Yotam Gigi [Tue, 3 Oct 2017 07:58:06 +0000 (09:58 +0200)]
skbuff: Add the offload_mr_fwd_mark field
Similarly to the offload_fwd_mark field, the offload_mr_fwd_mark field is
used to allow partial offloading of MFC multicast routes.
Switchdev drivers can offload MFC multicast routes to the hardware by
registering to the FIB notification chain. When one of the route output
interfaces is not offload-able, i.e. has different parent ID, the route
cannot be fully offloaded by the hardware. Examples to non-offload-able
devices are a management NIC, dummy device, pimreg device, etc.
Similar problem exists in the bridge module, as one bridge can hold
interfaces with different parent IDs. At the bridge, the problem is solved
by the offload_fwd_mark skb field.
Currently, when a route cannot go through full offload, the only solution
for a switchdev driver is not to offload it at all and let the packet go
through slow path.
Using the offload_mr_fwd_mark field, a driver can indicate that a packet
was already forwarded by hardware to all the devices with the same parent
ID as the input device. Further patches in this patch-set are going to
enhance ipmr to skip multicast forwarding to devices with the same parent
ID if a packets is marked with that field.
The reason why the already existing "offload_fwd_mark" bit cannot be used
is that a switchdev driver would want to make the distinction between a
packet that has already gone through L2 forwarding but did not go through
multicast forwarding, and a packet that has already gone through both L2
and multicast forwarding.
For example: when a packet is ingressing from a switchport enslaved to a
bridge, which is configured with multicast forwarding, the following
scenarios are possible:
- The packet can be trapped to the CPU due to exception while multicast
forwarding (for example, MTU error). In that case, it had already gone
through L2 forwarding in the hardware, thus A switchdev driver would
want to set the skb->offload_fwd_mark and not the
skb->offload_mr_fwd_mark.
- The packet can also be trapped due to a pimreg/dummy device used as one
of the output interfaces. In that case, it can go through both L2 and
(partial) multicast forwarding inside the hardware, thus a switchdev
driver would want to set both the skb->offload_fwd_mark and
skb->offload_mr_fwd_mark.
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellaox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Arjun Vynipadath [Tue, 3 Oct 2017 06:13:05 +0000 (11:43 +0530)]
cxgb4: Update comment for min_mtu
We have lost a comment for minimum mtu value set for netdevice with
'commit
d894be57ca92 ("ethernet: use net core MTU range checking in
more drivers"). Updating it accordingly.
Signed-off-by: Arjun Vynipadath <arjun@chelsio.com>
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 2 Oct 2017 22:16:03 +0000 (15:16 -0700)]
Merge branch '40GbE' of git://git./linux/kernel/git/jkirsher/next-queue
Jeff Kirsher says:
====================
40GbE Intel Wired LAN Driver Updates 2017-10-02
This series contains updates to i40e and i40evf.
Shannon Nelson fixes an issue where when a machine has more CPUs than
queue pairs, the counting gets a "little funky" and turns off Flow
Director. So to correct it, limit the number of LAN queues initially
allocated to be sure there are some left for Flow Director and other
features.
Lihong cleans up dead code by removing a condition check which cannot
ever be true.
Christophe Jaillet fixes a potential NULL pointer dereference, which
could happen if kzalloc() fails.
Filip corrects the reporting of supported link modes, which was incorrect
for some NICs. Added support for 'ethtool -m' command, which displays
information about QSFP+ modules.
Mariusz adds functions to read/write the LED registers to control the
LEDS, instead of accessing the registers directly whenever the LEDs
need to be controlled.
Jake fixes a regression where we introduced a scheduling while atomic,
so introduce a separate helper function which will manage its own need
for the mac_filter_hash_lock. Also cleaned up the "PF" parameter in
i40e_vc_disable_vf() since it is never used and is not needed. Fixed
a rare case where it is possible that a reset does not occur when
i40e_vc_disable_vf() is called, so modify i40e_reset_vf() to return a
bool to indicate whether it reset or not so that i40e_vc_disable_vf()
can wait until a reset actually occurs.
Alan adds the ability for the VF to request more or less underlying
allocated queues from the PF. Fixes the incorrect method for clearing
the vf_states variable with a NULL assignment, when we should be
using atomic bitops since we don't actually want to clear all the
flags. Fixed a resource leak, where the PF driver fails to inform
clients of a VF reset because we were incorrectly checking the
I40E_VF_STATE_PRE_ENABLE bit.
Mitch converts i40evf_map_rings_to_vectors() to a void function since
it cannot fail and allows us to clean up the checks for the function
return value.
Scott enables the driver(s) to pass traffic with VLAN tags using the
802.1ad Ethernet protocol.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Scott Peterson [Tue, 22 Aug 2017 10:57:54 +0000 (06:57 -0400)]
i40e: Stop dropping 802.1ad tags - eth proto 0x88a8
Enable i40e to pass traffic with VLAN tags using the 802.1ad ethernet
protocol ID (0x88a8).
This requires NIC firmware providing version 1.7 of the API. With
older NIC firmware 802.1ad tagged packets will continue to be dropped.
No VLAN offloads nor RSS are supported for 802.1ad VLANs.
Signed-off-by: Scott Peterson <scott.d.peterson@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Alan Brady [Tue, 22 Aug 2017 10:57:53 +0000 (06:57 -0400)]
i40e: fix client notify of VF reset
Currently there is a bug in which the PF driver fails to inform clients
of a VF reset which then causes clients to leak resources. The bug
exists because we were incorrectly checking the I40E_VF_STATE_PRE_ENABLE
bit.
When a VF is first init we go through a reset to initialize variables
and allocate resources but we don't want to inform clients of this first
reset since the client isn't fully enabled yet so we set a state bit
signifying we're in a "pre-enabled" client state. During the first
reset we should be clearing the bit, allowing all following resets to
notify the client of the reset when the bit is not set. This patch
fixes the issue by negating the 'test_and_clear_bit' check to accurately
reflect the behavior we want.
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>
Alan Brady [Tue, 22 Aug 2017 10:57:52 +0000 (06:57 -0400)]
i40e: fix handling of vf_states variable
Currently we inappropriately clear the vf_states variable with a null
assignment. This is problematic because we should be using atomic
bitops on this variable and we don't actually want to clear all the
flags. We should just clear the ones we know we want to clear.
Additionally remove the I40E_VF_STATE_FCOEENA bit because it is no
longer being used.
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>
Mitch Williams [Tue, 22 Aug 2017 10:57:51 +0000 (06:57 -0400)]
i40e: make i40evf_map_rings_to_vectors void
This function cannot fail, so why is it returning a value? And why are
we checking it? Why shouldn't we just make it void? Why is this commit
message made up of only questions?
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>
Alan Brady [Tue, 22 Aug 2017 10:57:50 +0000 (06:57 -0400)]
i40evf: Enable VF to request an alternate queue allocation
Currently the VF gets a default number of allocated queues from HW on
init and it could choose to enable or disable those allocated queues.
This makes it such that the VF can request more or less underlying
allocated queues from the PF.
First the VF negotiates the number of queues it wants that can be
supported by the PF and if successful asks for a reset. During reset
the PF will reallocate the HW queues for the VF and will then remap the
new queues.
