Ilias Apalodimas [Thu, 20 Feb 2020 07:41:55 +0000 (09:41 +0200)]
net: page_pool: API cleanup and comments
Functions starting with __ usually indicate those which are exported,
but should not be called directly. Update some of those declared in the
API and make it more readable.
page_pool_unmap_page() and page_pool_release_page() were doing
exactly the same thing calling __page_pool_clean_page(). Let's
rename __page_pool_clean_page() to page_pool_release_page() and
export it in order to show up on perf logs and get rid of
page_pool_unmap_page().
Finally rename __page_pool_put_page() to page_pool_put_page() since we
can now directly call it from drivers and rename the existing
page_pool_put_page() to page_pool_put_full_page() since they do the same
thing but the latter is trying to sync the full DMA area.
This patch also updates netsec, mvneta and stmmac drivers which use
those functions.
Suggested-by: Jonathan Lemon <jonathan.lemon@gmail.com> Acked-by: Toke Høiland-Jørgensen <toke@redhat.com> Acked-by: Jesper Dangaard Brouer <brouer@redhat.com> Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 20 Feb 2020 18:04:34 +0000 (10:04 -0800)]
Merge branch 'mlxsw-Preparation-for-RTNL-removal'
Ido Schimmel says:
====================
mlxsw: Preparation for RTNL removal
The driver currently acquires RTNL in its route insertion path, which
contributes to very large control plane latencies. This patch set
prepares mlxsw for RTNL removal from its route insertion path in a
follow-up patch set.
Patches #1-#2 protect shared resources - KVDL and counter pool - with
their own locks. All allocations of these resources are currently
performed under RTNL, so no locks were required.
Patches #3-#7 ensure that updates to mirroring sessions only take place
in case there are active mirroring sessions. This allows us to avoid
taking RTNL when it is unnecessary, as updating of the mirroring
sessions must be performed under RTNL for the time being.
Patches #8-#10 replace the use of APIs that assume that RTNL is taken
with their RCU counterparts. Specifically, patches #8 and #9 replace
__in_dev_get_rtnl() with __in_dev_get_rcu() under RCU read-side critical
section. Patch #10 replaces __dev_get_by_index() with
dev_get_by_index_rcu().
Patches #11-#15 perform small adjustments in the code to make it easier
to later introduce a router lock instead of relying on RTNL.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:08:00 +0000 (09:08 +0200)]
mlxsw: spectrum_nve: Make tunnel initialization symmetric
The device supports a single VTEP whose configuration is shared between
all VXLAN tunnels.
While the shared configuration is cleared upon the destruction of the
last tunnel - in mlxsw_sp_nve_tunnel_fini() - it is set in
mlxsw_sp_nve_fid_enable(), after calling mlxsw_sp_nve_tunnel_init().
Make tunnel initialization and destruction symmetric and set the
configuration in mlxsw_sp_nve_tunnel_init().
This will later allow us to protect the shared configuration with a
lock.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:59 +0000 (09:07 +0200)]
mlxsw: spectrum: Export function to check if RIF exists
After the previous patch, all the callers of mlxsw_sp_rif_find_by_dev()
outside of the routing code use it to understand if a RIF exists for the
passed netdev.
Therefore, export a function to check if a RIF exists and make
mlxsw_sp_rif_find_by_dev() internal to the routing code.
This will later allow us to more easily introduce the router lock which
will also protect the RIFs.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:58 +0000 (09:07 +0200)]
mlxsw: spectrum: Prevent RIF access outside of routing code
There are currently 5 users of mlxsw_sp_rif_find_by_dev() outside of the
routing code. Only one call site actually needs to dereference the
router interface (RIF). The rest merely need to know if a RIF exists for
the provided netdev.
Convert this call site to query the needed information directly from the
routing code instead of dereferencing the RIF.
This will later allow us to replace mlxsw_sp_rif_find_by_dev() with a
function that checks if a RIF exist.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:57 +0000 (09:07 +0200)]
mlxsw: spectrum_router: Prepare function for router lock introduction
The function de-associates the port-vlan from its router interface
(RIF). It is called both from the netdev notifier block and the inetaddr
notifier block that will soon hold the router lock.
