linux-2.6-microblaze.git
4 years agoenetc: Remove unused variable 'enetc_drv_name'
YueHaibing [Thu, 19 Mar 2020 06:46:37 +0000 (14:46 +0800)]
enetc: Remove unused variable 'enetc_drv_name'

commit ed0a72e0de16 ("net/freescale: Clean drivers from static versions")
leave behind this, remove it .

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoCrypto/chtls: add/delete TLS header in driver
Rohit Maheshwari [Thu, 19 Mar 2020 04:41:21 +0000 (10:11 +0530)]
Crypto/chtls: add/delete TLS header in driver

Kernel TLS forms TLS header in kernel during encryption and removes
while decryption before giving packet back to user application. The
similar logic is introduced in chtls code as well.

v1->v2:
- tls_proccess_cmsg() uses tls_handle_open_record() which is not required
  in TOE-TLS. Don't mix TOE with other TLS types.

Signed-off-by: Vinay Kumar Yadav <vinay.yadav@chelsio.com>
Signed-off-by: Rohit Maheshwari <rohitm@chelsio.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: Make skb_segment not to compute checksum if network controller supports checksumming
Yadu Kishore [Tue, 17 Mar 2020 08:38:38 +0000 (14:08 +0530)]
net: Make skb_segment not to compute checksum if network controller supports checksumming

Problem:
TCP checksum in the output path is not being offloaded during GSO
in the following case:
The network driver does not support scatter-gather but supports
checksum offload with NETIF_F_HW_CSUM.

Cause:
skb_segment calls skb_copy_and_csum_bits if the network driver
does not announce NETIF_F_SG. It does not check if the driver
supports NETIF_F_HW_CSUM.
So for devices which might want to offload checksum but do not support SG
there is currently no way to do so if GSO is enabled.

Solution:
In skb_segment check if the network controller does checksum and if so
call skb_copy_bits instead of skb_copy_and_csum_bits.

Testing:
Without the patch, ran iperf TCP traffic with NETIF_F_HW_CSUM enabled
in the network driver. Observed the TCP checksum offload is not happening
since the skbs received by the driver in the output path have
skb->ip_summed set to CHECKSUM_NONE.

With the patch ran iperf TCP traffic and observed that TCP checksum
is being offloaded with skb->ip_summed set to CHECKSUM_PARTIAL.
Also tested with the patch by disabling NETIF_F_HW_CSUM in the driver
to cover the newly introduced if-else code path in skb_segment.

Link: https://lore.kernel.org/netdev/CA+FuTSeYGYr3Umij+Mezk9CUcaxYwqEe5sPSuXF8jPE2yMFJAw@mail.gmail.com
Signed-off-by: Yadu Kishore <kyk.segfault@gmail.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'net-hns3-add-three-optimizations-for-mailbox-handling'
David S. Miller [Sun, 22 Mar 2020 03:22:25 +0000 (20:22 -0700)]
Merge branch 'net-hns3-add-three-optimizations-for-mailbox-handling'

Huazhong Tan says:

====================
net: hns3: add three optimizations for mailbox handling

This patchset includes three code optimizations for mailbox handling.

[patch 1] adds a response code conversion.
[patch 2] refactors some structure definitions about PF and
VF mailbox.
[patch 3] refactors the condition whether PF responds VF's mailbox.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: hns3: refactor mailbox response scheme between PF and VF
Huazhong Tan [Wed, 18 Mar 2020 03:57:07 +0000 (11:57 +0800)]
net: hns3: refactor mailbox response scheme between PF and VF

Currently, PF responds to VF depending on what mailbox it is
handling, it is a bit inflexible. The correct way is, PF should
check the mbx_need_resp field to decide whether gives response
to VF.

Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: hns3: refactor the mailbox message between PF and VF
Yufeng Mo [Wed, 18 Mar 2020 03:57:06 +0000 (11:57 +0800)]
net: hns3: refactor the mailbox message between PF and VF

For making the code more readable, this adds several new
structure to replace the msg field in structure
hclge_mbx_vf_to_pf_cmd and hclge_mbx_pf_to_vf_cmd.
Also uses macro to instead of some magic number.

Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: hns3: add a conversion for mailbox's response code
Jian Shen [Wed, 18 Mar 2020 03:57:05 +0000 (11:57 +0800)]
net: hns3: add a conversion for mailbox's response code

Currently, when mailbox handling fails, the PF driver
just responds 1 to the VF driver. It is not sufficient
for the VF driver to find out why its mailbox fails.

So the error should be responded to VF, but the error
is type int and the response field in struct
hclge_mbx_pf_to_vf_cmd is type u16, a conversion is
needed.

Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomptcp: Remove set but not used variable 'can_ack'
YueHaibing [Wed, 18 Mar 2020 02:01:57 +0000 (02:01 +0000)]
mptcp: Remove set but not used variable 'can_ack'

Fixes gcc '-Wunused-but-set-variable' warning:

net/mptcp/options.c: In function 'mptcp_established_options_dss':
net/mptcp/options.c:338:7: warning:
 variable 'can_ack' set but not used [-Wunused-but-set-variable]

commit dc093db5cc05 ("mptcp: drop unneeded checks")
leave behind this unused, remove it.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'selftests-expand-txtimestamp-with-new-features'
David S. Miller [Sun, 22 Mar 2020 03:14:13 +0000 (20:14 -0700)]
Merge branch 'selftests-expand-txtimestamp-with-new-features'

Jian Yang says:

====================
selftests: expand txtimestamp with new features

Current txtimestamp selftest issues requests with no delay, or fixed 50
usec delay. Nsec granularity is useful to measure fine-grained latency.
A configurable delay is useful to simulate the case with cold
cachelines.

This patchset adds new flags and features to the txtimestamp selftest,
including:
- Printing in nsec (-N)
- Polling interval (-b, -S)
- Using epoll (-E, -e)
- Printing statistics
- Running individual tests in txtimestamp.sh
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: txtimestamp: print statistics for timestamp events.
Jian Yang [Tue, 17 Mar 2020 19:25:09 +0000 (12:25 -0700)]
selftests: txtimestamp: print statistics for timestamp events.

Statistics on timestamps is useful to quantify average and tail latency.

Print timestamp statistics in count/avg/min/max format.

Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: txtimestamp: add support for epoll().
Jian Yang [Tue, 17 Mar 2020 19:25:08 +0000 (12:25 -0700)]
selftests: txtimestamp: add support for epoll().

Add the following new flags:
-e: use level-triggered epoll() instead of poll().
-E: use event-triggered epoll() instead of poll().

Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: txtimestamp: add new command-line flags.
Jian Yang [Tue, 17 Mar 2020 19:25:07 +0000 (12:25 -0700)]
selftests: txtimestamp: add new command-line flags.

A longer sleep duration between sendmsg()s makes more cachelines to be
evicted and results in higher latency. Making the duration configurable.

Add the following new flags:
-S: Configurable sleep duration.
-b: Busy loop instead of poll().

Remove the following flag:
-D: No delay between packets: subsumed by -S.

Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: txtimestamp: allow printing latencies in nsec.
Jian Yang [Tue, 17 Mar 2020 19:25:06 +0000 (12:25 -0700)]
selftests: txtimestamp: allow printing latencies in nsec.

Txtimestamp reports latencies in uses resolution, while nsec is needed
in cases such as measuring latencies on localhost.

Add the following new flag:
-N: print timestamps and durations in nsec (instead of usec)

Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: txtimestamp: allow individual txtimestamp tests.
Jian Yang [Tue, 17 Mar 2020 19:25:05 +0000 (12:25 -0700)]
selftests: txtimestamp: allow individual txtimestamp tests.

The wrapper script txtimestamp.sh executes a pre-defined list of testcases
sequentially without configuration options available.

Add an option (-r/--run) to setup the test namespace and pass remaining
arguments to txtimestamp binary. The script still runs all tests when no
argument is passed.

