Jeff Kirsher says:

====================
40GbE Intel Wired LAN Driver Updates 2018-01-10

This series contains updates to i40e and i40evf only.

Alice adds the displaying of priority xon/xoff packet stats, since we
were already keeping track of them.  Based on the recent changes, bump
the driver versions.

Jake changes how the driver determines whether or not the device is
currently up to resolve the possible issue of freeing data structures
and other memory before they have been fully allocated.  Refactored
the driver to simplify the locking behavior and to consistently use
spinlocks instead of an overloaded bit lock to protect MAC and filter
lists.  Created a helper function which can convert the AdminQ link
speed definition into a virtchnl definition.

Colin Ian King cleans up a redundant variable initialization.

Alex cleans up the driver to stop clearing the pending bit array for
each vector manually, since it is prone to dropping an interrupt and
based on the hardware specs, the pending bit array will be cleared
automatically in MSI-X mode.  Cleaned up flags for promiscuous mode to
resolve an issue where enabling & disabling promiscuous mode on a VF
would leave us in a high polling rate for the adminq task.  Cleaned up
code that was prone to race issues.

Jingjing renames pipeline personalization profile (ppp) to dynamic
device personalization (ddp) because it was being confused with the
well known point to point protocol.  Also removed checks for "track_id"
being zero, since it is valid for it to be zero for profiles that do
not have any 'write' commands.

v2: cleaned up commit message for patch 12 based on feedback from Sergei
    Shtylyov and Alex Duyck
v3: dropped patch 15 from the original series while Mariusz Stachura
    works on the changes that Jakub Kicinski has suggested
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Initiating a kdump via the command line can cause a pending interrupt
to be handled by the ibmvnic driver when initializing the sub-CRQ
irqs during driver initialization.

NIP [d000000000ca34f0] ibmvnic_interrupt_rx+0x40/0xd0 [ibmvnic]
LR [c000000008132ef0] __handle_irq_event_percpu+0xa0/0x2f0
Call Trace:
[c000000047fcfde0] [c000000008132ef0] __handle_irq_event_percpu+0xa0/0x2f0
[c000000047fcfea0] [c00000000813317c] handle_irq_event_percpu+0x3c/0x90
[c000000047fcfee0] [c00000000813323c] handle_irq_event+0x6c/0xd0
[c000000047fcff10] [c0000000081385e0] handle_fasteoi_irq+0xf0/0x250
[c000000047fcff40] [c0000000081320a0] generic_handle_irq+0x50/0x80
[c000000047fcff60] [c000000008014984] __do_irq+0x84/0x1d0
[c000000047fcff90] [c000000008027564] call_do_irq+0x14/0x24
[c00000003c92af00] [c000000008014b70] do_IRQ+0xa0/0x120
[c00000003c92af50] [c000000008002594] hardware_interrupt_common+0x114/0x180
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

add changes to t4_eth_xmit to enable vxlan segmentation
offload support.

Original work by: Santosh Rastapur <santosh@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Implement ndo_udp_tunnel_add and ndo_udp_tunnel_del
to support vxlan tunnelling.

Original work by: Santosh Rastapur <santosh@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add data structures and macros to be used in vxlan
offload.

Original work by: Santosh Rastapur <santosh@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Edward Cree says:

====================
sfc: support 25G configuration with ethtool

Adds support for advertise bits beyond the 32-bit legacy masks, and plumbs in
 translation of the new 25/50/100G bits to/from MCDI.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NEdward Cree <ecree@solarflare.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Store and handle ethtool link mode masks within the driver instead of
 just a single u32.  However, quite a significant amount of existing code
 wants to manipulate the masks directly, and thus now uses the first
 unsigned long (i.e. mask[0]) as though it were a legacy u32 mask.  This
 is ok because all the bits that code is interested in are in the first
 32 bits of the mask; but it might be a good idea to change them in
 future to use the proper bitmap API.
Signed-off-by: NEdward Cree <ecree@solarflare.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Only handles direct speed setting, not autoneg, because the driver is
 still trying to pretend it uses the legacy ethtool API which doesn't
 have advertised/supported bits for 25/50/100G.
Signed-off-by: NEdward Cree <ecree@solarflare.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jiri Pirko says:

====================
mlxsw qdisc refactoring

This patchset refactors the qdisc handling in mlxsw driver in order to make
it more object oriented like.
It helps readability, laying the groundwork for the offloading of
additional qdiscs by the driver
This patchset also makes the qdiscs statistics more generic.