Make sure that router code calls the internal version, as it will
already have the router lock held when the function is called.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:56 +0000 (09:07 +0200)]
mlxsw: spectrum_router: Prepare function for router lock introduction
The function removes the FDB entry that directs the macvlan's MAC to the
router port. It is called from both the netdev notifier block and the
inetaddr notifier block that will soon hold the router lock.
Make sure that only the netdev notifier calls the exported version, so
that is will take the router lock, which will already be held by the
inetaddr notifier.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:55 +0000 (09:07 +0200)]
mlxsw: spectrum_router: Do not assume RTNL is taken when resolving underlay device
The function that resolves the underlay device of the IPIP tunnel
assumes that RTNL is taken, but this will not be correct when RTNL is
removed from the route insertion path.
Convert the function to use dev_get_by_index_rcu() instead of
__dev_get_by_index() and make sure it is always called from an RCU
read-side critical section.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:54 +0000 (09:07 +0200)]
mlxsw: spectrum_router: Do not assume RTNL is taken during RIF teardown
IPv6 addresses are deleted in an atomic context, so the driver defers
the potential teardown of the associated router interface (RIF) to a
work item that takes RTNL.
The RIF is only destroyed if the associated netdev does not have any IP
addresses (both IPv4 and IPv6). The IPv4 device ('struct in_device') is
currently fetched via __in_dev_get_rtnl() which assumes RTNL is taken.
Since RTNL is going to be removed, convert it to use __in_dev_get_rcu()
from an RCU read-side critical section.
Note that the IPv6 device ('struct inet6_dev') is fetched via
__in6_dev_get(), which does not require RTNL.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:53 +0000 (09:07 +0200)]
mlxsw: spectrum_router: Do not assume RTNL is taken during nexthop init
RTNL is going to be removed from route insertion path, so use
__in_dev_get_rcu() from an RCU read-side critical section instead of
__in_dev_get_rtnl() which assumes RTNL is taken.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:52 +0000 (09:07 +0200)]
mlxsw: spectrum_span: Only update mirroring agents if present
In order not to needlessly schedule the work item that updates the
mirroring agents, only schedule it if there are any mirroring agents
present.
This is done by adding an atomic counter that counts the active
mirroring agents.
It is incremented / decremented whenever a mirroring agent is created /
destroyed. It is read before scheduling the work item and in the
devlink-resource occupancy callback.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:50 +0000 (09:07 +0200)]
mlxsw: spectrum_span: Prepare work item to update mirroring agents
The driver updates its mirroring agents whenever it receives a
notification about an event that can affect these. For example, the
addition of a route might require the driver to change the egress port
of an ERSPAN session.
Currently, RTNL needs to be held when these agents are updates, so the
driver either:
1. Calls directly into the mirroring code, in case RTNL is held
2. Schedules a work item that will take RTNL and call into the mirroring
code
Simplify this by having the mirroring code schedule the work item for
the update instead of requiring callers to schedule a work item
themselves.
The conversion of the callers will be done in the next patch to make
review easier.
This will later allow us to remove RTNL from different parts of the
driver. It will also allow us to only schedule the work item in case
there are active mirroring agents, which is information private to the
mirroring code.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:47 +0000 (09:07 +0200)]
mlxsw: spectrum: Protect counter pool with a lock
The counter pool is a shared resource. It is used by both the ACL code
to allocate counters for actions and by the routing code to allocate
counters for adjacency entries (for example).
Currently, all allocations are protected by RTNL, but this is going to
change with the removal of RTNL from the routing code.
Therefore, protect counter allocations with a spin lock.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Thu, 20 Feb 2020 07:07:46 +0000 (09:07 +0200)]
mlxsw: spectrum_kvdl: Protect allocations with a lock
The KVDL is used to store objects allocated throughout various places
in the driver. For example, both nexthops (adjacency entries) and ACL
actions are stored in the KVDL.
Currently, all allocations are protected by RTNL, but this is going to
change with the removal of RTNL from the routing code.