Signed-off-by: Jian Yang <jianyang@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'net-tls-Annotate-lockless-access-to-sk_prot'
David S. Miller [Sun, 22 Mar 2020 03:08:17 +0000 (20:08 -0700)]
Merge branch 'net-tls-Annotate-lockless-access-to-sk_prot'

Jakub Sitnicki says:

====================
net/tls: Annotate lockless access to sk_prot

We have recently noticed that there is a case of lockless read/write to
sk->sk_prot [0]. sockmap code on psock tear-down writes to sk->sk_prot,
while holding sk_callback_lock. Concurrently, tcp can access it. Usually to
read out the sk_prot pointer and invoke one of the ops,
sk->sk_prot->handler().

The lockless write (lockless in regard to concurrent reads) happens on the
following paths:

tcp_bpf_{recvmsg|sendmsg} / sock_map_unref
  sk_psock_put
    sk_psock_drop
      sk_psock_restore_proto
        WRITE_ONCE(sk->sk_prot, proto)

To prevent load/store tearing [1], and to make tooling aware of intentional
shared access [2], we need to annotate sites that access sk_prot with
READ_ONCE/WRITE_ONCE.

This series kicks off the effort to do it. Starting with net/tls.

[0] https://lore.kernel.org/bpf/a6bf279e-a998-84ab-4371-cd6c1ccbca5d@gmail.com/
[1] https://lwn.net/Articles/793253/
[2] https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/tls: Annotate access to sk_prot with READ_ONCE/WRITE_ONCE
Jakub Sitnicki [Tue, 17 Mar 2020 17:04:39 +0000 (18:04 +0100)]
net/tls: Annotate access to sk_prot with READ_ONCE/WRITE_ONCE

sockmap performs lockless writes to sk->sk_prot on the following paths:

tcp_bpf_{recvmsg|sendmsg} / sock_map_unref
  sk_psock_put
    sk_psock_drop
      sk_psock_restore_proto
        WRITE_ONCE(sk->sk_prot, proto)

To prevent load/store tearing [1], and to make tooling aware of intentional
shared access [2], we need to annotate other sites that access sk_prot with
READ_ONCE/WRITE_ONCE macros.

Change done with Coccinelle with following semantic patch:

@@
expression E;
identifier I;
struct sock *sk;
identifier sk_prot =~ "^sk_prot$";
@@
(
 E =
-sk->sk_prot
+READ_ONCE(sk->sk_prot)
|
-sk->sk_prot = E
+WRITE_ONCE(sk->sk_prot, E)
|
-sk->sk_prot
+READ_ONCE(sk->sk_prot)
 ->I
)

Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/tls: Read sk_prot once when building tls proto ops
Jakub Sitnicki [Tue, 17 Mar 2020 17:04:38 +0000 (18:04 +0100)]
net/tls: Read sk_prot once when building tls proto ops

Apart from being a "tremendous" win when it comes to generated machine
code (see bloat-o-meter output for x86-64 below) this mainly prepares
ground for annotating access to sk_prot with READ_ONCE, so that we don't
pepper the code with access annotations and needlessly repeat loads.

add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-46 (-46)
Function                                     old     new   delta
tls_init                                     851     805     -46
Total: Before=21063, After=21017, chg -0.22%

Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet/tls: Constify base proto ops used for building tls proto
Jakub Sitnicki [Tue, 17 Mar 2020 17:04:37 +0000 (18:04 +0100)]
net/tls: Constify base proto ops used for building tls proto

The helper that builds kTLS proto ops doesn't need to and should not modify
the base proto ops. Annotate the parameter as read-only.

Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'ionic-error-recovery-fixes'
David S. Miller [Sun, 22 Mar 2020 02:56:04 +0000 (19:56 -0700)]
Merge branch 'ionic-error-recovery-fixes'

Shannon Nelson says:

====================
ionic error recovery fixes

These are a few little patches to make error recovery a little
more safe and successful.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: check for NULL structs on teardown
Shannon Nelson [Fri, 20 Mar 2020 02:31:53 +0000 (19:31 -0700)]
ionic: check for NULL structs on teardown

Make sure the queue structs exist before trying to tear
them down to make for safer error recovery.

Fixes: 0f3154e6bcb3 ("ionic: Add Tx and Rx handling")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: clean irq affinity on queue deinit
Shannon Nelson [Fri, 20 Mar 2020 02:31:52 +0000 (19:31 -0700)]
ionic: clean irq affinity on queue deinit

Add a little more cleanup when tearing down the queues.

Fixes: 1d062b7b6f64 ("ionic: Add basic adminq support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: ignore eexist on rx filter add
Shannon Nelson [Fri, 20 Mar 2020 02:31:51 +0000 (19:31 -0700)]
ionic: ignore eexist on rx filter add

Don't worry if the rx filter add firmware request fails on
EEXIST, at least we know the filter is there.  Same for
the delete request, at least we know it isn't there.

Fixes: 2a654540be10 ("ionic: Add Rx filter and rx_mode ndo support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: only save good lif dentry
Shannon Nelson [Fri, 20 Mar 2020 02:31:50 +0000 (19:31 -0700)]
ionic: only save good lif dentry

Don't save the lif->dentry until we know we have
a good value.

Fixes: 1a58e196467f ("ionic: Add basic lif support")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: leave dev cmd request contents alone on FW timeout
Shannon Nelson [Fri, 20 Mar 2020 02:31:49 +0000 (19:31 -0700)]
ionic: leave dev cmd request contents alone on FW timeout

It is possible (but unlikely) that FW was busy and missed a heartbeat
check but is still alive and will process the pending request, so don't
clean the dev_cmd in this case.  This occasionally occurs when working
with a card that is supporting many devices and is trying to shut them
all down at once, but still wants to see that last LIF disable request.

Fixes: 97ca486592c0 ("ionic: add heartbeat check")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoionic: add timeout error checking for queue disable
Shannon Nelson [Fri, 20 Mar 2020 02:31:48 +0000 (19:31 -0700)]
ionic: add timeout error checking for queue disable

Short circuit the cleanup if we get a timeout error from
ionic_qcq_disable() so as to not have to wait too long
on shutdown when we already know the FW is not responding.

Fixes: 0f3154e6bcb3 ("ionic: Add Tx and Rx handling")
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agosoc: qcom: ipa: kill IPA_RX_BUFFER_ORDER
Alex Elder [Fri, 20 Mar 2020 16:02:20 +0000 (11:02 -0500)]
soc: qcom: ipa: kill IPA_RX_BUFFER_ORDER

Don't assume the receive buffer size is a power-of-2 number of pages.
Instead, define the receive buffer size independently, and then
compute the page order from that size when needed.

This fixes a build problem that arises when the ARM64_PAGE_SHIFT
config option is set to have a page size greater than 4KB.  The
problem was identified by Linux Kernel Functional Testing.

The IPA code basically assumed the page size to be 4KB.  A larger page
size caused the receive buffer size to become correspondingly larger
(32KB or 128KB for ARM64_16K_PAGES and ARM64_64K_PAGES, respectively).
The receive buffer size is used to compute an "aggregation byte limit"
value that gets programmed into the hardware, and the large page sizes
caused that limit value to be too big to fit in a 5 bit field.  This
triggered a BUILD_BUG_ON() call in ipa_endpoint_validate_build().

This fix causes a lot of receive buffer memory to be wasted if
system is configured for page size greater than 4KB.  But such a
misguided configuration will now build successfully.

Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Alex Elder <elder@linaro.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'mac80211-next-for-net-next-2020-03-20' of git://git.kernel.org/pub/scm...
David S. Miller [Fri, 20 Mar 2020 15:57:38 +0000 (08:57 -0700)]
Merge tag 'mac80211-next-for-net-next-2020-03-20' of git://git./linux/kernel/git/jberg/mac80211-next

Johannes Berg says:

====================
Another set of changes:
 * HE ranging (fine timing measurement) API support
 * hwsim gets virtio support, for use with wmediumd,
   to be able to simulate with multiple machines
 * eapol-over-nl80211 improvements to exclude preauth
 * IBSS reset support, to recover connections from
   userspace
 * and various others.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: dsa: sja1105: Add support for the SGMII port
Vladimir Oltean [Fri, 20 Mar 2020 11:29:37 +0000 (13:29 +0200)]
net: dsa: sja1105: Add support for the SGMII port

SJA1105 switches R and S have one SerDes port with an 802.3z
quasi-compatible PCS, hardwired on port 4. The other ports are still
MII/RMII/RGMII. The PCS performs rate adaptation to lower link speeds;
the MAC on this port is hardwired at gigabit. Only full duplex is
supported.

The SGMII port can be configured as part of the static config tables, as
well as through a dedicated SPI address region for its pseudo-clause-22
registers. However it looks like the static configuration is not
able to change some out-of-reset values (like the value of MII_BMCR), so
at the end of the day, having code for it is utterly pointless. We are
just going to use the pseudo-C22 interface.

Because the PCS gets reset when the switch resets, we have to add even
more restoration logic to sja1105_static_config_reload, otherwise the
SGMII port breaks after operations such as enabling PTP timestamping
which require a switch reset.

>From PHYLINK perspective, the switch supports *only* SGMII (it doesn't
support 1000Base-X). It also doesn't expose access to the raw config
word for in-band AN in registers MII_ADV/MII_LPA.
It is able to work in the following modes:
 - Forced speed
 - SGMII in-band AN slave (speed received from PHY)
 - SGMII in-band AN master (acting as a PHY)

The latter mode is not supported by this patch. It is even unclear to me
how that would be described. There is some code for it left in the
patch, but 'an_master' is always passed as false.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'net-bridge-vlan-options-nest-the-tunnel-options'
David S. Miller [Fri, 20 Mar 2020 15:52:20 +0000 (08:52 -0700)]
Merge branch 'net-bridge-vlan-options-nest-the-tunnel-options'

Nikolay Aleksandrov says:

====================
net: bridge: vlan options: nest the tunnel options

After a discussion with Roopa about the new tunnel vlan option, she
suggested that we'll be adding more tunnel options and attributes, so
it'd be better to have them all grouped together under one main vlan
entry tunnel attribute instead of making them all main attributes. Since
the tunnel code was added in this net-next cycle and still hasn't been
released we can easily nest the BRIDGE_VLANDB_ENTRY_TUNNEL_ID attribute
in BRIDGE_VLANDB_ENTRY_TUNNEL_INFO and allow for any new tunnel
attributes to be added there. In addition one positive side-effect is
that we can remove the outside vlan info flag which controlled the
operation (setlink/dellink) and move it under a new nested attribute so
user-space can specify it explicitly.

Thus the vlan tunnel format becomes:
 [BRIDGE_VLANDB_ENTRY]
     [BRIDGE_VLANDB_ENTRY_TUNNEL_INFO]
         [BRIDGE_VLANDB_TINFO_ID]
         [BRIDGE_VLANDB_TINFO_CMD]
         ...
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: bridge: vlan options: move the tunnel command to the nested attribute
Nikolay Aleksandrov [Fri, 20 Mar 2020 11:23:03 +0000 (13:23 +0200)]
net: bridge: vlan options: move the tunnel command to the nested attribute

Now that we have a nested tunnel info attribute we can add a separate
one for the tunnel command and require it explicitly from user-space. It
must be one of RTM_SETLINK/DELLINK. Only RTM_SETLINK requires a valid
tunnel id, DELLINK just removes it if it was set before. This allows us
to have all tunnel attributes and control in one place, thus removing
the need for an outside vlan info flag.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: bridge: vlan options: nest the tunnel id into a tunnel info attribute
Nikolay Aleksandrov [Fri, 20 Mar 2020 11:23:02 +0000 (13:23 +0200)]
net: bridge: vlan options: nest the tunnel id into a tunnel info attribute

While discussing the new API, Roopa mentioned that we'll be adding more
tunnel attributes and options in the future, so it's better to make it a
nested attribute, since this is still in net-next we can easily change it
and nest the tunnel id attribute under BRIDGE_VLANDB_ENTRY_TUNNEL_INFO.

The new format is:
 [BRIDGE_VLANDB_ENTRY]
     [BRIDGE_VLANDB_ENTRY_TUNNEL_INFO]
         [BRIDGE_VLANDB_TINFO_ID]

Any new tunnel attributes can be nested under
BRIDGE_VLANDB_ENTRY_TUNNEL_INFO.

Suggested-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomac80211: driver can remain on channel if not using chan_ctx
Yan-Hsuan Chuang [Thu, 12 Mar 2020 07:43:37 +0000 (15:43 +0800)]
mac80211: driver can remain on channel if not using chan_ctx

Some of the drivers are not using channel context, but let the
stack to control/switch channels instead. For such cases, driver
can still remain on channel because the mac80211 stack actually
supports it.

The stack will check if the driver is using chan_ctx and has
ops->remain_on_channel been hooked. Otherwise it will start its
ROC work to remain on channel. So, even if the driver is not
using chan_ctx, the driver is still capable of doing remain on
channel.

Signed-off-by: Yan-Hsuan Chuang <yhchuang@realtek.com>
Link: https://lore.kernel.org/r/20200312074337.16198-1-yhchuang@realtek.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonl80211: clarify code in nl80211_del_station()
Johannes Berg [Fri, 20 Mar 2020 10:38:35 +0000 (11:38 +0100)]
nl80211: clarify code in nl80211_del_station()

The long if chain of interface types is hard to read,
especially now with the additional condition after it.
Use a switch statement to clarify this code.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20200320113834.2c51b9e8e341.I3fa5dc3f7d3cb1dbbd77191d764586f7da993f3f@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agocfg80211: Configure PMK lifetime and reauth threshold for PMKSA entries
Veerendranath Jakkam [Thu, 12 Mar 2020 23:59:03 +0000 (01:59 +0200)]
cfg80211: Configure PMK lifetime and reauth threshold for PMKSA entries

Drivers that trigger roaming need to know the lifetime of the configured
PMKSA for deciding whether to trigger the full or PMKSA cache based
authentication. The configured PMKSA is invalid after the PMK lifetime
has expired and must not be used after that and the STA needs to
disassociate if the PMK expires. Hence the STA is expected to refresh
the PMK with a full authentication before this happens (e.g., when
reassociating to a new BSS the next time or by performing EAPOL
reauthentication depending on the AKM) to avoid unnecessary
disconnection.

The PMK reauthentication threshold is the percentage of the PMK lifetime
value and indicates to the driver to trigger a full authentication roam
(without PMKSA caching) after the reauthentication threshold time, but
before the PMK timer has expired. Authentication methods like SAE need
to be able to generate a new PMKSA entry without having to force a
disconnection after this threshold timeout. If no roaming occurs between
the reauthentication threshold time and PMK lifetime expiration,
disassociation is still forced.

The new attributes for providing these values correspond to the dot11
MIB variables dot11RSNAConfigPMKLifetime and
dot11RSNAConfigPMKReauthThreshold.

This type of functionality is already available in cases where user
space component is in control of roaming. This commit extends that same
capability into cases where parts or all of this functionality is
offloaded to the driver.

Signed-off-by: Veerendranath Jakkam <vjakkam@codeaurora.org>
Signed-off-by: Jouni Malinen <jouni@codeaurora.org>
Link: https://lore.kernel.org/r/20200312235903.18462-1-jouni@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: Read rx_stats with perCPU pointers
Seevalamuthu Mariappan [Wed, 18 Mar 2020 10:15:55 +0000 (15:45 +0530)]
mac80211: Read rx_stats with perCPU pointers

Use perCPU pointers to get rx_stats in sta_set_sinfo
when RSS is enabled

Signed-off-by: Seevalamuthu Mariappan <seevalam@codeaurora.org>
Link: https://lore.kernel.org/r/1584526555-25960-1-git-send-email-seevalam@codeaurora.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agovirt_wifi: implement ndo_get_iflink
Taehee Yoo [Thu, 5 Mar 2020 09:06:36 +0000 (09:06 +0000)]
virt_wifi: implement ndo_get_iflink

->ndo_get_iflink() is useful for finding lower interface.