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>
Jacob Keller [Tue, 22 Aug 2017 10:57:49 +0000 (06:57 -0400)]
i40e: ensure reset occurs when disabling VF
It is possible although rare that we may not reset when
i40e_vc_disable_vf() is called. This can lead to some weird
circumstances with some values not being properly set. Modify
i40e_reset_vf() to return a code indicating whether it reset or not.
Now, i40e_vc_disable_vf() can wait until a reset actually occurs. If it
fails to free up within a reasonable time frame we'll display a warning
message.
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>
Jacob Keller [Tue, 22 Aug 2017 10:57:48 +0000 (06:57 -0400)]
i40e: make use of i40e_vc_disable_vf
Replace i40e_vc_notify_vf_reset and i40e_reset_vf with a call to
i40e_vc_disable_vf which does this exact thing. This matches similar
code patterns throughout the driver.
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>
Jacob Keller [Tue, 22 Aug 2017 10:57:47 +0000 (06:57 -0400)]
i40e: drop i40e_pf *pf from i40e_vc_disable_vf()
It's never used, and the vf structure could get back to the PF if
necessary. Lets just drop the extra unneeded parameter.
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>
Jacob Keller [Tue, 22 Aug 2017 10:57:46 +0000 (06:57 -0400)]
i40e: don't hold spinlock while resetting VF
When we refactored handling of the PVID in commit
9af52f60b2d9
("i40e: use (add|rm)_vlan_all_mac helper functions when changing PVID")
we introduced a scheduling while atomic regression.
This occurred because we now held the spinlock across a call to
i40e_reset_vf(), which results in a usleep_range() call that triggers
a scheduling while atomic bug. This was rare as it only occurred if the
user configured a VLAN on a VF and also attempted to reconfigure the VF
from the host system with a port VLAN.
We do need to hold the lock while calling i40e_is_vsi_in_vlan(), but we
should not be holding it while we reset the VF.
We'll fix this by introducing a separate helper function
i40e_vsi_has_vlans which checks whether we have a PVID and whether the
VSI has configured VLANs. This helper function will manage its own need
for the mac_filter_hash_lock.
Then, we can move the acquiring of the spinlock until after we reset the
VF, which ensures that we do not sleep while holding the lock.
Using a separate function like this makes the code more clear and is
easier to read than attempting to release and re-acquire the spinlock
when we reset the VF.
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>
Mariusz Stachura [Tue, 22 Aug 2017 10:57:45 +0000 (06:57 -0400)]
i40e: use admin queue for setting LEDs behavior
Instead of accessing register directly, use newly added AQC in
order to blink LEDs. Introduce and utilize a new flag to prevent
excessive API version checking.
Signed-off-by: Mariusz Stachura <mariusz.stachura@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Filip Sadowski [Tue, 22 Aug 2017 10:57:44 +0000 (06:57 -0400)]
i40e: Add support for 'ethtool -m'
This patch adds support for 'ethtool -m' command which displays
information about (Q)SFP+ module plugged into NIC's cage.
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>
Filip Sadowski [Tue, 22 Aug 2017 10:57:43 +0000 (06:57 -0400)]
i40e: Fix reporting of supported link modes
This patch fixes incorrect reporting of supported link modes on some NICs.
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>
Christophe JAILLET [Sun, 6 Aug 2017 21:37:01 +0000 (23:37 +0200)]
i40e: Fix a potential NULL pointer dereference
If 'kzalloc()' fails, a NULL pointer will be dereferenced.
Return an error code (-ENOMEM) instead.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Lihong Yang [Thu, 27 Jul 2017 10:17:09 +0000 (03:17 -0700)]
i40e: remove logically dead code
This patch removes the !vf condition check that cannot be
true in i40e_ndo_set_vf_trust function
Detected by CoverityScan, CID
1397531 Logically dead code
Signed-off-by: Lihong Yang <lihong.yang@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Shannon Nelson [Tue, 25 Jul 2017 01:17:42 +0000 (18:17 -0700)]
i40e: limit lan queue count in large CPU count machine
When a machine has more CPUs than queue pairs, e.g. 512 cores, the
counting gets a little funky and turns off Flow Director with the
message:
not enough queues for Flow Director. Flow Director feature is disabled
This patch limits the number of lan queues initially allocated to
be sure we have some left for FD and other features.
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
David S. Miller [Mon, 2 Oct 2017 18:59:34 +0000 (11:59 -0700)]
Merge branch '100GbE' of git://git./linux/kernel/git/jkirsher/next-queue
Jeff Kirsher says:
====================
100GbE Intel Wired LAN Driver Updates 2017-10-02
This series contains updates to fm10k only.
Jake provides all but one of the changes in this series. Most are small
fixes, starting with ensuring prompt transmission of messages queued up
after each VF message is received and handled. Fix a possible race
condition between the watchdog task and the processing of mailbox
messages by just checking whether the mailbox is still open. Fix a
couple of GCC v7 warnings, including misspelled "fall through" comments
and warnings about possible truncation of calls to snprintf(). Cleaned
up a convoluted bitshift and read for the PFVFLRE register. Fixed a
potential divide by zero when finding the proper r_idx.
Markus Elfring fixes an issue which was found using Coccinelle, where
we should have been using seq_putc() instead of seq_puts().
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 2 Oct 2017 18:24:42 +0000 (11:24 -0700)]
Merge branch 'Thunderbolt-networking'
Mika Westerberg says:
====================
Thunderbolt networking
In addition of tunneling PCIe, Display Port and USB traffic, Thunderbolt
allows connecting two hosts (domains) over a Thunderbolt cable. It is
possible to tunnel arbitrary data packets over such connection using
high-speed DMA rings available in the Thunderbolt host controller.
In order to discover Thunderbolt services the other host supports, there is
a software protocol running on top of the automatically configured control
channel (ring 0). This protocol is called XDomain discovery protocol and it
uses XDomain properties to describe the host (domain) and the services it
supports.
Once both sides have agreed what services are supported they can enable
high-speed DMA rings to transfer data over the cable.
This series adds support for the XDomain protocol so that we expose each
remote connection as Thunderbolt XDomain device and each service as
Thunderbolt service device. On top of that we create an API that allows
writing drivers for these services and finally we provide an example
Thunderbolt service driver that creates virtual ethernet inferface that
allows tunneling networking packets over Thunderbolt cable. The API could
be used for creating other future Thunderbolt services, such as tunneling
SCSI over Thunderbolt, for example.
The XDomain protocol and networking support is also available in macOS and
Windows so this makes it possible to connect Linux to macOS and Windows as
well.
The patches are based on previous Thunderbolt networking patch series by
Amir Levy and Michael Jamet, that can be found here:
https://lwn.net/Articles/705998/
The main difference to that patch series is that we have the XDomain
protocol running in the kernel now so there is no need for a separate
userspace daemon.
Note this does not affect the existing functionality, so security levels
and NVM firmware upgrade continue to work as before (with the small
exception that now sysfs also shows the XDomain connections and services in
addition to normal Thunderbolt devices). It is also possible to connect up
to 5 Thunderbolt devices and then another host, and the network driver
works exactly the same.