Therefore, protect KVDL allocations with a lock. A mutex is used since
the free operation can block in Spectrum-2.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
net: phy: Better support for BCM54810
This patch series updates the broadcom PHY driver to better support the
BCM54810 and allow it to make use of the exiting
bcm54xx_adjust_rxrefclk() as well as fix suspend/resume for it.
Changes in v2:
- added Reviewed-by tags from Andrew for patches #1 and #3
- expanded commit message in #2 to explain the change
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Wed, 19 Feb 2020 20:00:49 +0000 (12:00 -0800)]
net: phy: broadcom: Wire suspend/resume for BCM54810
The BCM54810 PHY can use the standard BMCR Power down suspend, but needs
a custom resume routine which first clear the Power down bit, and then
re-initializes the PHY. While in low-power mode, the PHY only accepts
writes to the BMCR register. The datasheet clearly says it:
Reads or writes to any MII register other than MII Control register
(address 00h) while the device is in the standby power-down mode may
cause unpredictable results.
Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Wed, 19 Feb 2020 20:00:48 +0000 (12:00 -0800)]
net: phy: broadcom: Have bcm54xx_adjust_rxrefclk() check for flags
bcm54xx_adjust_rxrefclk() already checks for PHY_BRCM_AUTO_PWRDWN_ENABLE
and PHY_BRCM_DIS_TXCRXC_NOENRGY in order to set the appropriate bit. The
situation is a bit more complicated with the flag
PHY_BRCM_RX_REFCLK_UNUSED but essentially amounts to the same situation.
The default setting for the 125MHz clock is to be on for all PHYs and
we still treat BCM50610 and BCM50610M specifically with the polarity of
the bit reversed.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/sfc/efx.c:116:38: warning:
efx_default_channel_type defined but not used [-Wunused-const-variable=]
commit 83975485077d ("sfc: move channel alloc/removal code")
left behind this, remove it.
Reported-by: Hulk Robot <hulkci@huawei.com> Fixes: 83975485077d ("sfc: move channel alloc/removal code") Signed-off-by: YueHaibing <yuehaibing@huawei.com> Acked-by: Martin Habets <mhabets@solarflare.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 20 Feb 2020 00:25:12 +0000 (16:25 -0800)]
Merge branch 'hns3-next'
Huazhong Tan says:
====================
net: hns3: misc updates for -net-next
This series includes some misc updates for the HNS3
ethernet driver.
[patch 1] modifies an unsuitable print when setting dulex mode.
[patch 2] adds some debugfs info for TC and DWRR.
[patch 3] adds some debugfs info for loopback.
[patch 4] adds a missing help info for QS shaper in debugfs.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Yonglong Liu [Wed, 19 Feb 2020 01:23:33 +0000 (09:23 +0800)]
net: hns3: add missing help info for QS shaper in debugfs
HNS3 driver can dump QS shaper configs via debugfs, but missing
help info in debugfs for this operation.
Signed-off-by: Yonglong Liu <liuyonglong@huawei.com> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Yufeng Mo [Wed, 19 Feb 2020 01:23:32 +0000 (09:23 +0800)]
net: hns3: add support for dump MAC ID and loopback status in debugfs
The MAC ID and loopback status information are obtained from
the hardware, which will be helpful for debugging. This patch
adds support for these two items in debugfs.
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>
Yonglong Liu [Wed, 19 Feb 2020 01:23:31 +0000 (09:23 +0800)]
net: hns3: add enabled TC numbers and DWRR weight info in debugfs
The actual enabled TC numbers and the DWRR weight of each
TC may be helpful for debugging, so adds them into debugfs.
Signed-off-by: Yonglong Liu <liuyonglong@huawei.com> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Guangbin Huang [Wed, 19 Feb 2020 01:23:30 +0000 (09:23 +0800)]
net: hns3: modify an unsuitable print when setting unknown duplex to fibre
Currently, if device is in link down status and user uses
'ethtool -s' command to set speed but not specify duplex
mode, the duplex mode passed from ethtool to driver is
unknown value(255), and the fibre port will identify this
value as half duplex mode and print "only copper port
supports half duplex!". This message is confusing.