Test commands:
    ip link add dummy0 type dummy
    ip link add vw1 link dummy0 type virt_wifi
    ip link show vw1

Before:
    9: vw1: <BROADCAST,MULTICAST> ...
After:
    9: vw1@dummy0: <BROADCAST,MULTICAST> ...

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Link: https://lore.kernel.org/r/20200305090636.28221-1-ap420073@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: Allow deleting stations in ibss mode to reset their state
Nicolas Cavallari [Thu, 5 Mar 2020 13:57:54 +0000 (14:57 +0100)]
mac80211: Allow deleting stations in ibss mode to reset their state

Set the NL80211_EXT_FEATURE_DEL_IBSS_STA if the interface support IBSS
mode, so that stations can be reset from user space.

mac80211 already deletes stations by itself, so mac80211 drivers must
already support this.

This has been successfully tested with ath9k.

Signed-off-by: Nicolas Cavallari <nicolas.cavallari@green-communications.fr>
Link: https://lore.kernel.org/r/20200305135754.12094-2-cavallar@lri.fr
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agocfg80211: Add support for userspace to reset stations in IBSS mode
Nicolas Cavallari [Thu, 5 Mar 2020 13:57:53 +0000 (14:57 +0100)]
cfg80211: Add support for userspace to reset stations in IBSS mode

Sometimes, userspace is able to detect that a peer silently lost its
state (like, if the peer reboots). wpa_supplicant does this for IBSS-RSN
by registering for auth/deauth frames, but when it detects this, it is
only able to remove the encryption keys of the peer and close its port.

However, the kernel also hold other state about the station, such as BA
sessions, probe response parameters and the like.  They also need to be
resetted correctly.

This patch adds the NL80211_EXT_FEATURE_DEL_IBSS_STA feature flag
indicating the driver accepts deleting stations in IBSS mode, which
should send a deauth and reset the state of the station, just like in
mesh point mode.

Signed-off-by: Nicolas Cavallari <nicolas.cavallari@green-communications.fr>
Link: https://lore.kernel.org/r/20200305135754.12094-1-cavallar@lri.fr
[preserve -EINVAL return]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: consider WLAN_EID_EXT_HE_OPERATION for parsing CRC
Johannes Berg [Fri, 31 Jan 2020 11:12:58 +0000 (13:12 +0200)]
mac80211: consider WLAN_EID_EXT_HE_OPERATION for parsing CRC

We use the parsing CRC for checking if the beacon changed, and
if the WLAN_EID_EXT_HE_OPERATION extended element changes we
need to track it so we can react to that. Include it in the CRC
calculation.

Link: https://lore.kernel.org/r/20200131111300.891737-22-luca@coelho.fi
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonl80211: add PROTECTED_TWT nl80211 extended feature
Shaul Triebitz [Fri, 31 Jan 2020 11:12:59 +0000 (13:12 +0200)]
nl80211: add PROTECTED_TWT nl80211 extended feature

Add API for telling whether the driver supports protected TWT.
The protected_twt capability in the RSNXE will be based on this.

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/20200131111300.891737-23-luca@coelho.fi
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: HE: set missing bss_conf fields in AP mode
Shaul Triebitz [Fri, 31 Jan 2020 11:12:55 +0000 (13:12 +0200)]
mac80211: HE: set missing bss_conf fields in AP mode

In AP mode, set htc_trig_based_pkt_ext and frame_time_rts_th
for driver use.

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/20200131111300.891737-19-luca@coelho.fi
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonl80211: pass HE operation element to the driver
Shaul Triebitz [Fri, 31 Jan 2020 11:12:54 +0000 (13:12 +0200)]
nl80211: pass HE operation element to the driver

Pass the AP's HE operation element to the driver.

Signed-off-by: Shaul Triebitz <shaul.triebitz@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/20200131111300.891737-18-luca@coelho.fi
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonl80211/cfg80211: add support for non EDCA based ranging measurement
Avraham Stern [Fri, 31 Jan 2020 11:12:38 +0000 (13:12 +0200)]
nl80211/cfg80211: add support for non EDCA based ranging measurement

Add support for requesting that the ranging measurement will use
the trigger-based / non trigger-based flow instead of the EDCA based
flow.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Link: https://lore.kernel.org/r/20200131111300.891737-2-luca@coelho.fi
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: don't leave skb->next/prev pointing to stack
Johannes Berg [Fri, 20 Mar 2020 09:20:23 +0000 (10:20 +0100)]
mac80211: don't leave skb->next/prev pointing to stack

In beacon protection, don't leave skb->next/prev pointing to the
on-stack list, even if that's actually harmless since we don't use
them again afterwards.

While at it, check that the SKB on the list is still the same, as
that's required here. If not, the encryption (protection) code is
buggy.

Fixes: 0a3a84360b37 ("mac80211: Beacon protection using the new BIGTK (AP)")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20200320102021.1be7823fc05e.Ia89fb79a0469d32137c9a04315a1d2dfc7b7d6f5@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: update documentation about tx power
Qiujun Huang [Thu, 12 Mar 2020 14:44:24 +0000 (22:44 +0800)]
mac80211: update documentation about tx power

The structure member added at some point, but the kernel-doc was not
updated.

Signed-off-by: Qiujun Huang <hqjagain@gmail.com>
Link: https://lore.kernel.org/r/20200312144424.3023-1-hqjagain@gmail.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211: handle no-preauth flag for control port
Markus Theil [Thu, 12 Mar 2020 09:10:54 +0000 (10:10 +0100)]
mac80211: handle no-preauth flag for control port

This patch adds support for disabling pre-auth rx over the nl80211 control
port for mac80211.

Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
Link: https://lore.kernel.org/r/20200312091055.54257-3-markus.theil@tu-ilmenau.de
[fix indentation slightly, squash feature enablement]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonl80211: add no pre-auth attribute and ext. feature flag for ctrl. port
Markus Theil [Thu, 12 Mar 2020 09:10:53 +0000 (10:10 +0100)]
nl80211: add no pre-auth attribute and ext. feature flag for ctrl. port

If the nl80211 control port is used before this patch, pre-auth frames
(0x88c7) are send to userspace uncoditionally. While this enables userspace
to only use nl80211 on the station side, it is not always useful for APs.
Furthermore, pre-auth frames are ordinary data frames and not related to
the control port. Therefore it should for example be possible for pre-auth
frames to be bridged onto a wired network on AP side without touching
userspace.

For backwards compatibility to code already using pre-auth over nl80211,
this patch adds a feature flag to disable this behavior, while it remains
enabled by default. An additional ext. feature flag is added to detect this
from userspace.

Thanks to Jouni for pointing out, that pre-auth frames should be handled as
ordinary data frames.

Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
Link: https://lore.kernel.org/r/20200312091055.54257-2-markus.theil@tu-ilmenau.de
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agomac80211_hwsim: add frame transmission support over virtio
Erel Geron [Thu, 5 Mar 2020 13:32:14 +0000 (14:32 +0100)]
mac80211_hwsim: add frame transmission support over virtio

This allows communication with external entities.

It also required fixing up the netlink policy, since NLA_UNSPEC
attributes are no longer accepted.