This is third version of the patch series. The previous versions can be
be found here:
v2: https://lkml.org/lkml/2017/9/25/225
v1: https://lwn.net/Articles/734019/
Changes from the v2:
* Add comment regarding calculation of interrupt throttling value
* Add UUIDs as strings in comments on top of each declaration
* Add a patch removing __packed from existing ICM messages. They are all
32-bit aligned and should pack fine without the __packed.
* Move adding MAINTAINERS entries to a separate patches
* Added Michael and Yehezkel to be maintainers of the network driver
* Remove __packed from the new ICM messages. They should pack fine as
well without it.
* Call register_netdev() after all other initialization is done in the
network driver.
* Use build_skb() instead of copying. We allocate order 1 page here to
leave room for SKB shared info required by build_skb(). However, we do
not leave room for full NET_SKB_PAD because the NHI hardware does not
cope well if a frame crosses 4kB boundary. According comments in
__build_skb() that should still be fine.
* Added Reviewed-by tag from Andy.
Changes from the v1:
* Add include/linux/thunderbolt.h to MAINTAINERS
* Correct Linux version and date of new sysfs entries in
Documentation/ABI/testing/sysfs-bus-thunderbolt
* Move network driver from drivers/thunderbolt/net.c to
drivers/net/thunderbolt.c and update it to follow coding style in
drivers/net/*.
* Add MAINTAINERS entry for the network driver
* Minor cleanups
In case someone wants to try this out, the last patch adds documentation
how the networking driver can be used. In short, if you connect Linux to a
macOS or Windows, everything is done automatically (as those systems have
the networking service enabled by default). For Linux to Linux connection
one host needs to load the networking driver first (so that the other side
can locate the networking service and load the corresponding driver).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:46 +0000 (13:38 +0300)]
MAINTAINERS: Add entry for Thunderbolt network driver
I will be maintaining the Thunderbolt network driver along with Michael
and Yehezkel.
Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Amir Levy [Mon, 2 Oct 2017 10:38:45 +0000 (13:38 +0300)]
net: Add support for networking over Thunderbolt cable
ThunderboltIP is a protocol created by Apple to tunnel IP/ethernet
traffic over a Thunderbolt cable. The protocol consists of configuration
phase where each side sends ThunderboltIP login packets (the protocol is
determined by UUID in the XDomain packet header) over the configuration
channel. Once both sides get positive acknowledgment to their login
packet, they configure high-speed DMA path accordingly. This DMA path is
then used to transmit and receive networking traffic.
This patch creates a virtual ethernet interface the host software can
use in the same way as any other networking interface. Once the
interface is brought up successfully network packets get tunneled over
the Thunderbolt cable to the remote host and back.
The connection is terminated by sending a ThunderboltIP logout packet
over the configuration channel. We do this when the network interface is
brought down by user or the driver is unloaded.
Signed-off-by: Amir Levy <amir.jer.levy@intel.com>
Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:44 +0000 (13:38 +0300)]
MAINTAINERS: Add thunderbolt.h to the Thunderbolt driver entry
The new API header (include/linux/thunderbolt.h) is maintained by the
Thunderbolt driver maintainers.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:43 +0000 (13:38 +0300)]
thunderbolt: Allocate ring HopID automatically if requested
Thunderbolt services should not care which HopID (ring) they use for
sending and receiving packets over the high-speed DMA path, so make
tb_ring_alloc_rx() and tb_ring_alloc_tx() accept negative HopID. This
means that the NHI will allocate next available HopID for the caller
automatically.
These HopIDs will be allocated from the range which is not reserved for
the Thunderbolt protocol (8 .. hop_count - 1).
The allocated HopID can be retrieved from ring->hop field after the ring
has been allocated successfully if needed.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:42 +0000 (13:38 +0300)]
thunderbolt: Add function to retrieve DMA device for the ring
This is needed when Thunderbolt service drivers need to DMA map memory
before it is passed down to the ring.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:41 +0000 (13:38 +0300)]
thunderbolt: Add polling mode for rings
In order to support things like networking over Thunderbolt cable, there
needs to be a way to switch the ring to a mode where it can be polled
with the interrupt masked. We implement such mode so that the caller can
allocate a ring by passing pointer to a function that is then called
when an interrupt is triggered. Completed frames can be fetched using
tb_ring_poll() and the interrupt can be re-enabled when the caller is
finished with polling by using tb_ring_poll_complete().
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:40 +0000 (13:38 +0300)]
thunderbolt: Use spinlock in NHI serialization
This is needed because ring polling functionality can be called from
atomic contexts when networking and other high-speed traffic is
transferred over a Thunderbolt cable.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:39 +0000 (13:38 +0300)]
thunderbolt: Use spinlock in ring serialization
This makes it possible to enqueue frames also from atomic context which
is needed for example, when networking packets are sent over a
Thunderbolt cable.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:38 +0000 (13:38 +0300)]
thunderbolt: Move ring descriptor flags to thunderbolt.h
A Thunderbolt service driver might need to check if there was an error
with the descriptor when in frame mode. We also add two Rx specific
error flags RING_DESC_CRC_ERROR and RING_DESC_BUFFER_OVERRUN.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:37 +0000 (13:38 +0300)]
thunderbolt: Export ring handling functions to modules
These are used by Thunderbolt services to send and receive frames over
the high-speed DMA rings.
We also put the functions to tb_ namespace to make sure we do not
collide with others and add missing kernel-doc comments for the exported
functions.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:36 +0000 (13:38 +0300)]
thunderbolt: Add support for frame mode
When high-speed DMA paths are used to transfer arbitrary data over a
Thunderbolt link, DMA rings should be in frame mode instead of raw mode.
The latter is used by the control channel (ring 0). In frame mode each
data frame can hold up to 4kB payload.
This patch modifies the DMA ring code to allow configuring a ring to be
in frame mode by passing a new flag (RING_FLAG_FRAME) to the ring when
it is allocated. In addition there might be need to enable end-to-end
(E2E) workaround for the ring to prevent losing Rx frames in certain
situations. We add another flag (RING_FLAG_E2E) that can be used for
this purpose.
This code is based on the work done by Amir Levy and Michael Jamet.
Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:35 +0000 (13:38 +0300)]
thunderbolt: Configure interrupt throttling for all interrupts
This will keep the interrupt delivery rate reasonable. The value used
here (128 us) is a recommendation from the hardware people.
This code is based on the work done by Amir Levy and Michael Jamet.
Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:34 +0000 (13:38 +0300)]
thunderbolt: Add support for XDomain discovery protocol
When two hosts are connected over a Thunderbolt cable, there is a
protocol they can use to communicate capabilities supported by the host.
The discovery protocol uses automatically configured control channel
(ring 0) and is build on top of request/response transactions using
special XDomain primitives provided by the Thunderbolt base protocol.
The capabilities consists of a root directory block of basic properties
used for identification of the host, and then there can be zero or more
directories each describing a Thunderbolt service and its capabilities.
Once both sides have discovered what is supported the two hosts can
setup high-speed DMA paths and transfer data to the other side using
whatever protocol was agreed based on the properties. The software
protocol used to communicate which DMA paths to enable is service
specific.
This patch adds support for the XDomain discovery protocol to the
Thunderbolt bus. We model each remote host connection as a Linux XDomain
device. For each Thunderbolt service found supported on the XDomain
device, we create Linux Thunderbolt service device which Thunderbolt
service drivers can then bind to based on the protocol identification
information retrieved from the property directory describing the
service.
