OSDN Git Service

android-x86/kernel.git
6 years agoi40e: don't hold spinlock while resetting VF
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>
6 years agoi40e: use admin queue for setting LEDs behavior
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>
6 years agoi40e: Add support for 'ethtool -m'
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>
6 years agoi40e: Fix reporting of supported link modes
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>
6 years agoi40e: Fix a potential NULL pointer dereference
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>
6 years agoi40e: remove logically dead code
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>
6 years agoi40e: limit lan queue count in large CPU count machine
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>
6 years agoMerge branch 'mlxsw-Fixlets'
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>
6 years agomlxsw: spectrum_router: Drop a redundant condition
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>
6 years agomlxsw: spectrum_router: Fix a typo
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>
6 years agoselftests: rtnetlink.sh: add vxlan and fou test cases
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>
6 years agoMerge branch 'flow_dissector-dissect-tunnel-info'
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>
6 years agoflow_dissector: dissect tunnel info
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>
6 years agonet/dst: Make skb parameter of skb{metadata_dst, tunnel_info}() const
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>
6 years agosamples/bpf: fix warnings in xdp_monitor_user
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>
6 years agomlxsw: spectrum: fix uninitialized value in err
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>
6 years agoMerge branch 'bcm63xx_enet-small-fixes-and-cleanups'
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>
6 years agobcm63xx_enet: remove unneeded include
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>
6 years agobcm63xx_enet: drop unneeded NULL phy_clk check
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>
6 years agobcm63xx_enet: use managed functions for clock/ioremap
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>
6 years agobcm63xx_enet: do not rely on probe order
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>
6 years agobcm63xx_enet: do not write to random DMA channel on BCM6345
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>
6 years agobcm63xx_enet: correct clock usage
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>
6 years agonet-ipv6: remove unused IP6_ECN_clear() function
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>
6 years agoipv4: Namespaceify tcp_fastopen_blackhole_timeout knob
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>
6 years agoipv4: Namespaceify tcp_fastopen_key knob
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>
6 years agoipv4: Remove the 'publish' logic in tcp_fastopen_init_key_once
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>
6 years agoipv4: Namespaceify tcp_fastopen knob
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>
6 years agoMerge branch 'dsa_ptr-port'
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>
6 years agonet: dsa: remove tag ops from the switch tree
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>
6 years agonet: dsa: change dsa_ptr for a dsa_port
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>
6 years agonet: dsa: prepare master receive hot path
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>
6 years agonet: dsa: add tagging ops to port
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>
6 years agonet: dsa: use temporary dsa_device_ops variable
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>
6 years agonet: dsa: use cpu_dp in master code
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>
6 years agonet: dsa: add master helper to look up slaves
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>
6 years agonet: hns3: fix null pointer dereference before null check
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>
6 years agohv_netvsc: report stop_queue and wake_queue
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>
6 years agobpf: Fix compiler warning on info.map_ids for 32bit platform
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>
6 years agonet_sched: remove redundant assignment to ret
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>
6 years agonet: ipmr: make function ipmr_notifier_init static
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>
6 years agoibmvnic: Set state UP
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>
6 years agoRevert "net: dsa: bcm_sf2: Defer port enabling to calling port_enable"
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>
6 years agoMerge branch '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
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>
6 years agomkiss: remove redundant check on len being zero
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>
6 years agonet-ipv6: add support for sockopt(SOL_IPV6, IPV6_FREEBIND)
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>
6 years agonet: ipv6: send NS for DAD when link operationally up
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>
6 years agoi40e: refactor FW version checking
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>
6 years agoi40e: Enable VF to negotiate number of allocated queues
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>
6 years agoi40evf: fix ring to vector mapping
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>
6 years agoi40e: shutdown all IRQs and disable MSI-X when suspended
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>
6 years agoi40e: prevent service task from running while we're suspended
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>
6 years agoi40e: don't clear suspended state until we finish resuming
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>
6 years agoi40e: use newer generic PM support instead of legacy PM callbacks
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>
6 years agoi40e: use separate state bit for miscellaneous IRQ setup
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>
6 years agoi40evf: lower message level
Mitch Williams [Fri, 14 Jul 2017 13:27:01 +0000 (09:27 -0400)]
i40evf: lower message level

We see this message regularly on VF reset or unload (which invokes a
reset). It's essentially meaningless unless it's happening constantly.
To prevent consternation, lower the log level to debug so it's not seen
under normal circumstance.