So for fibre port, only the setting duplex is half, prints
error and returns.
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com> Signed-off-by: Huazhong Tan <tanhuazhong@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
mlxsw: Replace zero-length array with flexible-array member
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com> Reviewed-by: Ido Schimmel <idosch@mellanox.com> Tested-by: Ido Schimmel <idosch@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Petr Oros [Tue, 18 Feb 2020 09:35:55 +0000 (10:35 +0100)]
phy: avoid unnecessary link-up delay in polling mode
commit 93c0970493c71f ("net: phy: consider latched link-down status in
polling mode") removed double-read of latched link-state register for
polling mode from genphy_update_link(). This added extra ~1s delay into
sequence link down->up.
Following scenario:
- After boot link goes up
- phy_start() is called triggering an aneg restart, hence link goes
down and link-down info is latched.
- After aneg has finished link goes up. In phy_state_machine is checked
link state but it is latched "link is down". The state machine is
scheduled after one second and there is detected "link is up". This
extra delay can be avoided when we keep link-state register double read
in case when link was down previously.
With this solution we don't miss a link-down event in polling mode and
link-up is faster.
Details about this quirky behavior on Realtek phy:
Without patch:
T0: aneg is started, link goes down, link-down status is latched
T0+3s: state machine runs, up-to-date link-down is read
T0+4s: state machine runs, aneg is finished (BMSR_ANEGCOMPLETE==1),
here i read link-down (BMSR_LSTATUS==0),
T0+5s: state machine runs, aneg is finished (BMSR_ANEGCOMPLETE==1),
up-to-date link-up is read (BMSR_LSTATUS==1),
phydev->link goes up, state change PHY_NOLINK to PHY_RUNNING
With patch:
T0: aneg is started, link goes down, link-down status is latched
T0+3s: state machine runs, up-to-date link-down is read
T0+4s: state machine runs, aneg is finished (BMSR_ANEGCOMPLETE==1),
first BMSR read: BMSR_ANEGCOMPLETE==1 and BMSR_LSTATUS==0,
second BMSR read: BMSR_ANEGCOMPLETE==1 and BMSR_LSTATUS==1,
phydev->link goes up, state change PHY_NOLINK to PHY_RUNNING
Signed-off-by: Petr Oros <poros@redhat.com> Reviewed-by: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Several network drivers for chips that support TSO6 share the same code
for preparing the TCP header, so let's factor it out to a helper.
A difference is that some drivers reset the payload_len whilst others
don't do this. This value is overwritten by TSO anyway, therefore
the new helper resets it in general.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Heiner Kallweit [Tue, 18 Feb 2020 20:12:52 +0000 (21:12 +0100)]
r8152: use new helper tcp_v6_gso_csum_prep
Use new helper tcp_v6_gso_csum_prep in additional network drivers.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Acked-by: Hayes Wang <hayeswang@realtek.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Heiner Kallweit [Tue, 18 Feb 2020 20:09:17 +0000 (21:09 +0100)]
net: socionext: use new helper tcp_v6_gso_csum_prep
Use new helper tcp_v6_gso_csum_prep in additional network drivers.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Heiner Kallweit [Tue, 18 Feb 2020 20:07:16 +0000 (21:07 +0100)]
ionic: use new helper tcp_v6_gso_csum_prep
Use new helper tcp_v6_gso_csum_prep in additional network drivers.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Acked-by: Shannon Nelson <snelson@pensando.io> Signed-off-by: David S. Miller <davem@davemloft.net>
Heiner Kallweit [Tue, 18 Feb 2020 20:05:02 +0000 (21:05 +0100)]
e1000(e): use new helper tcp_v6_gso_csum_prep
Use new helper tcp_v6_gso_csum_prep in additional network drivers.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com> Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Heiner Kallweit [Tue, 18 Feb 2020 19:56:41 +0000 (20:56 +0100)]
net: core: add helper tcp_v6_gso_csum_prep
Several network drivers for chips that support TSO6 share the same code
for preparing the TCP header, so let's factor it out to a helper.
