Jeff Kirsher says:
====================
40GbE Intel Wired LAN Driver Updates 2018-02-26
This series contains updates to i40e and i40evf only.
Mariusz adds a new ethtool private flag for forcing true link state with
the requested changes from Jakub Kicinski.
Paweł fixes an issue where we were double locking the same resource
which would generate a kernel panic after bringing an interface up for
i40evf.
Alan modifies both drivers to use software values to determine if there
are packets stalled on the ring with the added benefit of being less CPU
intensive since we do not need to reach into the hardware to get the
values.
Colin Ian King provides a few fixes detected by Coverity, first was to
pass a struct by reference versus by value to be more efficient. Then
verify the VSI pointer is not NULL before trying to dereference it.
Cleaned up redundant checks that always return true.
Dan Carpenter fixes over indented lines of code.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
These two lines are indented too far.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The checks to see if key->dst.s6_addr and key->src.s6_addr are null
pointers are redundant because these are constant size arrays and
so the checks always return true. Fix this by removing the redundant
checks. Also replace filter->f with vf, allowing wide lines to be
condensed and to rejoin some split wide lines.
Detected by CoverityScan, CID#1465279 ("Array compared to 0")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Function i40e_find_vsi_from_id can potentially return null, hence
VSI may be null, so defensively check it is non-null before
dereferencing it to check the seid.
Fixes: e284fc2804 ("i40e: Add and delete cloud filter")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Passing struct virtchnl_filter f by value requires a 272 byte copy
on x86_64, so instead pass it by reference is much more efficient. Also
adjust some lines that are over 80 chars.
Detected by CoverityScan, CID#1465285 ("Big parameter passed by value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Harshitha Ramamurthy <harshitha.ramamurthy@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The i40e_detect_recover_hung function uses the i40e_get_tx_pending
function to determine if there are packets stalled on the ring.
i40e_get_tx_pending calculates the pending packets using the head
writeback value and HW tail. If the queue is stopped and we lose the
interrupt to update our next_to_clean then we a) won't get another
interrupt to clean because queue is stopped b) we won't catch the
problem with i40e_detect_recover_hung because the HW values look like
there's no packets waiting to be transmitted. Using the SW values we
can catch the issue because next_to_clean will be out of sync with head
writeback.
This has the added benefit being less CPU intensive because we don't
need to reach into the hardware to get the values.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Removes the locking of adapter->mac_vlan_list_lock resource in
i40evf_add_filter(). The locking part is moved above i40evf_add_filter().
i40evf_add_filter(), called by i40evf_addr_sync(), was trying to lock the
resource again and double locking generated a kernel panic after bringing
an interface up.
Fixes: 8946b56354 ("i40evf: use __dev_[um]c_sync routines in
.set_rx_mode")
Signed-off-by: Paweł Jabłoński <pawel.jablonski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch introduces new ethtool private flag used for
forcing true link state. Function i40e_force_link_state that implements
this functionality was added, it sets phy_type = 0 in order to
work-around firmware's LESM. False positive error messages were
suppressed.
The ndo_open() should not succeed if there were issues with forcing link
state to be UP.
Added I40E_PHY_TYPES_BITMASK define with all phy types OR-ed together in
one bitmask. Added after phy type definition, so it will be hard to
forget to include new phy types to the bitmask.
Signed-off-by: Mariusz Stachura <mariusz.stachura@intel.com>
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Variable dma is initialized with a value that is never read, later
on it is re-assigned a new value, hence the initialization is redundant
and can be removed.
Cleans up clang warning:
drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c:584:13: warning: Value
stored to 'dma' during its initialization is never read
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add support for build_skb() similar to:
commit 6f429223b3 ("ixgbe: Add support for build_skb")
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Based on commit e014272672 ("igb: Break out Rx buffer page management")
Consolidate Rx code paths to reduce duplication when we expand them in
the future.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Make it so that all rings allocations are made as part of q_vector.
The advantage to this is that we can keep all of the memory related to
a single interrupt in one page.