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>
6 years agoi40e: fix for flow director counters not wrapping as expected
Mariusz Stachura [Fri, 14 Jul 2017 13:27:00 +0000 (09:27 -0400)]
i40e: fix for flow director counters not wrapping as expected

An errata with GLQF_PCNT causes it to not wrap as expected. This
can cause an error in flow director statistics. This patch resets
affected counters just after reading.

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>
6 years agoi40e: relax warning message in case of version mismatch
Mariusz Stachura [Fri, 14 Jul 2017 13:10:19 +0000 (09:10 -0400)]
i40e: relax warning message in case of version mismatch

Fortville and Fort Park devices are often on different firmware release
schedules. This change relaxes the minor version warning message,
so it is only displayed for older FW warning version for old
firmware Fortville 3 or earlier.

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>
6 years agoi40e: simplify member variable accesses
Sudheer Mogilappagari [Fri, 14 Jul 2017 13:10:18 +0000 (09:10 -0400)]
i40e: simplify member variable accesses

This commit replaces usage of vsi->back in i40e_print_link_message()
(which is actually a PF pointer) with temp variable.

Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoi40e: Fix link down message when interface is brought up
Sudheer Mogilappagari [Fri, 14 Jul 2017 13:10:16 +0000 (09:10 -0400)]
i40e: Fix link down message when interface is brought up

i40e_print_link_message() is intended to compare new
link state with current link state and print log message
only if the new state is different from current state.

However in current driver the new state does not get updated
when link is going down because of the if condition. When an
interface is brought down, vsi->state is set to I40E_VSI_DOWN
in i40e_vsi_close() and later i40e_print_link_message() does
not get invoked in i40e_link_event due to if condition. Hence
link down message doesn't appear when link is going down. The
down state is seen  later during i40e_open() and old state
gets printed. The actual link state doesn't get updated in
i40e_close() or i40e_open() but when i40e_handle_link_event is
called inside i40e_clean_adminq_subtask.

This change allows i40e_print_link_message() to be called when
interface is going down and keeps the state information updated.

Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoi40e: Fix unqualified module message while bringing link up
Sudheer Mogilappagari [Fri, 14 Jul 2017 13:10:15 +0000 (09:10 -0400)]
i40e: Fix unqualified module message while bringing link up

In current driver, when ifconfig ethx up is done, the link state
doesn't transition to UP inside i40e_open(). It changes after AQ
command response is handled in i40e_handle_link_event().

When pf->hw.phy.link_info.link_info is DOWN inside i40e_open(),
The state is transient and invalid. So log message gets printed
based on incorrect info (i.e link_info and an_info).

This commit removes check for unqualified module inside
i40e_up_complete(). The existing check in i40e_handle_link_event()
logs the error message based on correct link state information.

Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
6 years agoi40e/i40evf: rename bytes_per_int to bytes_per_usec
Jacob Keller [Fri, 14 Jul 2017 13:10:14 +0000 (09:10 -0400)]
i40e/i40evf: rename bytes_per_int to bytes_per_usec

This value is not calculating bytes_per_int, which would actually just
be bytes/ITR_COUNTDOWN_START, but rather it's calculating bytes/usecs.

Rename the variable for clarity so that future developers understand
what the value is actually calculating.

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>
6 years agonet: ipv4: remove fib_info arg to fib_check_nh
David Ahern [Thu, 28 Sep 2017 03:41:59 +0000 (20:41 -0700)]
net: ipv4: remove fib_info arg to fib_check_nh

fib_check_nh does not use the fib_info arg; remove t.

Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: ipv4: remove fib_weight
David Ahern [Thu, 28 Sep 2017 02:08:00 +0000 (19:08 -0700)]
net: ipv4: remove fib_weight

fib_weight in fib_info is set but not used. Remove it and the
helpers for setting it.

Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'bpf-extend-info'
David S. Miller [Fri, 29 Sep 2017 05:17:06 +0000 (06:17 +0100)]
Merge branch 'bpf-extend-info'

Martin KaFai Lau says:

====================
bpf: Extend bpf_{prog,map}_info

This patch series adds more fields to bpf_prog_info and bpf_map_info.
Please see individual patch for details.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Test new fields in bpf_attr and bpf_{prog, map}_info
Martin KaFai Lau [Wed, 27 Sep 2017 21:37:56 +0000 (14:37 -0700)]
bpf: Test new fields in bpf_attr and bpf_{prog, map}_info

This patch tests newly added fields of the bpf_attr,
bpf_prog_info and bpf_map_info.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Swap the order of checking prog_info and map_info
Martin KaFai Lau [Wed, 27 Sep 2017 21:37:55 +0000 (14:37 -0700)]
bpf: Swap the order of checking prog_info and map_info

This patch swaps the checking order.  It now checks the map_info
first and then prog_info.  It is a prep work for adding
test to the newly added fields (the map_ids of prog_info field
in particular).

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: libbpf: Provide basic API support to specify BPF obj name
Martin KaFai Lau [Wed, 27 Sep 2017 21:37:54 +0000 (14:37 -0700)]
bpf: libbpf: Provide basic API support to specify BPF obj name

This patch extends the libbpf to provide API support to
allow specifying BPF object name.

In tools/lib/bpf/libbpf, the C symbol of the function
and the map is used.  Regarding section name, all maps are
under the same section named "maps".  Hence, section name
is not a good choice for map's name.  To be consistent with
map, bpf_prog also follows and uses its function symbol as
the prog's name.

This patch adds logic to collect function's symbols in libbpf.
There is existing codes to collect the map's symbols and no change
is needed.

The bpf_load_program_name() and bpf_map_create_name() are
added to take the name argument.  For the other bpf_map_create_xxx()
variants, a name argument is directly added to them.

In samples/bpf, bpf_load.c in particular, the symbol is also
used as the map's name and the map symbols has already been
collected in the existing code.  For bpf_prog, bpf_load.c does
not collect the function symbol name.  We can consider to collect
them later if there is a need to continue supporting the bpf_load.c.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Add map_name to bpf_map_info
Martin KaFai Lau [Wed, 27 Sep 2017 21:37:53 +0000 (14:37 -0700)]
bpf: Add map_name to bpf_map_info

This patch allows userspace to specify a name for a map
during BPF_MAP_CREATE.

The map's name can later be exported to user space
via BPF_OBJ_GET_INFO_BY_FD.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf: Add name, load_time, uid and map_ids to bpf_prog_info
Martin KaFai Lau [Wed, 27 Sep 2017 21:37:52 +0000 (14:37 -0700)]
bpf: Add name, load_time, uid and map_ids to bpf_prog_info

The patch adds name and load_time to struct bpf_prog_aux.  They
are also exported to bpf_prog_info.

The bpf_prog's name is passed by userspace during BPF_PROG_LOAD.
The kernel only stores the first (BPF_PROG_NAME_LEN - 1) bytes
and the name stored in the kernel is always \0 terminated.

The kernel will reject name that contains characters other than
isalnum() and '_'.  It will also reject name that is not null
terminated.

The existing 'user->uid' of the bpf_prog_aux is also exported to
the bpf_prog_info as created_by_uid.

The existing 'used_maps' of the bpf_prog_aux is exported to
the newly added members 'nr_map_ids' and 'map_ids' of
the bpf_prog_info.  On the input, nr_map_ids tells how
big the userspace's map_ids buffer is.  On the output,
nr_map_ids tells the exact user_map_cnt and it will only
copy up to the userspace's map_ids buffer is allowed.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Alexei Starovoitov <ast@fb.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agotcp: fix under-evaluated ssthresh in TCP Vegas
Hoang Tran [Wed, 27 Sep 2017 16:30:58 +0000 (18:30 +0200)]
tcp: fix under-evaluated ssthresh in TCP Vegas

With the commit 76174004a0f19785 (tcp: do not slow start when cwnd equals
ssthresh), the comparison to the reduced cwnd in tcp_vegas_ssthresh() would
under-evaluate the ssthresh.