Patch 1 moves the qdiscs declaration to the spectrum_qdisc.c
Patches 2-3 clean the offloaded stats requests. Patch 2 changes the RED
generic stats struct to be sharable by other offloaded qdiscs. Patch 3
changes the xstats request to be like the stats. Note that these patches
are outside the driver scope.
Patches 4-5 clean the statistics related functions and structs within the
driver.
Patches 6-7 decrease the need for the same parameters to be sent to many
functions.
Patches 8-11 create a functions pointers struct, to make the qdiscs
handling more object oriented like.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If a qdisc is being replaced by another qdisc of the same type, it can
simply override over its configuration.
However, if it replaces a qdisc of another type, it needs to be removed
before setting the new qdisc.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Create a generic qdisc replace function.
For that goal, add three functions to the qdisc ops struct:
* check_params: Checks if the given parameters are offloadable.
* replace: Offload the given parameters.
* clean_stats: clean the qdisc stats for the offloaded qdisc.
integrate RED offloading into using the new internal replace API.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add a destroy function to the qdiscs ops struct.
Create a generic qdisc destroy function, that clears the qdisc metadata as
well as calling the specific qdisc destroy function.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Qdisc struct have the Qdisc_class_ops struct.
This patch introduces the similar ops struct for the mlxsw_sp_qdisc_ops
struct. It allows better readability as well as code reusability for the
common parts of some functions like destroy.
The first operations to be added are the statistics getters.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Every qdisc op gets the qdisc handle ID as well as its location.  Each one
of them, beside replace, checks if the handle doesn't match the qdisc in
the given location, and if so, it returns without running the actual op.
Unite these checks to one comparison function and avoid sending the handle
id to these ops.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Tclass number is needed for most of the operations related to the qdisc in
the driver. Create a field for it in the mlxsw_sp_qdisc instead of passing
it to every function as parameter.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Improve readability by changing the clean stats function to handle only
RED. Qdiscs that will be offloaded in the future will have a clean stats
function of their own.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Clean RED offloaded stats and make them more generic by breaking the
generic qdisc stats to a struct of their own.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Change the value of the xstats requested from the driver for offloaded RED
to be incremental, like the normal stats.
It increases consistency - if a qdisc stops being offloaded its xstats
don't change.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Change the name of the stats struct to be generic, so it could be used for
other qdisc offload, that will be added in the next patches.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Move all the qdisc related data from the spectrum.h to spectrum_qdisc.c.
Create an init and fini functions for the qdiscs.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Reviewed-by: NYuval Mintz <yuvalm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fixes the following sparse warning:

net/ipv4/tcp.c:1736:6: warning:
 symbol 'tcp_recv_timestamp' was not declared. Should it be static?
Signed-off-by: NWei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix to return a negative error code from the xdp_rxq_info_reg() error
handling case instead of 0, as done elsewhere in this function.

Fixes: 0ddf5432 ("xdp/mlx5: setup xdp_rxq_info")
Signed-off-by: NWei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: NJesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We'd come in with SGE_FL_BUFFER_SIZE[0] and [1] both equal to 64KB and
the extant logic would flag that as an error. This was already fixed in
cxgb4 driver with "92ddcc7b cxgb4: Fix some small bugs in
t4_sge_init_soft() when our Page Size is 64KB".

Original Work by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: NArjun Vynipadath <arjun@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fixes: fc72d1d5 ("tuntap: XDP transmission")
Signed-off-by: NStephen Rothwell <sfr@canb.auug.org.au>
Acked-by: NJason Wang <jasowang@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

track_id == 0 is valid for “read only” profiles when
profile does not have any “write” commands.
Signed-off-by: NJingjing Wu <jingjing.wu@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

PPP name was going to be confusing since PPP already means point
to point protocol. It is decided to change pipeline personalization
profile(ppp) to dynamic device personalization(ddp).
Signed-off-by: NJingjing Wu <jingjing.wu@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Having the interrupts firing while we are polling causes extra overhead and
isn't needed for most systems out there. If an interrupt is lost us
experiencing a 2s latency spike before recovering is still not acceptable
and masks the issue. We are better off just identifying systems that lose
interrupts and instead enable workarounds for those systems.

To that end I am dropping the code that was strobing the interrupts as
there is a narrow window where having them enabled can actually cause
race issues anyway where a few stray packets might get misses if the
interrupt is re-enabled and fires before we call napi_complete.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

If you enabled and disabled promiscuous mode on a VF you could easily put
it into a state where it would start firing interrupts on all queues at a
rate of 50+ interrupts per second even though there was no traffic present.
The issue seems to have been a stray admin queue feature flag set that was
leaving us in a high polling rate for the adminq task.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

We should not be clearing the pending bit array for each vector manually.
The documentation for the hardware states that when in MSI-X mode the
pending bit array will be cleared automatically. Us clearing it ourselves
just results in multiple opportunities for us to drop an interrupt.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Variable read_size is initialized and this value is never read, it is
instead set inside the do-loop, hence the initialization is redundant
and can be removed. Cleans up clang warning:

drivers/net/ethernet/intel/i40e/i40e_nvm.c:390:6: warning: Value stored
to 'read_size' during its initialization is never read
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Bump the i40e driver from 2.1.14 to 2.3.2.