The goal is to bring the logic of handling rings closer to ixgbe.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Similar to commit a50c29dd09
("ixgbe: Make certain that all frames fit minimum size requirements")
Make sure that any packet we attempt to transmit will meet minimum
size requirements.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Following the logic from commit 2de6aa3a66
("ixgbe: Add support for padding packet")
Add support for providing a buffer with headroom and tail room
to allow for shared info, NET_SKB_PAD, and NET_IP_ALIGN. With this
combined with the DMA changes we can start using build_skb to build frames
around an incoming Rx buffer instead of having to memcpy the headers.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Add calls for netif_set_real_num_t/rx_queues() in ixgbevf_open().
Make sure that calls to ixgbevf_open() are rtnl protected and improve
the error handling when setting up multiple queues.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Based on commit 8649aaef40
("igb: Add support for using order 1 pages to receive large frames")
Add support for using 3K buffers in order 1 page. We are reserving 1K for
now to have space available for future tail room and head room when we
enable build_skb support.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Introduce legacy-rx private flag that will allow switching between the
old and new (build_skb based) Rx code paths. The implementation is the
same as in commit e08912985b
("igb: Add support for ethtool private flag to allow use of legacy Rx")
This provides a means of validating the legacy Rx path in the event that
we are forced to fall back. At some point in the future when we are
convinced we don't need it anymore we might be able to drop the legacy-rx
flag.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Based on commit 3456fd5342
("igb: Use page_address offset from page instead of masking virtual address")
Update the handling of page addresses so that we always refer to them using
a void pointer, and try to use the consistent name of va indicating we are
working with a virtual address.
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
On hardware which supports timestamping all packets, the timestamps are
recorded in the packet buffer, and the driver no longer uses or reads
the registers. This makes the logic for checking and clearing Rx
timestamp hangs meaningless.
If we run the ixgbe_ptp_rx_hang() function in this case, then the driver
will continuously spam the log output with "Clearing Rx timestamp hang".
These messages are spurious, and confusing to end users.
The original code in commit a9763f3cb5 ("ixgbe: Update PTP to support
X550EM_x devices", 2015-12-03) did have a flag PTP_RX_TIMESTAMP_IN_REGISTER
which was intended to be used to avoid the Rx timestamp hang check,
however it did not actually check the flag before calling the function.
Do so now in order to stop the checks and prevent the spurious log
messages.
Fixes: a9763f3cb5 ("ixgbe: Update PTP to support X550EM_x devices", 2015-12-03)
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
If indir == 0 in the ixgbe_set_rxfh(), it is unnecessary
to write the HW. Because redirection table is not changed.
Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Variable pool is being assigned zero and then in the following for-loop
is it being set to zero again. Remove the redundant first assignment.
Cleans up clang warning:
drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c:61:2: warning: Value stored
to 'pool' is never read
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch provides support to add or delete cloud filter for queue
channels created for ADq on VF.
We are using the HW's cloud filter feature and programming it to act
as a TC filter applied to a group of queues.
There are two possible modes for a VF when applying a cloud filter
1. Basic Mode: Intended to apply filters that don't need a VF to be
Trusted. This would include the following
Dest MAC + L4 port
Dest MAC + VLAN + L4 port
2. Advanced Mode: This mode is only for filters with combination that
requires VF to be Trusted.
Dest IP + L4 port
When cloud filters are applied on a trusted VF and for some reason
the same VF is later made as untrusted then all cloud filters
will be deleted. All cloud filters has to be re-applied in
such a case.
Cloud filters are also deleted when queue channel is deleted.
Testing-Hints:
=============
1. Adding Basic Mode filter should be possible on a VF in
Non-Trusted mode.
2. In Advanced mode all filters should be able to be created.
Steps:
======
1. Enable ADq and create TCs using TC mqprio command
2. Apply cloud filter.
3. Turn-off the spoof check.
4. Pass traffic.
Example:
========
1. tc qdisc add dev enp4s2 root mqprio num_tc 4 map 0 0 0 0 1 2 2 3\
queues 2@0 2@2 1@4 1@5 hw 1 mode channel
2. tc qdisc add dev enp4s2 ingress
3. ethtool -K enp4s2 hw-tc-offload on
4. ip link set ens261f0 vf 0 spoofchk off
5. tc filter add dev enp4s2 protocol ip parent ffff: prio 1 flower\
dst_ip 192.168.3.5/32 ip_proto udp dst_port 25 skip_sw hw_tc 2
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch enables a tc filter to be applied as a cloud
filter for the VF. This patch adds functions which parse the
tc filter, extract the necessary fields needed to configure the
filter and package them in a virtchnl message to be sent to the
PF to apply them.