This code is based on the work done by Amir Levy and Michael Jamet.
Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:33 +0000 (13:38 +0300)]
thunderbolt: Move tb_switch_phy_port_from_link() to thunderbolt.h
A Thunderbolt service might need to find the physical port from a link
the cable is connected to. For instance networking driver uses this
information to generate MAC address according the Apple ThunderboltIP
protocol.
Move this function to thunderbolt.h and rename it to
tb_phy_port_from_link() to reflect the fact that it does not take switch
as parameter.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:32 +0000 (13:38 +0300)]
thunderbolt: Move thunderbolt domain structure to thunderbolt.h
These are needed by Thunderbolt services so move them to thunderbolt.h
to make sure they are available outside of drivers/thunderbolt.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:31 +0000 (13:38 +0300)]
thunderbolt: Move enum tb_cfg_pkg_type to thunderbolt.h
These will be needed by Thunderbolt services when sending and receiving
XDomain control messages. While there change TB_CFG_PKG_PREPARE_TO_SLEEP
value to be decimal in order to be consistent with other members.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:30 +0000 (13:38 +0300)]
thunderbolt: Add support for XDomain properties
Thunderbolt XDomain discovery protocol uses directories which contain
properties and other directories to exchange information about what
capabilities the remote host supports. This also includes identification
information like device ID and name.
This adds support for parsing and formatting these properties and
establishes an API drivers can use in addition to the core Thunderbolt
driver. This API is exposed in a new header: include/linux/thunderbolt.h.
This code is based on the work done by Amir Levy and Michael Jamet.
Signed-off-by: Michael Jamet <michael.jamet@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:29 +0000 (13:38 +0300)]
thunderbolt: Remove __packed from ICM message structures
These messages are all 32-bit aligned and they should be packed without
the __packed attribute just fine. It also allows compiler to generate
better code on some architectures.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mika Westerberg [Mon, 2 Oct 2017 10:38:28 +0000 (13:38 +0300)]
byteorder: Move {cpu_to_be32, be32_to_cpu}_array() from Thunderbolt to core
We will be using these when communicating XDomain discovery protocol
over Thunderbolt link but they might be useful for other drivers as
well.
Make them available through byteorder/generic.h.
Suggested-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Michael Jamet <michael.jamet@intel.com>
Reviewed-by: Yehezkel Bernat <yehezkel.bernat@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 2 Oct 2017 18:20:22 +0000 (11:20 -0700)]
Merge branch 'mlxsw-Fixlets'
Jiri Pirko says:
====================
mlxsw: Fixlets
Couple of small nit fixes from Petr
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Petr Machata [Mon, 2 Oct 2017 10:21:58 +0000 (12:21 +0200)]
mlxsw: spectrum_router: Drop a redundant condition
Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Petr Machata [Mon, 2 Oct 2017 10:21:57 +0000 (12:21 +0200)]
mlxsw: spectrum_router: Fix a typo
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Westphal [Mon, 2 Oct 2017 10:05:29 +0000 (12:05 +0200)]
selftests: rtnetlink.sh: add vxlan and fou test cases
fou test lifted from ip-fou man page.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 2 Oct 2017 18:06:08 +0000 (11:06 -0700)]
Merge branch 'flow_dissector-dissect-tunnel-info'
Simon Horman says:
====================
flow_dissector: dissect tunnel info
Move dissection of tunnel info from the flower classifier to the flow
dissector where all other dissection occurs. This should not have any
behavioural affect on other users of the flow dissector.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Simon Horman [Mon, 2 Oct 2017 08:41:16 +0000 (10:41 +0200)]
flow_dissector: dissect tunnel info
Move dissection of tunnel info from the flower classifier to the flow
dissector where all other dissection occurs. This should not have any
behavioural affect on other users of the flow dissector.
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simon Horman [Mon, 2 Oct 2017 08:41:15 +0000 (10:41 +0200)]
net/dst: Make skb parameter of skb{metadata_dst, tunnel_info}() const
Make the skb parameter of skb_metadata_dst() and skb_tunnel_info()
const as they are not modified. This is in preparation for using
them in call-sites where skb is const.
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jacob Keller [Mon, 10 Jul 2017 20:23:13 +0000 (13:23 -0700)]
fm10k: prevent race condition of __FM10K_SERVICE_SCHED
Although very unlikely, it is possible that cancel_work_sync() may stop
the service_task before it actually started. In this case, the
__FM10K_SERVICE_SCHED bit will never be cleared. This results in the
service task being unable to reschedule in the future. Add a helper
function which sets the service disable bit, waits for the service task
to stop and clears the schedule bit, thus avoiding the race condition.
We know the schedule bit is safe to clear because the cancel_work_sync()
guarantees the service task is not running.
Add a helper function also to restart the service task, for symmetry.
This is not strictly needed but helps the mental model of how to stop
and start the service task.
This race could only happen in fm10k_suspend/fm10k_resume as this is the
only place where the service task is actually restarted. Thus,
suspend/resume testing would be ideal. However, note that the chance of
this happening is very slim as the service event is scheduled for
immediate execution, and you would have to trigger a suspend at almost
the exact same time as the service task was scheduled.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:12 +0000 (13:23 -0700)]
fm10k: move fm10k_prepare_for_reset and fm10k_handle_reset
A future patch needs these functions defined earlier in the file. Move
them closer to above where they will be called.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:11 +0000 (13:23 -0700)]
fm10k: avoid divide by zero in rare cases when device is resetting
It is possible that under rare circumstances the device is undergoing
a reset, such as when a PFLR occurs, and the device may be transmitting
simultaneously. In this case, we might attempt to divide by zero when
finding the proper r_idx. Instead, lets read the num_tx_queues once,
and make sure it's non-zero.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:10 +0000 (13:23 -0700)]
fm10k: don't loop while resetting VFs due to VFLR event
We've always had a really weird looping construction for resetting VFs.
We read the VFLRE register and reset the VF if the corresponding bit is
set, which makes sense. However we loop continuously until we no longer
have any bits left unset. At first this makes sense, as a sort of "keep
trying until we succeed" concept.
Unfortunately this causes a problem if we happen to surprise remove
while this code is executing, because in this case we'll always read all
1s for the VFLRE register. This results in a hard lockup on the CPU
because the loop will never terminate.
Because our own reset function will clear the VFLR event register
always, (except when we've lost PCIe link obviously) there is no real
reason to loop. In practice, we'll loop over once and find that no VFs
are pending anymore.
Lets just check once. Since we're clear the notification when we reset
there's no benefit to the loop. Additionally, there shouldn't be a race
as future VLFRE events should trigger an interrupt. Additionally, we
didn't warn or do anything in the looped case anyways.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:09 +0000 (13:23 -0700)]
fm10k: simplify reading PFVFLRE register
We're doing a really convoluted bitshift and read for the PFVFLRE
register. Just reading the PFVFLRE(1), shifting it by 32, then reading
PFVFLRE(0) should be sufficient.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:08 +0000 (13:23 -0700)]
fm10k: avoid needless delay when loading driver
When we load the driver, we set the last_reset to be in the future,
which delays the initial driver reset. Additionally, the service task
isn't scheduled to run automatically until the timer runs out. This
causes a needless delay of the first reset to begin talking to the
switch manager.