Bump the i40evf driver from 3.0.1 to 3.2.2
Signed-off-by: NAlice Michael <alice.michael@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

We introduced the virtchnl interface in order to have an interface for
talking to a virtual device driver which was host-driver agnostic. This
interface has its own definitions, including one for link speed.

The host driver has to talk to the virtchnl interface using these new
definitions in order to remain compatible. Today, the i40e link_speed
enumerations are value-exact matches for the virtchnl interface, so it
was originally decided to simply use a typecast.

However, this is unsafe, and makes it easier for future drivers to
continue this unsafe practice. There is nothing guaranteeing these
values are exact, and the type-cast would hide any compiler warning
which indicates the problem.

Rather than rely on this type cast, introduce a helper function which
can convert the AdminQ link speed definition into a virtchnl
definition. This can then be used by host driver implementations in
order to safely convert to the interface recognized by the virtual
functions.

If the link speed is not able to be represented by the virtchnl
definitions we'll report UNKNOWN which is the safest result.

This will ensure that should the driver specific link_speeds actual bit
definitions change, we do not report them incorrectly according to the
VF.

Additionally, this provides a better pattern for future drivers to copy,
as it is more likely a future device may not use the exact same bit-wise
definition as the current virtchnl interface.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

We currently notify a VF of the link state after ENABLE_QUEUES, which is
the last thing a VF does after being configured. Guests may not actually
ENABLE_QUEUES until they get configured, and thus between driver load
and device configuration the VF may show inaccurate link status.

Fix this by also sending the link state after GET_VF_RESOURCES. Although
we could remove the message following ENABLE_QUEUES, it's not that
significant of a loss, so this patch just keeps both to ensure maximum
compatibility with guests on various OSes.

Specifically, without this patch guests running FreeBSD will display
inaccurate link state until the device is brought up. This is mostly
a cosmetic issue but can be confusing to system administrators.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

If i40evf_open() is called quickly at the same time as a reset occurs
(such as via ethtool) it is possible for the device to attempt to open
while a reset is in progress. This occurs because the driver was not
holding the critical task bit lock during i40evf_open, nor was it
holding it around the call to i40evf_up_complete() in
i40evf_reset_task().

We didn't hold the lock previously because calls to i40evf_down() would
take the bit lock directly, and this would have caused a deadlock.

To avoid this, we'll move the bit lock handling out of i40evf_down() and
into the callers of this function. Additionally, we'll now hold the bit
lock over the entire set of steps when going up or down, to ensure that
we remain consistent.

Ultimately this causes us to serialize the transitions between down and
up properly, and avoid changing status while we're resetting.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Although not strictly necessary, it is customary to reverse the order in
which we release locks that we acquire. This helps preserve lock
ordering during future refactors, which can help avoid potential
deadlock situations.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Stop overloading the __I40EVF_IN_CRITICAL_TASK bit lock to protect the
mac_filter_list and vlan_filter_list. Instead, implement a spinlock to
protect these two lists, similar to how we protect the hash in the i40e
PF code.

Ensure that every place where we access the list uses the spinlock to
ensure consistency, and stop holding the critical section around blocks
of code which only need access to the macvlan filter lists.

This refactor helps simplify the locking behavior, and is necessary as
a future refactor to the __I40EVF_IN_CRITICAL_TASK would cause
a deadlock otherwise.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

In i40evf_reset_task we use netif_running() to determine whether or not
the device is currently up. This allows us to properly free queue memory
and shut down things before we request the hardware reset.

It turns out that we cannot be guaranteed of netif_running() returning
false until the device is fully up, as the kernel core code sets
__LINK_STATE_START prior to calling .ndo_open. Since we're not holding
the rtnl_lock(), it's possible that the driver's i40evf_open handler
function is currently being called while we're resetting.

We can't simply hold the rtnl_lock() while checking netif_running() as
this could cause a deadlock with the i40evf_open() function.
Additionally, we can't avoid the deadlock by holding the rtnl_lock()
over the whole reset path, as this essentially serializes all resets,
and can cause massive delays if we have multiple VFs on a system.

Instead, lets just check our own internal state __I40EVF_RUNNING state
field. This allows us to ensure that the state is correct and is only
set after we've finished bringing the device up.

Without this change we might free data structures about device queues
and other memory before they've been fully allocated.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NAndrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>