A difference is that some drivers reset the payload_len whilst others
don't do this. This value is overwritten by TSO anyway, therefore
the new helper resets it in general.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com> Signed-off-by: David S. Miller <davem@davemloft.net>
works just fine. But currently non-initial network namespaces have no
way of kowing which congestion algorithms are available or allowed other
than through trial and error by writing the names of the algorithms into
the aforementioned file.
Since we already allow changing the congestion algorithm in non-initial
network namespaces by exposing the tcp_congestion_control file there is
no reason to not also expose the
tcp_{allowed,available}_congestion_control files to non-initial network
namespaces. After this change a container with a separate network
namespace will show:
root@f1:~# ls -al /proc/sys/net/ipv4/tcp_* | grep congestion
-rw-r--r-- 1 root root 0 Feb 19 11:54 /proc/sys/net/ipv4/tcp_allowed_congestion_control
-r--r--r-- 1 root root 0 Feb 19 11:54 /proc/sys/net/ipv4/tcp_available_congestion_control
-rw-r--r-- 1 root root 0 Feb 19 11:54 /proc/sys/net/ipv4/tcp_congestion_control
Link: https://github.com/lxc/lxc/issues/3267 Reported-by: Haw Loeung <haw.loeung@canonical.com> Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Lorenzo Bianconi [Wed, 19 Feb 2020 09:57:37 +0000 (10:57 +0100)]
net: mvneta: align xdp stats naming scheme to mlx5 driver
Introduce "rx" prefix in the name scheme for xdp counters
on rx path.
Differentiate between XDP_TX and ndo_xdp_xmit counters
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> Acked-by: Jesper Dangaard Brouer <brouer@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Most of the CGX register config is restricted to mapped RVU PFs,
this patch cleans up these permission checks spread across
the rvu_cgx.c file by moving the checks to a common fn().
Signed-off-by: Sunil Goutham <sgoutham@marvell.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Wed, 19 Feb 2020 18:44:25 +0000 (10:44 -0800)]
Merge tag 'mlx5-updates-2020-01-24' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
This series adds two moderate updates and some misc small patches to
mlx5 driver.
1) From Aya, Add the missing devlink health dump callbacks support for
both rx and tx health reporters.
First patch of the series is extending devlink API to set binary fmsg
data.
All others patches in the series are adding the mlx5 devlink health
callbacks support and the needed FW commands.
2) Also from Aya, Support for FEC modes based on 50G per lane links.
Part of this series, Aya adds one missing link mode define "FEC_LLRS"
to include/uapi/linux/ethtool.h.
3) From Joe, Use proper logging and tracing line terminations
4) From Christophe, Remove a useless 'drain_workqueue()'
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Aya Levin [Tue, 11 Feb 2020 22:32:53 +0000 (14:32 -0800)]
net/mlx5e: Add support for FEC modes based on 50G per lane links
Introduce new FEC modes:
- RS-FEC-(544,514)
- LL_RS-FEC-(272,257+1)
Add support in ethtool for set and get callbacks for the new modes
above. While RS-FEC-(544,514) is mapped to exsiting RS FEC mode,
LL_RS-FEC-(272,257+1) is mapped to a new ethtool link mode: LL-RS.
Add support for FEC on 50G per lane link modes up to 400G. The new link
modes uses a u16 fields instead of u8 fields for the legacy link modes.
Signed-off-by: Aya Levin <ayal@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Aya Levin [Tue, 11 Feb 2020 22:32:52 +0000 (14:32 -0800)]
ethtool: Add support for low latency RS FEC
Add support for low latency Reed Solomon FEC as LLRS.
The LL-FEC is defined by the 25G/50G ethernet consortium,
in the document titled "Low Latency Reed Solomon Forward Error Correction"
Signed-off-by: Aya Levin <ayal@mellanox.com> Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com> CC: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Aya Levin [Tue, 11 Feb 2020 22:32:51 +0000 (14:32 -0800)]
net/mlxe5: Separate between FEC and current speed
FEC mode is per link type, not necessary per speed. This patch access
FEC register by link modes instead of speeds. This patch will allow
further enhacment of link modes supporting FEC with the same speed
(different lane type).