Signed-off-by: Harshitha Ramamurthy <harshitha.ramamurthy@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch handles the request from ADq enabled VF to allocate
bandwidth to each traffic class which means for each VSI.
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds support to configure bandwidth for the traffic
classes via tc tool. The required information is passed to the PF
which is used in the process of setting up the traffic classes.
Signed-off-by: Harshitha Ramamurthy <harshitha.ramamurthy@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch takes care of freeing up all the VSIs, queues and
other ADq related software and hardware resources, when a user
requests for deletion of ADq on VF.
Example command:
tc qdisc del dev eth0 root
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch allocates number of queues requested by the user as a part
of TC command when ADq is enabled on a VF.
In order to be consistent in design with PF implementation of ADq,
don't allow to set channels via ethtool from VF when ADq is already
enabled. This means the users will not be able to change the number of
queues/channels via ethtool for a VF when ADq is ON. In order to be
able to use set channels, users will be required to disable ADq first
and then try setting the channels again.
When ADq is enabled on VF, it goes through a reset during which VSIs
and queues are re-configured. Meanwhile if we receive link status
message from PF even before the queues are re-configured, just ignore
this link up message.
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch enables ADq and creates queue channels on a VF. An ADq
enabled VF can have up to 4 VSIs and each one of them represents
a traffic class and this is termed as a queue channel. Each of these
VSIs can have up to 4 queues. This patch services the request for
enabling ADq and adds queue channel based on the TC mqprio info
provided by the user in the VF.
Initially a check is made to see if spoof check is OFF, if not ADq
will not be enabled. PF notifies VF for a reset in order to complete
the creation of ADq resources i.e. creation of additional VSIs and
allocation of queues as per TC information, all in the reset path.
Steps:
======
1. Turn off the spoof check
2. Enable ADq using tc mqprio command with or without rate limit.
3. Pass traffic.
Example:
========
% ip link set dev eth0 vf 0 spoofchk off
% tc qdisc add dev $iface root mqprio num_tc 4 map\
0 0 0 0 1 1 1 1 2 2 2 2 3 3 3 3 queues\
4@0 4@4 4@8 4@8 hw 1 mode channel
Expected results:
=================
1. Total number of queues for the VF should be sum of queues of all TCs.
2. Traffic flow should be normal without errors.
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch introduces the callback to the ndo_setup_tc function
in the VF driver. We add a wrapper function to make room for the
upcoming cloud filter patches which add calls to different functions
from setup_tc.
First, we add support for capability exchange for ADQ between the
PF and VF. Next, we add support to take in the mqprio configuration
and configure queues as per the traffic classes, rate limit and the
priorities specified by the user. This is done by passing the channel
config to the PF driver through a virtchannel message.
The flags and bits added, track if ADq is enabled, set max number of
traffic classes to 4 and provide ability to negotiate capability with
the PF.
Signed-off-by: Harshitha Ramamurthy <harshitha.ramamurthy@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
One of the previous patch fixes the link up issue by ignoring it if
i40evf is not in __I40EVF_RUNNING state. However this doesn't fix the
race condition when queues are disabled esp for ADq on VF. Hence check
if all queues are enabled before starting all queues.
Signed-off-by: Avinash Dayanand <avinash.dayanand@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When the PF resets the VF, the VF puts out a warning message
indicating that the VF received a reset message from the PF.
Make this message more clear so that we do not mistakenly
think that the PF is undergoing a reset.
Signed-off-by: Harshitha Ramamurthy <harshitha.ramamurthy@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Similar to changes done to the PF driver in commit 6622f5cdba ("i40e:
make use of __dev_uc_sync and __dev_mc_sync"), replace our
home-rolled method for updating the internal status of MAC filters with
__dev_uc_sync and __dev_mc_sync.
These new functions use internal state within the netdev struct in order
to efficiently break the question of "which filters in this list need to
be added or removed" into singular "add this filter" and "delete this
filter" requests.
This vastly improves our handling of .set_rx_mode especially with large
number of MAC filters being added to the device, and even results in
a simpler .set_rx_mode handler.