We can avoid this by simply not setting last_reset and immediately
scheduling the service task while in probe. This allows the device to
wake up faster, and avoids this delay.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:07 +0000 (13:23 -0700)]
fm10k: add missing fall through comment
Newer versions of GCC starting with 7 now additionally warn when a case
statement may fall through without an explicit comment mentioning it.
Add such a comment to silence the warning, as this is expected.
Unfortunately the comment must come directly before the next case
statement, so we put it outside the #ifdef. Otherwise, the compiler
cannot properly detect it and thus the warning is displayed regardless.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:06 +0000 (13:23 -0700)]
fm10k: avoid possible truncation of q_vector->name
New versions of GCC since version 7 began warning about possible
truncation of calls to snprintf. We can fix this and avoid false
positives. First, we should pass the full buffer size to snprintf,
because it guarantees a NULL character as part of its passed length, so
passing len-1 is simply wasting a byte of possible storage.
Second, if we make the ri and ti variables unsigned, the compiler is
able to correctly reason that the value never gets larger than 256, so
it doesn't need to warn about the full space required to print a signed
integer.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:05 +0000 (13:23 -0700)]
fm10k: fix typos on fall through comments
Newer versions of GCC since version 7 now warn when a case statement may
fall through without an explicit comment. "Fallthough" does not count as
it is misspelled. Fix the typos for these comments to appease the new
warnings.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 10 Jul 2017 20:23:04 +0000 (13:23 -0700)]
fm10k: stop spurious link down messages when Tx FIFO is full
In fm10k_get_host_state_generic, we check the mailbox tx_read() function
to ensure that the mailbox is still open. This function also checks to
make sure we have space to transmit another message. Unfortunately, if
we just recently sent a bunch of messages (such as enabling hundreds of
VLANs on a VF) this can result in a race where the watchdog task thinks
the link went down just because we haven't had time to process all these
messages yet.
Instead, lets just check whether the mailbox is still open. This ensures
that we don't race with the Tx FIFO, and we only link down once the
mailbox is not open.
This is safe, because if the FIFO fills up and we're unable to send
a message for too long, we'll end up triggering the timeout detection
which results in a reset. Additionally, since we still check to ensure
the mailbox state is OPEN, we'll transition to link down whenever the
mailbox closes as well.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Markus Elfring [Mon, 8 May 2017 16:18:09 +0000 (18:18 +0200)]
fm10k: Use seq_putc() in fm10k_dbg_desc_break()
Two single characters should be put into a sequence.
Thus use the corresponding function "seq_putc".
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Wed, 8 Mar 2017 23:55:43 +0000 (15:55 -0800)]
fm10k: reschedule service event if we stall the PF<->SM mailbox
When we are handling PF<->VF mailbox messages, it is possible that the
VF will send us so many messages that the PF<->SM FIFO will fill up. In
this case, we stop the loop and wait until the service event is
rescheduled.
Normally this should happen due to an interrupt. But it is possible that
we don't get another interrupt for a while and it isn't until the
service timer actually reschedules us. Instead, simply reschedule
immediately which will cause the service event to be run again as soon
as we exit.
This ensures that we promptly handle all of the PF<->VF messages with
minimal delay, while still giving time for the SM mailbox to drain.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Jacob Keller [Mon, 2 Oct 2017 14:17:50 +0000 (07:17 -0700)]
fm10k: ensure we process SM mbx when processing VF mbx
When we process VF mailboxes, the driver is likely going to also queue
up messages to the switch manager. This process merely queues up the
FIFO, but doesn't actually begin the transmission process. Because we
hold the mailbox lock during this VF processing, the PF<->SM mailbox is
not getting processed at this time. Ensure that we actually process the
PF<->SM mailbox in between each PF<->VF mailbox.
This should ensure prompt transmission of the messages queued up after
each VF message is received and handled.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Krishneil Singh <krishneil.k.singh@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Stephen Hemminger [Sun, 1 Oct 2017 21:07:34 +0000 (14:07 -0700)]
samples/bpf: fix warnings in xdp_monitor_user
Make local functions static to fix
HOSTCC samples/bpf/xdp_monitor_user.o
samples/bpf/xdp_monitor_user.c:64:7: warning: no previous prototype for ‘gettime’ [-Wmissing-prototypes]
__u64 gettime(void)
^~~~~~~
samples/bpf/xdp_monitor_user.c:209:6: warning: no previous prototype for ‘print_bpf_prog_info’ [-Wmissing-prototypes]
void print_bpf_prog_info(void)
^~~~~~~~~~~~~~~~~~~
Fixes: 3ffab5460264 ("samples/bpf: xdp_monitor tool based on tracepoints")
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Colin Ian King [Sun, 1 Oct 2017 16:27:35 +0000 (17:27 +0100)]
mlxsw: spectrum: fix uninitialized value in err
In the unlikely event that mfc->mfc_un.res.ttls[i] is 255 for all
values of i from 0 to MAXIVS-1, the err is not set at all and hence
has a garbage value on the error return at the end of the function,
so initialize it to 0. Also, the error return check on err and goto
to err: inside the for loop makes it impossible for err to be zero
at the end of the for loop, so we can remove the redundant err check
at the end of the loop.
Detected by CoverityScan CID#
1457207 ("Unitialized scalar value")
Fixes: c011ec1bbfd6 ("mlxsw: spectrum: Add the multicast routing offloading logic")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 2 Oct 2017 06:05:11 +0000 (23:05 -0700)]
Merge branch 'bcm63xx_enet-small-fixes-and-cleanups'
Jonas Gorski says:
====================
bcm63xx_enet: small fixes and cleanups
This patch set fixes a few theoretical issues and cleans up the code a
bit. It also adds a bit more managed function usage to simplify clock
and iomem usage.
Based on net-next.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Jonas Gorski [Sun, 1 Oct 2017 11:02:20 +0000 (13:02 +0200)]
bcm63xx_enet: remove unneeded include
We don't use anyhing from that file, so drop it.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jonas Gorski [Sun, 1 Oct 2017 11:02:19 +0000 (13:02 +0200)]
bcm63xx_enet: drop unneeded NULL phy_clk check
clk_disable and clk_unprepare are NULL-safe, so need to duplicate the
NULL check of the functions.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jonas Gorski [Sun, 1 Oct 2017 11:02:18 +0000 (13:02 +0200)]
bcm63xx_enet: use managed functions for clock/ioremap
Use managed functions where possible to reduce the amount of resource
handling on error and remove paths.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jonas Gorski [Sun, 1 Oct 2017 11:02:17 +0000 (13:02 +0200)]
bcm63xx_enet: do not rely on probe order
Do not rely on the shared device being probed before the enet(sw)
devices. This makes it easier to eventually move out the shared
device as a dma controller driver (what it should be).
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jonas Gorski [Sun, 1 Oct 2017 11:02:16 +0000 (13:02 +0200)]
bcm63xx_enet: do not write to random DMA channel on BCM6345
The DMA controller regs actually point to DMA channel 0, so the write to
ENETDMA_CFG_REG will actually modify a random DMA channel.