Signed-off-by: Aya Levin <ayal@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Aya Levin [Tue, 11 Feb 2020 22:32:50 +0000 (14:32 -0800)]
net/mlx5e: Advertise globaly supported FEC modes
Ethtool advertise supported link modes on an interface. Per each FEC
mode, query if there is a link type which supports it. If so, add this
FEC mode to the supported FEC modes list. Prior to this patch, ethtool
advertised only the supported FEC modes on the current link type.
Add an explicit mapping between internal FEC modes and ethtool link mode
bits. With this change, adding new FEC modes in the downstream patch
would be easier.
Signed-off-by: Aya Levin <ayal@mellanox.com> Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Aya Levin [Tue, 11 Feb 2020 22:32:49 +0000 (14:32 -0800)]
net/mlx5e: Enforce setting of a single FEC mode
Ethtool command allow setting of several FEC modes in a single set
command. The driver can only set a single FEC mode at a time. With this
patch driver will reply not-supported on setting several FEC modes.
Signed-off-by: Aya Levin <ayal@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Aya Levin [Tue, 11 Feb 2020 22:32:48 +0000 (14:32 -0800)]
net/mlx5e: Set FEC to auto when configured mode is not supported
When configuring FEC mode, driver tries to set it for all available
link types. If a link type doesn't support a FEC mode, set this link
type to auto (FW best effort). Prior to this patch, when a link type
didn't support a FEC mode is was set to no FEC.
Signed-off-by: Aya Levin <ayal@mellanox.com> Reviewed-by: Eran Ben Elisha <eranbe@mellanox.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Joe Perches [Tue, 11 Feb 2020 22:32:47 +0000 (14:32 -0800)]
mlx5: Use proper logging and tracing line terminations
netdev_err should use newline termination but mlx5_health_report
is used in a trace output function devlink_health_report where
no newline should be used.
Remove the newlines from a couple formats and add a format string
of "%s\n" to the netdev_err call to not directly output the
logging string.
Also use snprintf to avoid any possible output string overrun.
Signed-off-by: Joe Perches <joe@perches.com> Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Aya Levin [Tue, 11 Feb 2020 22:32:46 +0000 (14:32 -0800)]
net/mlx5e: Support dump callback in RX reporter
Add support for SQ's FW dump on RX reporter's events. Use Resource dump
API to retrieve the relevant data: RX slice, RQ dump, RX buffer and
ICOSQ dump (depends on the error). Wrap it in formatted messages and
store the binary output in devlink core.
Aya Levin [Tue, 11 Feb 2020 22:32:45 +0000 (14:32 -0800)]
net/mlx5e: Support dump callback in TX reporter
Add support for SQ's FW dump on TX reporter's events. Use Resource dump
API to retrieve the relevant data: SX slice, SQ dump and SQ buffer. Wrap
it in formatted messages and store the binary output in devlink core.
Aya Levin [Tue, 11 Feb 2020 22:32:43 +0000 (14:32 -0800)]
net/mlx5: Add support for resource dump
On driver load:
- Initialize resource dump data structure and memory access tools (mkey
& pd).
- Read the resource dump's menu which contains the FW segment
identifier. Each record is identified by the segment name (ASCII).
During the driver's course of life, users (like reporters) may request
dumps per segment. The user should create a command providing the
segment identifier (SW enumeration) and command keys. In return, the
user receives a command context. In order to receive the dump, the user
should supply the command context and a memory (aligned to a PAGE) on
which the dump content will be written. Since the dump may be larger
than the given memory, the user may resubmit the command until received
an indication of end-of-dump. It is the user's responsibility to destroy
the command.
Aya Levin [Tue, 11 Feb 2020 22:32:42 +0000 (14:32 -0800)]
devlink: Force enclosing array on binary fmsg data
Add a new API for start/end binary array brackets [] to force array
around binary data as required from JSON. With this restriction, re-open
API to set binary fmsg data.