Signed-off-by: Erel Geron <erelx.geron@intel.com>
[port to backports, inline the ID, use 29 as the ID as requested,
 drop != NULL checks, reduce ifdefs]
Link: https://lore.kernel.org/r/20200305143212.c6e4c87d225b.I7ce60bf143e863dcdf0fb8040aab7168ba549b99@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agocfg80211: fix documentation format
Johannes Berg [Fri, 20 Mar 2020 13:41:11 +0000 (14:41 +0100)]
cfg80211: fix documentation format

Kernel-doc complains if the line isn't prefixed with an
asterisk, fix that.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20200320144110.2786ad5fb234.I369d103d11c71e39e3a3f97ed68a528c5b875f1e@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
4 years agonet: dsa: sja1105: Avoid error message for unknown PHY mode on disabled ports
Vladimir Oltean [Thu, 19 Mar 2020 20:12:10 +0000 (22:12 +0200)]
net: dsa: sja1105: Avoid error message for unknown PHY mode on disabled ports

When sja1105_init_mii_settings iterates over the port list, it prints
this message for disabled ports, because they don't have a valid
phy-mode:

[    4.778702] sja1105 spi2.0: Unsupported PHY mode unknown!

Suggested-by: Andrew Lunn <andrew@lunn.ch>
Suggested-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vivien Didelot <vivien.didelot@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Fri, 20 Mar 2020 04:33:27 +0000 (21:33 -0700)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
pull request: bluetooth-next 2020-03-19

Here's the main bluetooth-next pull request for the 5.7 kernel.

 - Added wideband speech support to mgmt and the ability for HCI drivers
   to declare support for it.
 - Added initial support for L2CAP Enhanced Credit Based Mode
 - Fixed suspend handling for several use cases
 - Fixed Extended Advertising related issues
 - Added support for Realtek 8822CE device
 - Added DT bindings for QTI chip WCN3991
 - Cleanups to replace zero-length arrays with flexible-array members
 - Several other smaller cleanups & fixes
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'veth-xdp-stats'
David S. Miller [Fri, 20 Mar 2020 04:24:59 +0000 (21:24 -0700)]
Merge branch 'veth-xdp-stats'

Lorenzo Bianconi says:

====================
add more xdp stats to veth driver

Align veth xdp stats accounting to mellanox, intel and marvell
implementation. Introduce the following xdp counters:
- rx_xdp_tx
- rx_xdp_tx_errors
- tx_xdp_xmit
- tx_xdp_xmit_errors
- rx_xdp_redirect
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoveth: remove atomic64_add from veth_xdp_xmit hotpath
Lorenzo Bianconi [Thu, 19 Mar 2020 16:41:29 +0000 (17:41 +0100)]
veth: remove atomic64_add from veth_xdp_xmit hotpath

Remove atomic64_add from veth_xdp_xmit hotpath and rely on
xdp_xmit_err/xdp_tx_err counters

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoveth: introduce more xdp counters
Lorenzo Bianconi [Thu, 19 Mar 2020 16:41:28 +0000 (17:41 +0100)]
veth: introduce more xdp counters

Introduce xdp_xmit counter in order to distinguish between XDP_TX and
ndo_xdp_xmit stats. Introduce the following ethtool counters:
- rx_xdp_tx
- rx_xdp_tx_errors
- tx_xdp_xmit
- tx_xdp_xmit_errors
- rx_xdp_redirect

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoveth: distinguish between rx_drops and xdp_drops
Lorenzo Bianconi [Thu, 19 Mar 2020 16:41:27 +0000 (17:41 +0100)]
veth: distinguish between rx_drops and xdp_drops

Distinguish between rx_drops and xdp_drops since the latter is already
reported in rx_packets. Report xdp_drops in ethtool statistics

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoveth: introduce more specialized counters in veth_stats
Lorenzo Bianconi [Thu, 19 Mar 2020 16:41:26 +0000 (17:41 +0100)]
veth: introduce more specialized counters in veth_stats

Introduce xdp_tx, xdp_redirect and rx_drops counters in veth_stats data
structure. Move stats accounting in veth_poll. Remove xdp_xmit variable
in veth_xdp_rcv_one/veth_xdp_rcv_skb and rely on veth_stats counters.
This is a preliminary patch to align veth xdp statistics to mlx, intel
and marvell xdp implementation

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoveth: move xdp stats in a dedicated structure
Lorenzo Bianconi [Thu, 19 Mar 2020 16:41:25 +0000 (17:41 +0100)]
veth: move xdp stats in a dedicated structure

Move xdp stats in veth_stats data structure. This is a preliminary patch
to align xdp statistics to mlx5, ixgbe and mvneta drivers

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agosysfs: fix static inline declaration of sysfs_groups_change_owner()
Christian Brauner [Thu, 19 Mar 2020 14:47:41 +0000 (15:47 +0100)]
sysfs: fix static inline declaration of sysfs_groups_change_owner()

The CONFIG_SYSFS declaration of sysfs_group_change_owner() is different
from the !CONFIG_SYSFS version and thus causes build failurs when
!CONFIG_SYSFS is set.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 303a42769c4c ("sysfs: add sysfs_group{s}_change_owner()")
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'mscc-RGMII'
David S. Miller [Fri, 20 Mar 2020 04:14:46 +0000 (21:14 -0700)]
Merge branch 'mscc-RGMII'

Antoine Tenart says:

====================
net: phy: mscc: add support for RGMII MAC mode

This series adds support for the RGMII MAC mode for the VSC8584 PHY
family and for RGMII_ID modes (Tx and/or Rx).

I decided to drop the custom delay for now. I made some tests and it
seemed to be working quite well. If we find out we really need to lower
the delay, which I doubt, I'll send support for it.

Since v2:
  - Dropped support for custom dt bindings.
  - Add the 2ns delay based on the interface mode.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mscc: RGMII skew delay configuration
Antoine Tenart [Thu, 19 Mar 2020 14:19:58 +0000 (15:19 +0100)]
net: phy: mscc: RGMII skew delay configuration

This patch adds support for configuring the RGMII skew delays in Rx and
Tx. The Rx and Tx skews are set based on the interface mode. By default
their configuration is set to the default value in hardware (0.2ns);
this means the driver do not rely anymore on the bootloader
configuration.

Then based on the interface mode being used, a 2ns delay is added:
- RGMII_ID adds it for both Rx and Tx.
- RGMII_RXID adds it for Rx.
- RGMII_TXID adds it for Tx.

Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mscc: add support for RGMII MAC mode
Antoine Tenart [Thu, 19 Mar 2020 14:19:57 +0000 (15:19 +0100)]
net: phy: mscc: add support for RGMII MAC mode

This patch adds support for connecting VSC8584 PHYs to the MAC using
RGMII.

Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'mlxsw-Offload-TC-action-skbedit-priority'
David S. Miller [Fri, 20 Mar 2020 04:09:20 +0000 (21:09 -0700)]
Merge branch 'mlxsw-Offload-TC-action-skbedit-priority'

Ido Schimmel says:

====================
mlxsw: Offload TC action skbedit priority

Petr says:

The TC action "skbedit priority P" has the effect of assigning skbprio of P
to SKBs that it's applied on. In HW datapath of a switch, the corresponding
action is assignment of internal switch priority. Spectrum switches allow
setting of packet priority based on an ACL action, which is good match for
the skbedit priority gadget. This patchset therefore implements offloading
of this action to the Spectrum ACL engine.

After a bit of refactoring in patch #1, patch #2 extends the skbedit action
to support offloading of "priority" subcommand.

On mlxsw side, in patch #3, the QOS_ACTION flexible action is added, with
fields necessary for priority adjustment. In patch #4, "skbedit priority"
is connected to that action.