Since DMA controller registers do not exist on BCM6345, guard the write
with the usual check for dma_has_sram.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jonas Gorski [Sun, 1 Oct 2017 11:02:15 +0000 (13:02 +0200)]
bcm63xx_enet: correct clock usage
Check the return code of prepare_enable and change one last instance of
enable only to prepare_enable. Also properly disable and release the
clock in error paths and on remove for enetsw.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Maciej Żenczykowski [Wed, 27 Sep 2017 03:37:22 +0000 (20:37 -0700)]
net-ipv6: remove unused IP6_ECN_clear() function
This function is unused, and furthermore it is buggy since it suffers
from the same issue that requires IP6_ECN_set_ce() to take a pointer
to the skb so that it may (in case of CHECKSUM_COMPLETE) update skb->csum
Instead of fixing it, let's just outright remove it.
Tested: builds, and 'git grep IP6_ECN_clear' comes up empty
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Haishuang Yan [Wed, 27 Sep 2017 03:35:43 +0000 (11:35 +0800)]
ipv4: Namespaceify tcp_fastopen_blackhole_timeout knob
Different namespace application might require different time period in
second to disable Fastopen on active TCP sockets.
Tested:
Simulate following similar situation that the server's data gets dropped
after 3WHS.
C ---- syn-data ---> S
C <--- syn/ack ----- S
C ---- ack --------> S
S (accept & write)
C? X <- data ------ S
[retry and timeout]
And then print netstat of TCPFastOpenBlackhole, the counter increased as
expected when the firewall blackhole issue is detected and active TFO is
disabled.
# cat /proc/net/netstat | awk '{print $91}'
TCPFastOpenBlackhole
1
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Haishuang Yan [Wed, 27 Sep 2017 03:35:42 +0000 (11:35 +0800)]
ipv4: Namespaceify tcp_fastopen_key knob
Different namespace application might require different tcp_fastopen_key
independently of the host.
David Miller pointed out there is a leak without releasing the context
of tcp_fastopen_key during netns teardown. So add the release action in
exit_batch path.
Tested:
1. Container namespace:
# cat /proc/sys/net/ipv4/tcp_fastopen_key:
2817fff2-
f803cf97-
eadfd1f3-
78c0992b
cookie key in tcp syn packets:
Fast Open Cookie
Kind: TCP Fast Open Cookie (34)
Length: 10
Fast Open Cookie:
1e5dd82a8c492ca9
2. Host:
# cat /proc/sys/net/ipv4/tcp_fastopen_key:
107d7c5f-
68eb2ac7-
02fb06e6-
ed341702
cookie key in tcp syn packets:
Fast Open Cookie
Kind: TCP Fast Open Cookie (34)
Length: 10
Fast Open Cookie:
e213c02bf0afbc8a
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Haishuang Yan [Wed, 27 Sep 2017 03:35:41 +0000 (11:35 +0800)]
ipv4: Remove the 'publish' logic in tcp_fastopen_init_key_once
The 'publish' logic is not necessary after commit
dfea2aa65424 ("tcp:
Do not call tcp_fastopen_reset_cipher from interrupt context"), because
in tcp_fastopen_cookie_gen,it wouldn't call tcp_fastopen_init_key_once.
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Haishuang Yan [Wed, 27 Sep 2017 03:35:40 +0000 (11:35 +0800)]
ipv4: Namespaceify tcp_fastopen knob
Different namespace application might require enable TCP Fast Open
feature independently of the host.
This patch series continues making more of the TCP Fast Open related
sysctl knobs be per net-namespace.
Reported-by: Luca BRUNO <lucab@debian.org>
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Sun, 1 Oct 2017 03:15:08 +0000 (04:15 +0100)]
Merge branch 'dsa_ptr-port'
Vivien Didelot says:
====================
net: dsa: change dsa_ptr for a dsa_port
With DSA, a master net_device is physically wired to a dedicated CPU
switch port. For interaction with the DSA layer, the struct net_device
contains a dsa_ptr, which currently points to a dsa_switch_tree object.
This is only valid for a switch fabric with a single CPU port. In order
to support switch fabrics with multiple CPU ports, we first need to
change the type of dsa_ptr to what it really is: a dsa_port object.
This is what this patchset does. The first patches adds a
dsa_master_get_slave helper and cleans up portions of DSA core to make
the next patches more readable. These next patches prepare the xmit and
receive hot paths and finally change dsa_ptr.
Changes in v2:
- introduce dsa_master_get_slave helper to simplify patch 6
- keep hot path data at beginning of dsa_port for cacheline 1
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot [Fri, 29 Sep 2017 21:19:21 +0000 (17:19 -0400)]
net: dsa: remove tag ops from the switch tree
Now that the dsa_ptr is a dsa_port instance, there is no need to keep
the tag operations in the dsa_switch_tree structure. Remove it.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot [Fri, 29 Sep 2017 21:19:20 +0000 (17:19 -0400)]
net: dsa: change dsa_ptr for a dsa_port
With DSA, a master net device (CPU facing interface) has a dsa_ptr
pointer to which hangs a dsa_switch_tree. This is not correct because a
master interface is wired to a dedicated switch port, and because we can
theoretically have several master interfaces pointing to several CPU
ports of the same switch fabric.
Change the master interface's dsa_ptr for the CPU dsa_port pointer.
This is a step towards supporting multiple CPU ports.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot [Fri, 29 Sep 2017 21:19:19 +0000 (17:19 -0400)]
net: dsa: prepare master receive hot path
In preparation to make DSA master devices point to their corresponding
CPU port instead of the whole tree, add copies of dst and rcv in the
dsa_port structure so that we keep fast access in the receive hot path.
Also keep the copies at the beginning of the dsa_port structure in order
to ensure they are available in cacheline 1.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot [Fri, 29 Sep 2017 21:19:18 +0000 (17:19 -0400)]
net: dsa: add tagging ops to port
The DSA tagging protocol operations are specific to each CPU port,
thus the dsa_device_ops pointer belongs to the dsa_port structure.
>From now on assign a slave's xmit copy from its CPU port tagging
operations. This will ease the future support for multiple CPU ports.
Also keep the tag_ops at the beginning of the dsa_port structure so that
we ensure copies for hot path are in cacheline 1.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot [Fri, 29 Sep 2017 21:19:17 +0000 (17:19 -0400)]
net: dsa: use temporary dsa_device_ops variable
When resolving the DSA tagging protocol used by a CPU switch, use a
temporary "tag_ops" variable to store the dsa_device_ops instead of
using directly dst->tag_ops. This will make the future patches moving
this pointer around easier to read.
There is no functional changes.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot [Fri, 29 Sep 2017 21:19:16 +0000 (17:19 -0400)]
net: dsa: use cpu_dp in master code
Make it clear that the master device is linked to a CPU port by using
"cpu_dp" for the dsa_port variable in master.c instead of "port", then
use a "port" variable to describe the port index, as usually seen in
other places of DSA core.
This will make the future patch touching dsa_ptr more readable. There is
no functional changes.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vivien Didelot [Fri, 29 Sep 2017 21:19:15 +0000 (17:19 -0400)]
net: dsa: add master helper to look up slaves
The DSA tagging code does not need to know about the DSA architecture,
it only needs to return the slave device corresponding to the source
port index (and eventually the source device index for cascade-capable
switches) parsed from the frame received on the master device.