Signed-off-by: Hoang Tran <hoang.tran@uclouvain.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: bridge: add per-port group_fwd_mask with less restrictions
Nikolay Aleksandrov [Wed, 27 Sep 2017 13:12:44 +0000 (16:12 +0300)]
net: bridge: add per-port group_fwd_mask with less restrictions

We need to be able to transparently forward most link-local frames via
tunnels (e.g. vxlan, qinq). Currently the bridge's group_fwd_mask has a
mask which restricts the forwarding of STP and LACP, but we need to be able
to forward these over tunnels and control that forwarding on a per-port
basis thus add a new per-port group_fwd_mask option which only disallows
mac pause frames to be forwarded (they're always dropped anyway).
The patch does not change the current default situation - all of the others
are still restricted unless configured for forwarding.
We have successfully tested this patch with LACP and STP forwarding over
VxLAN and qinq tunnels.

Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'hns3-dcb'
David S. Miller [Thu, 28 Sep 2017 17:35:12 +0000 (10:35 -0700)]
Merge branch 'hns3-dcb'

Yunsheng Lin says:

====================
Add support for DCB feature in hns3 driver

The patchset contains some enhancement related to DCB before
adding support for DCB feature.

This patchset depends on the following patchset:
https://patchwork.ozlabs.org/cover/815646/
https://patchwork.ozlabs.org/cover/816145/

High Level Architecture:

                   [ lldpad ]
                       |
                       |
                       |
                 [ hns3_dcbnl ]
                       |
                       |
                       |
                 [ hclge_dcb ]
                   /      \
                /            \
             /                  \
     [ hclge_main ]        [ hclge_tm ]

Current patch-set support following functionality:
   Use of lldptool to configure the tc schedule mode, tc
   bandwidth(if schedule mode is ETS), prio_tc_map and
   PFC parameter.

V3: Drop mqprio support

V2: Fix for not defining variables in local loop.

V1: Initial Submit.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add DCB support when interacting with network stack
Yunsheng Lin [Wed, 27 Sep 2017 01:45:32 +0000 (09:45 +0800)]
net: hns3: Add DCB support when interacting with network stack

When using lldptool to configure DCB parameter, hclge_dcb module
call the client_ops->setup_tc to tell network stack which queue
and priority is using for specific tc.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Setting for fc_mode and dcb enable flag in TM module
Yunsheng Lin [Wed, 27 Sep 2017 01:45:31 +0000 (09:45 +0800)]
net: hns3: Setting for fc_mode and dcb enable flag in TM module

After the DCB feature is supported, fc_mode and dcb enable flag
must be set according to the DCB parameter.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add dcb netlink interface for the support of DCB feature
Yunsheng Lin [Wed, 27 Sep 2017 01:45:30 +0000 (09:45 +0800)]
net: hns3: Add dcb netlink interface for the support of DCB feature

This patch add dcb netlink interface by calling the interface from
hclge_dcb module.

This patch also update Makefile in order to build hns3_dcbnl module.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add hclge_dcb module for the support of DCB feature
Yunsheng Lin [Wed, 27 Sep 2017 01:45:29 +0000 (09:45 +0800)]
net: hns3: Add hclge_dcb module for the support of DCB feature

The hclge_dcb module calls the interface from hclge_main/tm
and provide interface for the dcb netlink interface.

This patch also update Makefiles required to build the DCB
supported code in HNS3 Ethernet driver and update the existing
Kconfig file in the hisilicon folder.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add some interface for the support of DCB feature
Yunsheng Lin [Wed, 27 Sep 2017 01:45:28 +0000 (09:45 +0800)]
net: hns3: Add some interface for the support of DCB feature

This patch add some interface and export some interface from
hclge_tm and hclgc_main to support the upcoming DCB feature.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add tc-based TM support for sriov enabled port
Yunsheng Lin [Wed, 27 Sep 2017 01:45:27 +0000 (09:45 +0800)]
net: hns3: Add tc-based TM support for sriov enabled port

When sriov is enabled and TM is in tc-based mode, vf's TM
parameters is not set in TM initialization process.
This patch add the tc_based TM support for sriov enabled
using the information in vport struct.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add support for port shaper setting in TM module
Yunsheng Lin [Wed, 27 Sep 2017 01:45:26 +0000 (09:45 +0800)]
net: hns3: Add support for port shaper setting in TM module

This patch add a tm_port_shaper cmd and set port shaper
to HCLGE_ETHER_MAX_RATE on TM initialization process.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add support for PFC setting in TM module
Yunsheng Lin [Wed, 27 Sep 2017 01:45:25 +0000 (09:45 +0800)]
net: hns3: Add support for PFC setting in TM module