Patch #5 implements a new forwarding selftest, suitable for both SW- and
HW-datapath testing.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: forwarding: Add an skbedit priority selftest
Petr Machata [Thu, 19 Mar 2020 13:47:24 +0000 (15:47 +0200)]
selftests: forwarding: Add an skbedit priority selftest

Add a test that runs traffic through a port such that skbedit priority
action acts on it during forwarding. Test that at egress, it is classified
correctly according to the new priority at a PRIO qdisc.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_flower: Offload FLOW_ACTION_PRIORITY
Petr Machata [Thu, 19 Mar 2020 13:47:23 +0000 (15:47 +0200)]
mlxsw: spectrum_flower: Offload FLOW_ACTION_PRIORITY

Offload action skbedit priority when keyed to a flower classifier. The
skb->priority field in Linux is very generic, so only allow setting the
bottom 8 priorities and bounce anything else.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: core: Add QOS_ACTION
Petr Machata [Thu, 19 Mar 2020 13:47:22 +0000 (15:47 +0200)]
mlxsw: core: Add QOS_ACTION

The QOS_ACTION is used for manipulating the QoS attributes of a packet.
Add the corresponding defines and helpers, in particular for the
switch_priority override.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: tc_skbedit: Make the skbedit priority offloadable
Petr Machata [Thu, 19 Mar 2020 13:47:21 +0000 (15:47 +0200)]
net: tc_skbedit: Make the skbedit priority offloadable

The skbedit action "priority" is used for adjusting SKB priority. Allow
drivers to offload the action by introducing two new skbedit getters and a
new flow action, and initializing appropriately in tc_setup_flow_action().

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: tc_skbedit: Factor a helper out of is_tcf_skbedit_{mark, ptype}()
Petr Machata [Thu, 19 Mar 2020 13:47:20 +0000 (15:47 +0200)]
net: tc_skbedit: Factor a helper out of is_tcf_skbedit_{mark, ptype}()

The two functions is_tcf_skbedit_mark() and is_tcf_skbedit_ptype() have a
very similar structure. A follow-up patch will add one more such function.
Instead of more cut'n'pasting, extract a helper function that checks
whether a TC action is an skbedit with the required flag. Convert the two
existing functions into thin wrappers around the helper.

Signed-off-by: Petr Machata <petrm@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: stmmac: dwmac_lib: remove unnecessary checks in dwmac_dma_reset()
Dejin Zheng [Thu, 19 Mar 2020 13:16:38 +0000 (21:16 +0800)]
net: stmmac: dwmac_lib: remove unnecessary checks in dwmac_dma_reset()

it will check the return value of dwmac_dma_reset() in the
stmmac_init_dma_engine() function and report an error if the
return value is not zero. so don't need check here.

Signed-off-by: Dejin Zheng <zhengdejin5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: mscc: add missing check on a phy_write return value
Antoine Tenart [Thu, 19 Mar 2020 12:48:19 +0000 (13:48 +0100)]
net: phy: mscc: add missing check on a phy_write return value

Commit a5afc1678044 ("net: phy: mscc: add support for VSC8584 PHY")
introduced a call to 'phy_write' storing its return value to a variable
called 'ret'. But 'ret' never was checked for a possible error being
returned, and hence was not used at all. Fix this by checking the return
value and exiting the function if an error was returned.

As this does not fix a known bug, this commit is mostly cosmetic and not
sent as a fix.

Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ipa: Remove unused including <linux/version.h>
YueHaibing [Thu, 19 Mar 2020 12:12:00 +0000 (12:12 +0000)]
net: ipa: Remove unused including <linux/version.h>

Remove including <linux/version.h> that don't need it.

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: ipa: fix platform_no_drv_owner.cocci warnings
YueHaibing [Thu, 19 Mar 2020 12:11:27 +0000 (12:11 +0000)]
net: ipa: fix platform_no_drv_owner.cocci warnings

Remove .owner field if calls are used which set it automatically
Generated by: scripts/coccinelle/api/platform_no_drv_owner.cocci

Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoliquidio: remove set but not used variable 's'
YueHaibing [Thu, 19 Mar 2020 12:07:43 +0000 (12:07 +0000)]
liquidio: remove set but not used variable 's'

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/cavium/liquidio/lio_main.c: In function 'octeon_chip_specific_setup':
drivers/net/ethernet/cavium/liquidio/lio_main.c:1378:8: warning:
 variable 's' set but not used [-Wunused-but-set-variable]

It's not used since commit b6334be64d6f ("net/liquidio: Delete driver version assignment")

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: sched: Do not assume RTNL is held in tunnel key action helpers
Ido Schimmel [Thu, 19 Mar 2020 11:33:10 +0000 (13:33 +0200)]
net: sched: Do not assume RTNL is held in tunnel key action helpers

The cited commit removed RTNL from tc_setup_flow_action(), but the
function calls two tunnel key action helpers that use rtnl_dereference()
to fetch the action's parameters. This leads to "suspicious RCU usage"
warnings [1][2].

Change the helpers to use rcu_dereference_protected() while requiring
the action's lock to be held. This is safe because the two helpers are
only called from tc_setup_flow_action() which acquires the lock.

[1]
[  156.950855] =============================
[  156.955463] WARNING: suspicious RCU usage
[  156.960085] 5.6.0-rc5-custom-47426-gdfe43878d573 #2409 Not tainted
[  156.967116] -----------------------------
[  156.971728] include/net/tc_act/tc_tunnel_key.h:31 suspicious rcu_dereference_protected() usage!
[  156.981583]
[  156.981583] other info that might help us debug this:
[  156.981583]
[  156.990675]
[  156.990675] rcu_scheduler_active = 2, debug_locks = 1
[  156.998205] 1 lock held by tc/877:
[  157.002187]  #0: ffff8881cbf7bea0 (&(&p->tcfa_lock)->rlock){+...}, at: tc_setup_flow_action+0xbe/0x4f78
[  157.012866]
[  157.012866] stack backtrace:
[  157.017886] CPU: 2 PID: 877 Comm: tc Not tainted 5.6.0-rc5-custom-47426-gdfe43878d573 #2409
[  157.027253] Hardware name: Mellanox Technologies Ltd. MSN2100-CB2FO/SA001017, BIOS 5.6.5 06/07/2016
[  157.037389] Call Trace:
[  157.040170]  dump_stack+0xfd/0x178
[  157.044034]  lockdep_rcu_suspicious+0x14a/0x153
[  157.049157]  tc_setup_flow_action+0x89f/0x4f78
[  157.054227]  fl_hw_replace_filter+0x375/0x640
[  157.064348]  fl_change+0x28ec/0x4f6b
[  157.088843]  tc_new_tfilter+0x15e2/0x2260
[  157.176801]  rtnetlink_rcv_msg+0x8d6/0xb60
[  157.190915]  netlink_rcv_skb+0x177/0x460
[  157.208884]  rtnetlink_rcv+0x21/0x30
[  157.212925]  netlink_unicast+0x5d0/0x7f0
[  157.227728]  netlink_sendmsg+0x981/0xe90
[  157.245416]  ____sys_sendmsg+0x76d/0x8f0
[  157.255348]  ___sys_sendmsg+0x10f/0x190
[  157.320308]  __sys_sendmsg+0x115/0x1f0
[  157.342553]  __x64_sys_sendmsg+0x7d/0xc0
[  157.346987]  do_syscall_64+0xc1/0x600
[  157.351142]  entry_SYSCALL_64_after_hwframe+0x49/0xbe