For this purpose, provide an inline dsa_master_get_slave helper which
validates the device and port indexes and look up the slave device.
This makes the tagging rcv functions more concise and robust, and also
makes dsa_get_cpu_port obsolete.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Colin Ian King [Fri, 29 Sep 2017 19:51:23 +0000 (20:51 +0100)]
net: hns3: fix null pointer dereference before null check
pointer ndev is being dereferenced with the call to netif_running
before it is being null checked. Re-order the code to only dereference
ndev after it has been null checked.
Detected by CoverityScan, CID#
1457206 ("Dereference before null check")
Fixes: 9df8f79a4d29 ("net: hns3: Add DCB support when interacting with network stack")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simon Xiao [Fri, 29 Sep 2017 18:39:46 +0000 (11:39 -0700)]
hv_netvsc: report stop_queue and wake_queue
Report the numbers of events for stop_queue and wake_queue in
ethtool stats.
Example:
ethtool -S eth0
NIC statistics:
...
stop_queue: 7
wake_queue: 7
...
Signed-off-by: Simon Xiao <sixiao@microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Martin KaFai Lau [Fri, 29 Sep 2017 17:52:17 +0000 (10:52 -0700)]
bpf: Fix compiler warning on info.map_ids for 32bit platform
This patch uses u64_to_user_ptr() to cast info.map_ids to a userspace ptr.
It also tags the user_map_ids with '__user' for sparse check.
Fixes: cb4d2b3f03d8 ("bpf: Add name, load_time, uid and map_ids to bpf_prog_info")
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Colin Ian King [Fri, 29 Sep 2017 14:01:16 +0000 (15:01 +0100)]
net_sched: remove redundant assignment to ret
The assignment of -EINVAL to variable ret is redundant as it
is being overwritten on the following error exit paths or
to the return value from the following call to basic_set_parms.
Fix this up by removing it. Cleans up clang warning message:
net/sched/cls_basic.c:185:2: warning: Value stored to 'err' is never read
Fixes: 1d8134fea2eb ("net_sched: use idr to allocate basic filter handles")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Colin Ian King [Fri, 29 Sep 2017 13:34:22 +0000 (14:34 +0100)]
net: ipmr: make function ipmr_notifier_init static
The function ipmr_notifier_init is local to the source and does
not need to be in global scope, so make it static.
Cleans up sparse warning:
warning: symbol 'ipmr_notifier_init' was not declared. Should it be static?
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mick Tarsel [Thu, 28 Sep 2017 20:53:18 +0000 (13:53 -0700)]
ibmvnic: Set state UP
State is initially reported as UNKNOWN. Before register call
netif_carrier_off(). Once the device is opened, call netif_carrier_on() in
order to set the state to UP.
Signed-off-by: Mick Tarsel <mjtarsel@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Thu, 28 Sep 2017 18:19:06 +0000 (11:19 -0700)]
Revert "net: dsa: bcm_sf2: Defer port enabling to calling port_enable"
This reverts commit
e85ec74ace29 ("net: dsa: bcm_sf2: Defer port
enabling to calling port_enable") because this now makes an unbind
followed by a bind to fail connecting to the ingrated PHY.
What this patch missed is that we need the PHY to be enabled with
bcm_sf2_gphy_enable_set() before probing it on the MDIO bus. This is
correctly done in the ops->setup() function, but by the time
ops->port_enable() runs, this is too late. Upon unbind we would power
down the PHY, and so when we would bind again, the PHY would be left
powered off.
Fixes: e85ec74ace29 ("net: dsa: bcm_sf2: Defer port enabling to calling port_enable")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Sun, 1 Oct 2017 02:31:17 +0000 (03:31 +0100)]
Merge branch '40GbE' of git://git./linux/kernel/git/jkirsher/next-queue
Jeff Kirsher says:
====================
40GbE Intel Wired LAN Driver Updates 2017-09-29
This series contains updates to i40e and i40evf only.
Jake provides several of the changes starting with the renaming of a
variable to clarify what the value is actually calculating. Found we
were misusing the __I40E_RECOVERY_PENDING bit to determine when we
should actually request a new IRQ in i40e_setup_misc_vector(), which
lead to a design mistake, so to resolve the issue, use a separate
state bit for miscellaneous IRQ setup and fix up the design while we
are at it. Cleaned up the old legacy PM support in the driver since
we support the newer generic PM callbacks. Fixed a failure to
hibernate issue, where on some platforms with a large number of CPUs,
we would allocate many IRQ vectors which we would try to migrate to
CPU0 when hibernating.
Sudheer cleans up a check for unqualified module inside i40e_up_complete()
because the link state information is in flux at time, so log messages
are getting logged with incorrect link state information. Also provided
additional log message cleanups and simplify member variable access in
the printing of the link messages.
Mariusz relaxes the firmware check since Fortville and Fort Park NICs
can and do have different firmware versions, so only warn for older
Fortville firmware. Fixed an errata with a flow director statistic that
was not wrapping as expected, simply reset after reading.
Mitch prevents consternation by lowering the log level to debug on a
message seen regularly on VF reset or unload, which is meaningless under
normal circumstances. Refactor the firmware version checking since
Fortville and Fort Park devices can have different firmware versions.
Alan fixes a ring to vector mapping, where the past implementation
attempted to map each Tx and Rx ring to its own vector, however we use
combined queues so we should be mapping the Tx/Rx rings together on one
vector. Adds the ability for the VF to request a different number of
queues allocated to it.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Colin Ian King [Wed, 27 Sep 2017 21:45:13 +0000 (22:45 +0100)]
mkiss: remove redundant check on len being zero
The check on len is redundant as it is always greater than 1,
so just remove it and make the printk less complex.
Detected by CoverityScan, CID#
1226729 ("Logically dead code")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Maciej Żenczykowski [Wed, 27 Sep 2017 04:32:42 +0000 (21:32 -0700)]
net-ipv6: add support for sockopt(SOL_IPV6, IPV6_FREEBIND)
So far we've been relying on sockopt(SOL_IP, IP_FREEBIND) being usable
even on IPv6 sockets.
However, it turns out it is perfectly reasonable to want to set freebind
on an AF_INET6 SOCK_RAW socket - but there is no way to set any SOL_IP
socket option on such a socket (they're all blindly errored out).
One use case for this is to allow spoofing src ip on a raw socket
via sendmsg cmsg.
Tested:
built, and booted
# python
>>> import socket
>>> SOL_IP = socket.SOL_IP
>>> SOL_IPV6 = socket.IPPROTO_IPV6
>>> IP_FREEBIND = 15
>>> IPV6_FREEBIND = 78
>>> s = socket.socket(socket.AF_INET6, socket.SOCK_DGRAM, 0)
>>> s.getsockopt(SOL_IP, IP_FREEBIND)
0
>>> s.getsockopt(SOL_IPV6, IPV6_FREEBIND)
0
>>> s.setsockopt(SOL_IPV6, IPV6_FREEBIND, 1)
>>> s.getsockopt(SOL_IP, IP_FREEBIND)
1
>>> s.getsockopt(SOL_IPV6, IPV6_FREEBIND)
1
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mike Manning [Mon, 25 Sep 2017 21:01:36 +0000 (22:01 +0100)]
net: ipv6: send NS for DAD when link operationally up
The NS for DAD are sent on admin up as long as a valid qdisc is found.