Under some circumstances, such as when attached to a bridge, we may
receive a request to delete our own permanent address. Prevent deletion
of this address during i40evf_addr_unsync so that we don't accidentally
stop receiving traffic.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The MAC, FW Version and NPAR check used to determine
if shutting off the FW LLDP engine is supported is not
using the usual feature check mechanism.
This patch fixes the problem by moving the feature check
to i40e_sw_init in order to set a flag in pf->hw_features
that ethtool will use for priv_flags disable operation.
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Broadcast filters can now cause overflow promiscuous to trigger when
adding "too many" VLANs to all the ports of a device and the driver
needs a way to exit overflow promiscuous once triggered.
Currently the driver looks to see if there are "too many" filters and/or
we have any failed filters to determine when it is safe to exit overflow
promiscuous. If we trigger overflow promiscuous with broadcast filters,
any new filters added will be "auto-failed" until we exit overflow
promiscuous. Since the user can't manually remove the failed broadcast
filters for VLANs (nor should we expect the user to do such), there is
no way to exit overflow promiscuous without reloading the driver.
The easiest way to do this is to remove the shortcut to "auto-fail"
filters in overflow promiscuous. If the user removes the VLANs, the
failed filters will be removed and since we're no longer "auto-failing"
new filters, we'll eventually get a good set of filters and exit
overflow promiscuous.
This has the side benefit of making filter state more explicit in that
if a filter says it's failed we know for a fact it failed and not just
assuming it will if we're in overflow promiscuous. This is nice because
if the user removes some filters and then adds some, even if we're in
overflow promiscuous, the filter might succeed; we were just assuming it
won't because the user hasn't rectified other existing failed filters.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This code here is quite complex and easy to screw up. Let's see if we
can't improve the readability and maintainability a bit. This refactors
out promisc_changed into two variables 'old_overflow' and 'new_overflow'
which makes it a bit clearer when we're concerned about when and how
overflow promiscuous is changed. This also makes so that we no longer
need to pass a boolean pointer to i40e_aqc_add_filters. Instead we can
simply check if we changed the overflow promiscuous flag since the
function start.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When iterating through the linked list of VLAN filters, make the
iterator the same type as that of the linked list.
Signed-off-by: Harshitha Ramamurthy <harshitha.ramamurthy@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
When adding a bunch of VLANs to all the ports on a device, it's possible
to run out of space for broadcast filters. The driver should trigger
overflow promiscuous in this circumstance to prevent traffic from being
unexpectedly dropped.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Could a Bad Person do Bad Things to a server if they found these
addresses printed in the log? Who knows? But let's not take that risk.
Remove pointers from a bunch of printks. In some cases, I was able to
adjust the message to indicate whether or not the value was null. In
others, I just removed the entire message as there was really no hope of
saving it.
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
A spin lock is taken here so we should use GFP_ATOMIC.
Fixes: 504398f0a7 ("i40evf: use spinlock to protect (mac|vlan)_filter_list")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Fixes the following sparse warning:
drivers/net/ethernet/intel/i40e/i40e_main.c:5440:5: warning:
symbol 'i40e_get_link_speed' was not declared. Should it be static?
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch replaces the existing mechanism for determining the correct
value to program for adaptive ITR with yet another new and more
complicated approach.
The basic idea from a 30K foot view is that this new approach will push the
Rx interrupt moderation up so that by default it starts in low latency and
is gradually pushed up into a higher latency setup as long as doing so
increases the number of packets processed, if the number of packets drops
to 4 to 1 per packet we will reset and just base our ITR on the size of the
packets being received. For Tx we leave it floating at a high interrupt
delay and do not pull it down unless we start processing more than 112
packets per interrupt. If we start exceeding that we will cut our interrupt
rates in half until we are back below 112.
The side effect of these patches are that we will be processing more
packets per interrupt. This is both a good and a bad thing as it means we
will not be blocking processing in the case of things like pktgen and XDP,
but we will also be consuming a bit more CPU in the cases of things such as
network throughput tests using netperf.
One delta from this versus the ixgbe version of the changes is that I have
made the interrupt moderation a bit more aggressive when we are in bulk
mode by moving our "goldilocks zone" up from 48 to 96 to 56 to 112. The
main motivation behind moving this is to address the fact that we need to
update less frequently, and have more fine grained control due to the
separate Tx and Rx ITR times.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch is mostly prep-work for replacing the current approach to
programming the dynamic aka adaptive ITR. Specifically here what we are
doing is splitting the Tx and Rx ITR each into two separate values.