This patch add a pfc_pause_en cmd, and use it to configure
PFC option according to fc_mode in hdev->tm_info.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Add support for dynamically buffer reallocation
Yunsheng Lin [Wed, 27 Sep 2017 01:45:24 +0000 (09:45 +0800)]
net: hns3: Add support for dynamically buffer reallocation

Current buffer allocation can only happen at init, when
doing buffer reallocation after init, care must be taken
care of memory which priv_buf points to.
This patch fixes it by using a dynamic allocated temporary
memory. Because we only do buffer reallocation at init or
when setting up the DCB parameter, and priv_buf is only
used at buffer allocation process, so it is ok to use a
dynamic allocated temporary memory.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet: hns3: Support for dynamically assigning tx buffer to TC
Yunsheng Lin [Wed, 27 Sep 2017 01:45:23 +0000 (09:45 +0800)]
net: hns3: Support for dynamically assigning tx buffer to TC

This patch add support of dynamically assigning tx buffer to
TC when the TC is enabled.
It will save buffer for rx direction to avoid packet loss.

Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoarp: make arp_hdr_len() return unsigned int
Alexey Dobriyan [Tue, 26 Sep 2017 20:12:28 +0000 (23:12 +0300)]
arp: make arp_hdr_len() return unsigned int

Negative ARP header length are not a thing.

Constify arguments while I'm at it.

Space savings:

add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-3 (-3)
function                        old     new   delta
arpt_do_table                  1163    1160      -3

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet-next/hinic: Fix a case of Tx Queue is Stopped forever
Aviad Krawczyk [Tue, 26 Sep 2017 18:11:33 +0000 (02:11 +0800)]
net-next/hinic: Fix a case of Tx Queue is Stopped forever

Fix the following scenario:
1. tx_free_poll is running on cpu X
2. xmit function is running on cpu Y and fails to get sq wqe
3. tx_free_poll frees wqes on cpu X and checks the queue is not stopped
4. xmit function stops the queue after failed to get sq wqe
5. The queue is stopped forever

Signed-off-by: Aviad Krawczyk <aviad.krawczyk@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet-next/hinic: Set Rxq irq to specific cpu for NUMA
Aviad Krawczyk [Tue, 26 Sep 2017 17:57:50 +0000 (01:57 +0800)]
net-next/hinic: Set Rxq irq to specific cpu for NUMA

Set Rxq irq to specific cpu for allocating and receiving the skb from
the same node.

Signed-off-by: Aviad Krawczyk <aviad.krawczyk@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoMerge branch 'bpf-verifier-disassembly-improvements'
David S. Miller [Thu, 28 Sep 2017 17:23:19 +0000 (10:23 -0700)]
Merge branch 'bpf-verifier-disassembly-improvements'

Edward Cree says:

====================
bpf/verifier: disassembly improvements

Fix the output of print_bpf_insn() for ALU ops that don't look like
 compound assignment (i.e. BPF_END and BPF_NEG).

Sample output for a short test program:
0: (b4) (u32) r0 = (u32) 0
1: (dc) r0 = be32 r0
2: (84) r0 = (u32) -r0
3: (95) exit
processed 4 insns, stack depth 0
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf/verifier: improve disassembly of BPF_NEG instructions
Edward Cree [Tue, 26 Sep 2017 15:35:29 +0000 (16:35 +0100)]
bpf/verifier: improve disassembly of BPF_NEG instructions

BPF_NEG takes only one operand, unlike the bulk of BPF_ALU[64] which are
 compound-assignments.  So give it its own format in print_bpf_insn().

Signed-off-by: Edward Cree <ecree@solarflare.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agobpf/verifier: improve disassembly of BPF_END instructions
Edward Cree [Tue, 26 Sep 2017 15:35:13 +0000 (16:35 +0100)]
bpf/verifier: improve disassembly of BPF_END instructions

print_bpf_insn() was treating all BPF_ALU[64] the same, but BPF_END has a
 different structure: it has a size in insn->imm (even if it's BPF_X) and
 uses the BPF_SRC (X or K) to indicate which endianness to use.  So it
 needs different code to print it.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agocxgb4: make function ch_flower_stats_cb, fixes warning
Colin Ian King [Tue, 26 Sep 2017 15:14:09 +0000 (16:14 +0100)]
cxgb4: make function ch_flower_stats_cb, fixes warning

The function ch_flower_stats_cb is local to the source and does not need
to be in global scope, so make it static.