A race condition exists by which these packets will not egress the
interface if the operational state of the lower device is not yet up.
The solution is to delay DAD until the link is operationally up
according to RFC2863. Rather than only doing this, follow the existing
code checks by deferring IPv6 device initialization altogether. The fix
allows DAD on devices like tunnels that are controlled by userspace
control plane. The fix has no impact on regular deployments, but means
that there is no IPv6 connectivity until the port has been opened in
the case of port-based network access control, which should be
desirable.
Signed-off-by: Mike Manning <mmanning@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mitch Williams [Fri, 14 Jul 2017 13:27:09 +0000 (09:27 -0400)]
i40e: refactor FW version checking
The i40e driver now supports two different devices with two different
firmware versions. So be smart about how we handle these. Move the FW
version macros to the appropriate header file, and add a convenience
macro that checks the version based on the device. Then use this macro
to check whether or not the driver can use the new link info API.
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>
Alan Brady [Fri, 14 Jul 2017 13:27:08 +0000 (09:27 -0400)]
i40e: Enable VF to negotiate number of allocated queues
Currently the PF allocates a default number of queues for each VF and
cannot be changed. This patch enables the VF to request a different
number of queues allocated to it. This patch also adds a new virtchnl
op and capability flag to facilitate this negotiation.
After the PF receives a request message, it will set a requested number
of queues for that VF. Then when the VF resets, its VSI will get a new
number of queues allocated to it.
This is a best effort request and since we only allocate a guaranteed
default number, if the VF tries to ask for more than the guaranteed
number, there may not be enough in HW to accommodate it unless other
queues for other VFs are freed. It should also be noted decreasing the
number queues allocated to a VF to below the default will NOT enable the
allocation of more than 32 VFs per PF and will not free queues guaranteed
to each VF by default.
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>
Alan Brady [Fri, 14 Jul 2017 13:27:07 +0000 (09:27 -0400)]
i40evf: fix ring to vector mapping
The current implementation for mapping queues to vectors is broken
because it attempts to map each Tx and Rx ring to its own vector,
however we use combined queues so we should actually be mapping the
Tx/Rx rings together on one vector.
Also in the current implementation, in the case where we have more
queues than vectors, we attempt to group the queues together into
'chunks' and map each 'chunk' of queues to a vector. Chunking them
together would be more ideal if, and only if, we only had RSS because of
the way the hashing algorithm works but in the case of a future patch
that enables VF ADq, round robin assignment is better and still works
with RSS.
This patch resolves both those issues and simplifies the code needed to
accomplish this. Instead of treating the case where we have more queues
than vectors as special, if we notice our vector index is greater than
vectors, reset the vector index to zero and continue mapping. This
should ensure that in both cases, whether we have enough vectors for
each queue or not, the queues get appropriately mapped.
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>
Jacob Keller [Fri, 14 Jul 2017 13:27:06 +0000 (09:27 -0400)]
i40e: shutdown all IRQs and disable MSI-X when suspended
On some platforms with a large number of CPUs, we will allocate many IRQ
vectors. When hibernating, the system will attempt to migrate all of the
vectors back to CPU0 when shutting down all the other CPUs. It is
possible that we have so many vectors that it cannot re-assign them to
CPU0. This is even more likely if we have many devices installed in one
platform.
The end result is failure to hibernate, as it is not possible to
shutdown the CPUs. We can avoid this by disabling MSI-X and clearing our
interrupt scheme when the device is suspended. A more ideal solution
would be some method for the stack to properly handle this for all
drivers, rather than on a case-by-case basis for each driver to fix
itself.
However, until this more ideal solution exists, we can do our part and
shutdown our IRQs during suspend, which should allow systems with
a large number of CPUs to safely suspend or hibernate.
It may be worth investigating if we should shut down even further when
we suspend as it may make the path cleaner, but this was the minimum fix
for the hibernation issue mentioned here.
Testing-hints:
This affects systems with a large number of CPUs, and with multiple
devices enabled. Without this change, those platforms are unable to
hibernate at all.
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>
Jacob Keller [Fri, 14 Jul 2017 13:27:05 +0000 (09:27 -0400)]
i40e: prevent service task from running while we're suspended
Although the service task does check the suspended status before
running, it might already be part way through running when we go to
suspend. Lets ensure that the service task is stopped and will not be
restarted again until we finish resuming. This ensures that service task
code does not cause strange interactions with the suspend/resume
handlers.
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>
Jacob Keller [Fri, 14 Jul 2017 13:27:04 +0000 (09:27 -0400)]
i40e: don't clear suspended state until we finish resuming
When handling suspend and resume callbacks we want to make sure that (a)
we don't suspend again if we're already suspended and (b) we don't
resume again if we're already resuming. Lets make sure we test_and_set
the __I40E_SUSPENDED bit in i40e_suspend which ensures that a suspend
call when already suspended will exit early. Additionally, if
__I40E_SUSPENDED is not set when we begin resuming, exit early as well.
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>
Jacob Keller [Fri, 14 Jul 2017 13:27:03 +0000 (09:27 -0400)]
i40e: use newer generic PM support instead of legacy PM callbacks
Stop using the old legacy PM support, since we now have stable support
for the newer generic PM callbacks.
This has several advantages. First, we no longer have to manage our
own pci_save_state() and power changes, as it's preferred to have the
PCI stack do this. Second, these routines get called for both hibernate
and suspend to ram, so we can have the driver properly handle all the
suspend/resume flows that it needs to.
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>
Jacob Keller [Fri, 14 Jul 2017 13:27:02 +0000 (09:27 -0400)]
i40e: use separate state bit for miscellaneous IRQ setup
We currently (mis)use the __I40E_RECOVERY_PENDING bit to determine when
we should actually request a new IRQ in i40e_setup_misc_vector().
This led to a design mistake where we open-coded the re-setup of the
miscellaneous vector in i40e_resume() instead of using the function
provided. If we did not open-code this and instead tried to use the
i40e_setup_misc_vector() function, it would lead to never reallocating
the IRQ.
This would lead to a second i40e_suspend() call failing to free the
vector due to a NULL pointer dereference.
A future patch is going to re-work how the i40e_suspend() and
i40e_resume() flows work to clear all IRQ vectors, which would require
us to use i40e_setup_misc_vector() directly. Since during this time the
__I40E_RECOVERY_PENDING bit is set, we'll never re-allocate the vector.
Rather than leaving the open-coded setup in i40e_resume() lets just fix
the problem properly in i40e_setup_misc_vector().
Introduce a new state bit which indicates when the IRQ has been
assigned, which will be set when i40e_setup_misc_vector is first called.
This ultimately resolves the issue of re-requesting the vector, without
overloading the __I40E_RECOVERY_PENDING state. This ensures that the
suspend/resume cycle can use the setup function instead of open-coding
the re-request during resume.
Additionally, since the only callers of i40e_stop_misc_vector also want
to free it, move this code directly into the function to avoid
duplication. Due to the new functionality, rename it to
i40e_free_misc_vector().
This lets us drop the extra calls to free and re-enable the vector
during i40e_suspend() and i40e_resume(). We don't need to call
i40e_setup_misc_Vector() in i40e_resume() because it gets called by the
i40e_rebuild() call.
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>