[2]
[  157.432346] =============================
[  157.436937] WARNING: suspicious RCU usage
[  157.441537] 5.6.0-rc5-custom-47426-gdfe43878d573 #2409 Not tainted
[  157.448559] -----------------------------
[  157.453204] include/net/tc_act/tc_tunnel_key.h:43 suspicious rcu_dereference_protected() usage!
[  157.463042]
[  157.463042] other info that might help us debug this:
[  157.463042]
[  157.472112]
[  157.472112] rcu_scheduler_active = 2, debug_locks = 1
[  157.479529] 1 lock held by tc/877:
[  157.483442]  #0: ffff8881cbf7bea0 (&(&p->tcfa_lock)->rlock){+...}, at: tc_setup_flow_action+0xbe/0x4f78
[  157.494119]
[  157.494119] stack backtrace:
[  157.499114] CPU: 2 PID: 877 Comm: tc Not tainted 5.6.0-rc5-custom-47426-gdfe43878d573 #2409
[  157.508485] Hardware name: Mellanox Technologies Ltd. MSN2100-CB2FO/SA001017, BIOS 5.6.5 06/07/2016
[  157.518628] Call Trace:
[  157.521416]  dump_stack+0xfd/0x178
[  157.525293]  lockdep_rcu_suspicious+0x14a/0x153
[  157.530425]  tc_setup_flow_action+0x993/0x4f78
[  157.535505]  fl_hw_replace_filter+0x375/0x640
[  157.545650]  fl_change+0x28ec/0x4f6b
[  157.570204]  tc_new_tfilter+0x15e2/0x2260
[  157.658199]  rtnetlink_rcv_msg+0x8d6/0xb60
[  157.672315]  netlink_rcv_skb+0x177/0x460
[  157.690278]  rtnetlink_rcv+0x21/0x30
[  157.694320]  netlink_unicast+0x5d0/0x7f0
[  157.709129]  netlink_sendmsg+0x981/0xe90
[  157.726813]  ____sys_sendmsg+0x76d/0x8f0
[  157.736725]  ___sys_sendmsg+0x10f/0x190
[  157.801721]  __sys_sendmsg+0x115/0x1f0
[  157.823967]  __x64_sys_sendmsg+0x7d/0xc0
[  157.828403]  do_syscall_64+0xc1/0x600
[  157.832558]  entry_SYSCALL_64_after_hwframe+0x49/0xbe

Fixes: b15e7a6e8d31 ("net: sched: don't take rtnl lock during flow_action setup")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: bridge: vlan: include stats in dumps if requested
Nikolay Aleksandrov [Thu, 19 Mar 2020 10:14:14 +0000 (12:14 +0200)]
net: bridge: vlan: include stats in dumps if requested

This patch adds support for vlan stats to be included when dumping vlan
information. We have to dump them only when explicitly requested (thus the
flag below) because that disables the vlan range compression and will make
the dump significantly larger. In order to request the stats to be
included we add a new dump attribute called BRIDGE_VLANDB_DUMP_FLAGS which
can affect dumps with the following first flag:
  - BRIDGE_VLANDB_DUMPF_STATS
The stats are intentionally nested and put into separate attributes to make
it easier for extending later since we plan to add per-vlan mcast stats,
drop stats and possibly STP stats. This is the last missing piece from the
new vlan API which makes the dumped vlan information complete.

A dump request which should include stats looks like:
 [BRIDGE_VLANDB_DUMP_FLAGS] |= BRIDGE_VLANDB_DUMPF_STATS

A vlandb entry attribute with stats looks like:
 [BRIDGE_VLANDB_ENTRY] = {
     [BRIDGE_VLANDB_ENTRY_STATS] = {
         [BRIDGE_VLANDB_STATS_RX_BYTES]
         [BRIDGE_VLANDB_STATS_RX_PACKETS]
         ...
     }
 }

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomptcp: rename fourth ack field
Paolo Abeni [Thu, 19 Mar 2020 10:06:30 +0000 (11:06 +0100)]
mptcp: rename fourth ack field

The name is misleading, it actually tracks the 'fully established'
status.

Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge tag 'mlx5-updates-2020-03-17' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Thu, 19 Mar 2020 02:13:37 +0000 (19:13 -0700)]
Merge tag 'mlx5-updates-2020-03-17' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2020-03-17

1) Compiler warnings and cleanup for the connection tracking series
2) Bug fixes for the connection tracking series
3) Fix devlink port register sequence
4) Last five patches in the series, By Eli cohen
   Add the support for forwarding traffic between two eswitch uplink
   representors (Hairpin for eswitch), using mlx5 termination tables
   to change the direction of a packet in hw from RX to TX pipeline.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: phy: realtek: read actual speed to detect downshift
Heiner Kallweit [Wed, 18 Mar 2020 22:07:24 +0000 (23:07 +0100)]
net: phy: realtek: read actual speed to detect downshift

At least some integrated PHY's in RTL8168/RTL8125 chip versions support
downshift, and the actual link speed can be read from a vendor-specific
register. Info about this register was provided by Realtek.
More details about downshift configuration (e.g. number of attempts)
aren't available, therefore the downshift tunable is not implemented.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonet: sched: Fix hw_stats_type setting in pedit loop
Petr Machata [Wed, 18 Mar 2020 17:42:29 +0000 (19:42 +0200)]
net: sched: Fix hw_stats_type setting in pedit loop

In the commit referenced below, hw_stats_type of an entry is set for every
entry that corresponds to a pedit action. However, the assignment is only
done after the entry pointer is bumped, and therefore could overwrite
memory outside of the entries array.

The reason for this positioning may have been that the current entry's
hw_stats_type is already set above, before the action-type dispatch.
However, if there are no more actions, the assignment is wrong. And if
there are, the next round of the for_each_action loop will make the
assignment before the action-type dispatch anyway.

Therefore fix this issue by simply reordering the two lines.

Fixes: 74522e7baae2 ("net: sched: set the hw_stats_type in pedit loop")
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 'mlxsw-spectrum_cnt-Expose-counter-resources'
David S. Miller [Wed, 18 Mar 2020 23:46:20 +0000 (16:46 -0700)]
Merge branch 'mlxsw-spectrum_cnt-Expose-counter-resources'

Ido Schimmel says:

====================
mlxsw: spectrum_cnt: Expose counter resources

Jiri says:

Capacity and utilization of existing flow and RIF counters are currently
unavailable to be seen by the user. Use the existing devlink resources
API to expose the information:

$ sudo devlink resource show pci/0000:00:10.0 -v
pci/0000:00:10.0:
  name kvd resource_path /kvd size 524288 unit entry dpipe_tables none
  name span_agents resource_path /span_agents size 8 occ 0 unit entry dpipe_tables none
  name counters resource_path /counters size 79872 occ 44 unit entry dpipe_tables none
    resources:
      name flow resource_path /counters/flow size 61440 occ 4 unit entry dpipe_tables none
      name rif resource_path /counters/rif size 18432 occ 40 unit entry dpipe_tables none
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: mlxsw: Add tc action hw_stats tests
Jiri Pirko [Wed, 18 Mar 2020 13:48:57 +0000 (15:48 +0200)]
selftests: mlxsw: Add tc action hw_stats tests

Add tests for mlxsw hw_stats types.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_cnt: Expose devlink resource occupancy for counters
Jiri Pirko [Wed, 18 Mar 2020 13:48:56 +0000 (15:48 +0200)]
mlxsw: spectrum_cnt: Expose devlink resource occupancy for counters

Implement occupancy counting for counters and expose over devlink
resource API.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_cnt: Consolidate subpools initialization
Jiri Pirko [Wed, 18 Mar 2020 13:48:55 +0000 (15:48 +0200)]
mlxsw: spectrum_cnt: Consolidate subpools initialization

Put all init operations related to subpools into
mlxsw_sp_counter_sub_pools_init().

