issues by simply incrementing head to
account for the already transmitted messages, before we reset tx_pulled.
We do this via the same logic as fm10k_mbx_head_pull.

We account for the tail_len which includes all data not yet transmitted,
once we account for the acked data which means re-reading the HEAD
variable from the message header. Then, we drop messages until we've
dropped more than the new tx_pulled value. At this point, resetting
tail_len and tx_pulled, but not tx.head and tx.tail will result in
prevention of the phantom message. It also prevents us from dropping
untransmitted messages upon attempting to Tx into a connect or
disconnect header.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This comment is no longer true due to a couple of mailbox locking
refactors, and we now don't actually do any rtnl protected operations
directly in the mailbox path. Remove this comment as it is factually
incorrect and confusing.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The VF will send a message to request multicast addresses with the
default VID. In the current code, if the PF has statically assigned a
VLAN to a VF, then the VF will not get the multicast addresses. Fix up
all of the various VLAN messages to use identical checks (since each
check was different). Also use set as a variable, so that it simplifies
our check for whether VLAN matches the pf_vid.

The new logic will allow set of a VLAN if it is zero, automatically
converting to the default VID. Otherwise it will allow setting the PF
VID, or any VLAN if PF has not statically assigned a VLAN. This is
consistent behavior, and allows VF to request either 0 or the
default_vid without silently failing.

Note that we need the check for zero since VFs might not get the default
VID message in time to actually request non-zero VLANs.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change makes it so that we only trigger the data path reset if the
fabric is ready to handle traffic.  The general idea is to avoid
triggering the reset unless the switch API is ready for us.  Otherwise
we can just postpone the reset until we receive a switch ready
notification.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Modify behavior of Malicious Driver Detection events. Presently, the
hardware disables the VF queues and re-assigns them to the PF. This
causes the VF in question to continuously Tx hang, because it assumes
that it can transmit over the queues in question. For transient events,
this results in continuous logging of malicious events.

New behavior is to reset the LPORT and VF state, so that the VF will
have to reset and re-enable itself. This does mean that malicious VFs
will possibly be able to continue and attempt malicious events again.
However, it is expected that system administrators will step in and
manually remove or disable the VF in question.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch ensures that VLAN traffic on the default VID will go to the
corresponding VLAN device if it exists. To do this, mask the rx_ring VID
if we have an active VLAN on that VID.

For this to work correctly, we need to update fm10k_process_skb_fields
to correctly mask off the VLAN_PRIO_MASK bits and compare them
separately, otherwise we incorrectly compare the priority bits with the
cleared flag. This also happens to fix a related bug where having
priority bits set causes us to incorrectly classify traffic.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Fix some style issues in debugfs code, and correct ordering of void and
__always_unused. Technically, the order does not matter, but preferred
style is to put the macro between the type and name.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This function is no longer used now that we have updated fm10k_slot_warn
functionality.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

If we store the sw_vid at reset of PF, then we accidentally prevent the
VF from receiving the message to update its default VID. This only
occurs if the VF is created before the PF has come up, which is the
standard way of creating VFs when using the module parameter.

This fixes an issue where we request the incorrect MAC/VLAN
combinations, and prevents us from accidentally reporting some frames as
VLAN tagged.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

We re-sync upon going up, so there is little reason to worry about not
syncing immediately with switch. This prevents an error that occurs if
you add a VLAN interface while down.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change adds the MAC address to the list of values recorded on driver
load.  The MAC address represents the serial number of the unit and allows
us to track the value should a card be replaced in a system.

The log message should now be similar in output to that of ixgbe.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This change pulls out the optimization that assumed that all fragments
would be limited to page size.  That hasn't been the case for some time now
and to assume this is incorrect as the TCP allocator can provide up to a
32K page fragment.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Acked-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Update the netdev permanent address during fm10k_reinit enables the user
to immediately see the new MAC address on the VF even if the device
isn't up. The previous code required that the device by opened before
changes would appear.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This is useful in cases where we connect to a slot at Gen3, but the slot
is behind a bus which only connected at Gen2. This generally only
happens when a PCIe switch is in the sequence of devices, and can be
very confusing when you see slow performance with no obvious cause.

I am aware this patch has a few lines that break 80 characters, but
there does not seem to be a readable way to format them to less than 80
characters. Suggestions welcome.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This allows us to correctly add a VLAN even if it matches our default
VID. However, we don't want to remove the VID rules once that VLAN is
deleted. Correctly remove the stack layers information of the VLAN, but
then return to forwarding that VID as untagged frames. If we deleted the
VID rules here, we would begin dropping traffic due to VLAN membership
violations.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The service task reads some registers as part of its normal routine,
even while the interface is down. Normally this is ok. However, during
suspend we have disabled the PCI device. Due to this, registers will
read in the same way as a surprise-remove event. Disable the service
task while we suspend, and re-enable it after we resume. If we don't do
this, the device could be UP when you suspend and come back from resume
as closed (since fm10k closes the device when it gets a surprise
remove).
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NKrishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>