The first value current_itr represents the current value of the register.
The second value target_itr represents the desired value of the register.
The general plan by doing this is to allow for deferring the update of the
ITR value under certain circumstances. For now we will work with what we
have, but in the future I hope to change the behavior so that we always
only update one ITR at a time using some simple logic to determine which
ITR requires an update.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
While testing code for the recent ITR changes I found that updating the Tx
ITR appeared to have no effect with everything defaulting to the Rx ITR. A
bit of digging narrowed it down the fact that we were asking the PF to
associate all causes with ITR 0 as we weren't populating the itr_idx values
for either Rx or Tx.
To correct it I have added the configuration for these values to this
patch. In addition I did some minor clean-up to just add a local pointer
for the vector map instead of dereferencing it based off of the index
repeatedly. In my opinion this makes the resultant code a bit more readable
and saves us a few characters.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Instead of using the register value for the defines when setting up the
ring ITR we can just use the actual values and avoid the use of shifts and
macros to translate between the values we have and the values we want.
This helps to make the code more readable as we can quickly translate from
one value to the other.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The CLEARPBA bit in the dynamic interrupt control register actually has
no effect either way on the hardware. As per errata 28 in the XL710
specification update the interrupt is actually cleared any time the
register is written with the INTENA_MSK bit set to 0. As such the act of
toggling the enable bit actually will trigger the interrupt being
cleared and could lead to potential lost events if auto-masking is
not enabled.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch is a further clean-up related to the change over to using
q_vector->reg_idx when accessing the ITR registers. Specifically the code
appears to have several other spots where we were computing the register
offset manually and this resulted in errors in a few spots.
Specifically in the i40evf functions for mapping queues to vectors it
appears we may have had an off by 1 error since (v_idx - 1) for the first
q_vector with an index of 0 would result in us returning -1 if I am not
mistaken.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Currently in i40e_set_priv_flags we use new_flags to check for the
I40E_FLAG_DISABLE_FW_LLDP flag. This is an issue for a few a reasons.
DISABLE_FW_LLDP is persistent across reboots/driver reloads. This means
we need some way to detect if FW LLDP is enabled on init. We do this by
trying to init_dcb and if it fails with EPERM we know LLDP is disabled
in FW.
This could be a problem on older FW versions or NPAR enabled PFs because
there are situations where the FW could disable LLDP, but they do _not_
support using this flag to change it. If we do end up in this
situation, the flag will be set, then when the user tries to change any
priv flags, the driver thinks the user is trying to disable FW LLDP on a
FW that doesn't support it and essentially forbids any priv flag
changes.
The fix is simple, instead of checking if this flag is set, we should be
checking if the user is trying to _change_ the flag on unsupported FW
versions.
This patch also adds a comment explaining that the cmpxchg is the point
of no return. Once we put the new flags into pf->flags we can't back
out.
Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds a warning message when the link-down-on-close flag is
setting on. The warning is printed only on MFP devices
Signed-off-by: Paweł Jabłoński <pawel.jablonski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch adds necessary delay for 4.33 firmware to recover after
EMP reset. Without this patch driver occasionally reinitializes
structures too quickly to communicate with firmware after EMP reset
causing AdminQ to timeout.
Signed-off-by: Filip Sadowski <filip.sadowski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The logic for dynamic ITR update is confusing at best as there were odd
paths chosen for how to find the rings associated with a given queue based
on the vector index and other inconsistencies throughout the code.
This patch is an attempt to clean up the logic so that we can more easily
understand what is going on. Specifically if there is a Rx or Tx ring that
is enabled in dynamic mode on the q_vector it is allowed to override the
other side of the interrupt moderation. While it isn't correct all this
patch is doing is cleaning up the logic for now so that when we come
through and fix it we can more easily identify that this is wrong.
The other big change made here is that we replace references to:
vsi->rx_rings[q_vector->v_idx]->itr_setting
with:
q_vector->rx.ring->itr_setting
The general idea is we can avoid the long pointer chase since just
accessing q_vector->rx.ring is a single pointer access versus having to
chase down vsi->rx_rings, and then finding the pointer in the array, and
finally chasing down the itr_setting from there.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