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_cnt: Move config validation along with resource register
Jiri Pirko [Wed, 18 Mar 2020 13:48:54 +0000 (15:48 +0200)]
mlxsw: spectrum_cnt: Move config validation along with resource register

Move the validation of subpools configuration, to avoid possible over
commitment to resource registration. Add WARN_ON to indicate bug
in the code.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_cnt: Expose subpool sizes over devlink resources
Jiri Pirko [Wed, 18 Mar 2020 13:48:53 +0000 (15:48 +0200)]
mlxsw: spectrum_cnt: Expose subpool sizes over devlink resources

Implement devlink resources support for counter pools. Move the subpool
sizes calculations into the new resources register function.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_cnt: Add entry_size_res_id for each subpool and use it to query entry...
Jiri Pirko [Wed, 18 Mar 2020 13:48:52 +0000 (15:48 +0200)]
mlxsw: spectrum_cnt: Add entry_size_res_id for each subpool and use it to query entry size

Add new field to subpool struct that would indicate which
resource id should be used to query the entry size for
the subpool from the device.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_cnt: Move sub_pools under per-instance pool struct
Jiri Pirko [Wed, 18 Mar 2020 13:48:51 +0000 (15:48 +0200)]
mlxsw: spectrum_cnt: Move sub_pools under per-instance pool struct

Currently, the global static array of subpools is used. Make it
per-instance as multiple instances of the mlxsw driver can have
different values.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoselftests: spectrum-2: Adjust tc_flower_scale limit according to current counter...
Jiri Pirko [Wed, 18 Mar 2020 13:48:50 +0000 (15:48 +0200)]
selftests: spectrum-2: Adjust tc_flower_scale limit according to current counter count

With the change that made the code to query counter bank size from device
instead of using hard-coded value, the number of available counters
changed for Spectrum-2. Adjust the limit in the selftests.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agomlxsw: spectrum_cnt: Query bank size from FW resources
Jiri Pirko [Wed, 18 Mar 2020 13:48:49 +0000 (15:48 +0200)]
mlxsw: spectrum_cnt: Query bank size from FW resources

The bank size is different between Spectrum versions. Also it is
a resource that can be queried. So instead of hard coding the value in
code, query it from the firmware.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agocxgb4: rework TC filter rule insertion across regions
Rahul Lakkireddy [Wed, 18 Mar 2020 10:54:51 +0000 (16:24 +0530)]
cxgb4: rework TC filter rule insertion across regions

Chelsio NICs have 3 filter regions, in following order of priority:
1. High Priority (HPFILTER) region (Highest Priority).
2. HASH region.
3. Normal FILTER region (Lowest Priority).

Currently, there's a 1-to-1 mapping between the prio value passed
by TC and the filter region index. However, it's possible to have
multiple TC rules with the same prio value. In this case, if a region
is exhausted, no attempt is made to try inserting the rule in the
next available region.

So, rework and remove the 1-to-1 mapping. Instead, dynamically select
the region to insert the filter rule, as long as the new rule's prio
value doesn't conflict with existing rules across all the 3 regions.

Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agonetfilter: revert introduction of egress hook
Daniel Borkmann [Wed, 18 Mar 2020 09:33:22 +0000 (10:33 +0100)]
netfilter: revert introduction of egress hook

This reverts the following commits:

  8537f78647c0 ("netfilter: Introduce egress hook")
  5418d3881e1f ("netfilter: Generalize ingress hook")
  b030f194aed2 ("netfilter: Rename ingress hook include file")

>From the discussion in [0], the author's main motivation to add a hook
in fast path is for an out of tree kernel module, which is a red flag
to begin with. Other mentioned potential use cases like NAT{64,46}
is on future extensions w/o concrete code in the tree yet. Revert as
suggested [1] given the weak justification to add more hooks to critical
fast-path.

  [0] https://lore.kernel.org/netdev/cover.1583927267.git.lukas@wunner.de/
  [1] https://lore.kernel.org/netdev/20200318.011152.72770718915606186.davem@davemloft.net/

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: David Miller <davem@davemloft.net>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Nacked-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agoMerge branch 's390-qeth-next'
David S. Miller [Wed, 18 Mar 2020 23:33:36 +0000 (16:33 -0700)]
Merge branch 's390-qeth-next'

Julian Wiedmann says:

====================
s390/qeth: updates 2020-03-18

please apply the following patch series for qeth to netdev's net-next
tree.

This consists of three parts:
1) support for __GFP_MEMALLOC,
2) several ethtool enhancements (.set_channels, SW Timestamping),
3) the usual cleanups.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: use dev->reg_state
Julian Wiedmann [Wed, 18 Mar 2020 12:54:55 +0000 (13:54 +0100)]
s390/qeth: use dev->reg_state

To check whether a netdevice has already been registered, look at
NETREG_REGISTERED to replace some hacks I added a while ago.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: remove gratuitous NULL checks
Julian Wiedmann [Wed, 18 Mar 2020 12:54:54 +0000 (13:54 +0100)]
s390/qeth: remove gratuitous NULL checks

qeth_do_ioctl() is only reached through our own net_device_ops, so we
can trust that dev->ml_priv still contains what we put there earlier.

qeth_bridgeport_an_set() is an internal function that doesn't require
such sanity checks.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: add phys_to_virt() translation for AOB
Julian Wiedmann [Wed, 18 Mar 2020 12:54:53 +0000 (13:54 +0100)]
s390/qeth: add phys_to_virt() translation for AOB

Data addresses in the AOB are absolute, and need to be translated before
being fed into kmem_cache_free(). Currently this phys_to_virt() is a no-op.
Also see commit 2db01da8d25f ("s390/qdio: fill SBALEs with absolute addresses").

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: don't report hard-coded driver version
Julian Wiedmann [Wed, 18 Mar 2020 12:54:52 +0000 (13:54 +0100)]
s390/qeth: don't report hard-coded driver version

Versions are meaningless for an in-kernel driver.
Instead use the UTS_RELEASE that is set by ethtool_get_drvinfo().

Cc: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: add SW timestamping support for IQD devices
Julian Wiedmann [Wed, 18 Mar 2020 12:54:51 +0000 (13:54 +0100)]
s390/qeth: add SW timestamping support for IQD devices

This adds support for SOF_TIMESTAMPING_TX_SOFTWARE.
No support for non-IQD devices, since they orphan the skb in their xmit
path.

To play nice with TX bulking, set the timestamp when the buffer that
contains the skb(s) is actually flushed out to HW.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: balance the TX queue selection for IQD devices
Julian Wiedmann [Wed, 18 Mar 2020 12:54:50 +0000 (13:54 +0100)]
s390/qeth: balance the TX queue selection for IQD devices

For ucast traffic, qeth_iqd_select_queue() falls back to
netdev_pick_tx(). This will potentially use skb_tx_hash() to distribute
the flow over all active TX queues - so txq 0 is a valid selection, and
qeth_iqd_select_queue() needs to check for this and put it on some other
queue. As a result, the distribution for ucast flows is unbalanced and
hits QETH_IQD_MIN_UCAST_TXQ heavier than the other queues.

Open-coding a custom variant of skb_tx_hash() isn't an option, since
netdev_pick_tx() also gives us eg. access to XPS. But we can pull a
little trick: add a single TC class that excludes the mcast txq, and
thus encourage skb_tx_hash() to not pick the mcast txq.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: allow configuration of TX queues for IQD devices
Julian Wiedmann [Wed, 18 Mar 2020 12:54:49 +0000 (13:54 +0100)]
s390/qeth: allow configuration of TX queues for IQD devices

Similar to the support for z/VM NICs, but we need to take extra care
about the dedicated mcast queue:

1. netdev_pick_tx() is unaware of this limitation and might select the
   mcast txq. Catch this.
2. require at least _two_ TX queues - one for ucast, one for mcast.
3. when reducing the number of TX queues, there's a potential race
   where netdev_cap_txqueue() over-rules the selected txq index and
   falls back to index 0. This would place ucast traffic on the mcast
   queue, and result in TX errors.
   So for IQD, reject a reduction while the interface is running.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: allow configuration of TX queues for z/VM NICs
Julian Wiedmann [Wed, 18 Mar 2020 12:54:48 +0000 (13:54 +0100)]
s390/qeth: allow configuration of TX queues for z/VM NICs

Add support for ETHTOOL_SCHANNELS to change the count of active
TX queues.

Since all TX queue structs are pre-allocated and -registered, we just
need to trivially adjust dev->real_num_tx_queues.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
4 years agos390/qeth: remove prio-queueing support for z/VM NICs
Julian Wiedmann [Wed, 18 Mar 2020 12:54:47 +0000 (13:54 +0100)]
s390/qeth: remove prio-queueing support for z/VM NICs

z/VM NICs don't offer HW QoS for TX rings. So just use netdev_pick_tx()
to distribute the connections equally over all enabled TX queues.

We start with just 1 enabled TX queue (this matches the typical
configuration without prio-queueing). A follow-on patch will allow users
to enable additional TX queues.

Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>