Cleans up sparse warnings:
symbol 'ch_flower_stats_cb' 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>
6 years agoMerge branch 'rtnl-pushdown-prep'
David S. Miller [Thu, 28 Sep 2017 17:20:49 +0000 (10:20 -0700)]
Merge branch 'rtnl-pushdown-prep'

Florian Westphal says:

====================
rtnetlink: preparation patches for further rtnl lock pushdown/removal

Patches split large rtnl_fill_ifinfo into smaller chunks
to better see which parts

1. require rtnl
2. do not require it at all
3. rely on rtnl locking now but could be converted

Changes since v3:

I dropped the 'ifalias' patch, I have a change to decouple ifalias and
rtnl mutex, I will send it once this series has been merged.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agortnetlink: rtnl_have_link_slave_info doesn't need rtnl
Florian Westphal [Tue, 26 Sep 2017 11:58:43 +0000 (13:58 +0200)]
rtnetlink: rtnl_have_link_slave_info doesn't need rtnl

it can be switched to rcu.

Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agortnetlink: add helpers to dump netnsid information
Florian Westphal [Tue, 26 Sep 2017 11:58:42 +0000 (13:58 +0200)]
rtnetlink: add helpers to dump netnsid information

Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agortnetlink: add helpers to dump vf information
Florian Westphal [Tue, 26 Sep 2017 11:58:41 +0000 (13:58 +0200)]
rtnetlink: add helpers to dump vf information

similar to earlier patches, split out more parts of this function to
better see what is happening and where we assume rtnl is locked.

Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agortnetlink: add helper to put master and link ifindexes
Florian Westphal [Tue, 26 Sep 2017 11:58:40 +0000 (13:58 +0200)]
rtnetlink: add helper to put master and link ifindexes

rtnl_fill_ifinfo currently requires caller to hold the rtnl mutex.
Unfortunately the function is quite large which makes it harder to see
which spots require the lock, which spots assume it and which ones could
do without.

Add helpers to factor out the ifindex dumping, one can use rcu to avoid
rtnl dependency.

Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoselftests: rtnetlink.sh: add rudimentary vrf test
Florian Westphal [Tue, 26 Sep 2017 05:40:42 +0000 (07:40 +0200)]
selftests: rtnetlink.sh: add rudimentary vrf test

Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet_sched: use idr to allocate u32 filter handles
Cong Wang [Mon, 25 Sep 2017 17:13:51 +0000 (10:13 -0700)]
net_sched: use idr to allocate u32 filter handles

Instead of calling u32_lookup_ht() in a loop to find
a unused handle, just switch to idr API to allocate
new handles. u32 filters are special as the handle
could contain a hash table id and a key id, so we
need two IDR to allocate each of them.

Cc: Chris Mi <chrism@mellanox.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet_sched: use idr to allocate basic filter handles
Cong Wang [Mon, 25 Sep 2017 17:13:50 +0000 (10:13 -0700)]
net_sched: use idr to allocate basic filter handles

Instead of calling basic_get() in a loop to find
a unused handle, just switch to idr API to allocate
new handles.

Cc: Chris Mi <chrism@mellanox.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agonet_sched: use idr to allocate bpf filter handles
Cong Wang [Mon, 25 Sep 2017 17:13:49 +0000 (10:13 -0700)]
net_sched: use idr to allocate bpf filter handles

Instead of calling cls_bpf_get() in a loop to find
a unused handle, just switch to idr API to allocate
new handles.

Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Chris Mi <chrism@mellanox.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
6 years agoinetpeer: speed up inetpeer_invalidate_tree()
Eric Dumazet [Mon, 25 Sep 2017 16:14:14 +0000 (09:14 -0700)]
inetpeer: speed up inetpeer_invalidate_tree()

As measured in my prior patch ("sch_netem: faster rb tree removal"),
rbtree_postorder_for_each_entry_safe() is nice looking but much slower
than using rb_next() directly, except when tree is small enough
to fit in CPU caches (then the cost is the same)

From: Eric Dumazet <edumazet@google.com>

Signed-off-by: David S. Miller <davem@davemloft.net>