Fabio Estevam [Tue, 18 Feb 2020 15:34:44 +0000 (12:34 -0300)]
net: fec: Prevent unbind operation
After performing an unbind/bind operation the network is no longer
functional on i.MX6 (which has a single FEC instance):
# echo 2188000.ethernet > /sys/bus/platform/drivers/fec/unbind
# echo 2188000.ethernet > /sys/bus/platform/drivers/fec/bind
[ 10.756519] pps pps0: new PPS source ptp0
[ 10.792626] libphy: fec_enet_mii_bus: probed
[ 10.799330] fec 2188000.ethernet eth0: registered PHC device 1
# udhcpc -i eth0
udhcpc: started, v1.31.1
[ 14.985211] fec 2188000.ethernet eth0: no PHY, assuming direct connection to switch
[ 14.993140] libphy: PHY fixed-0:00 not found
[ 14.997643] fec 2188000.ethernet eth0: could not attach to PHY
On SoCs with two FEC instances there are some cases where one FEC instance
depends on the other one being present. One such example is i.MX28, which
has the following FEC dependency as noted in the comments:
/*
* The i.MX28 dual fec interfaces are not equal.
* Here are the differences:
*
* - fec0 supports MII & RMII modes while fec1 only supports RMII
* - fec0 acts as the 1588 time master while fec1 is slave
* - external phys can only be configured by fec0
*
* That is to say fec1 can not work independently. It only works
* when fec0 is working. The reason behind this design is that the
* second interface is added primarily for Switch mode.
*
* Because of the last point above, both phys are attached on fec0
* mdio interface in board design, and need to be configured by
* fec0 mii_bus.
*/
Prevent the unbind operation to avoid these issues.
Signed-off-by: Fabio Estevam <festevam@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
YueHaibing [Tue, 18 Feb 2020 06:21:54 +0000 (14:21 +0800)]
net: ena: remove set but not used variable 'hash_key'
drivers/net/ethernet/amazon/ena/ena_com.c: In function ena_com_hash_key_allocate:
drivers/net/ethernet/amazon/ena/ena_com.c:1070:50:
warning: variable hash_key set but not used [-Wunused-but-set-variable]
commit 6a4f7dc82d1e ("net: ena: rss: do not allocate key when not supported")
introduced this, but not used, so remove it.
Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: YueHaibing <yuehaibing@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
net: netlink: Replace zero-length array with flexible-array member
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
net: switchdev: Replace zero-length array with flexible-array member
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
bpf, sockmap: Replace zero-length array with flexible-array member
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
NFC: digital: Replace zero-length array with flexible-array member
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
net: usb: cdc-phonet: Replace zero-length array with flexible-array member
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
Russell King [Mon, 17 Feb 2020 16:03:11 +0000 (16:03 +0000)]
net: phy: allow bcm84881 to be a module
Now that the phylib module loading issue has been resolved, we can
allow this PHY driver to be built as a module.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk> Acked-by: Florian Fainelli <f.fainelli@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ursula Braun [Mon, 17 Feb 2020 15:24:55 +0000 (16:24 +0100)]
net/smc: reduce port_event scheduling
IB event handlers schedule the port event worker for further
processing of port state changes. This patch reduces the number of
schedules to avoid duplicate processing of the same port change.
Reviewed-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Ursula Braun <ubraun@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Karsten Graul [Mon, 17 Feb 2020 15:24:54 +0000 (16:24 +0100)]
net/smc: simplify normal link termination
smc_lgr_terminate() and smc_lgr_terminate_sched() both result in soft
link termination, smc_lgr_terminate_sched() is scheduling a worker for
this task. Take out complexity by always using the termination worker
and getting rid of smc_lgr_terminate() completely.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Ursula Braun <ubraun@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Karsten Graul [Mon, 17 Feb 2020 15:24:53 +0000 (16:24 +0100)]
net/smc: remove unused parameter of smc_lgr_terminate()
The soft parameter of smc_lgr_terminate() is not used and obsolete.
Remove it.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Ursula Braun <ubraun@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Karsten Graul [Mon, 17 Feb 2020 15:24:52 +0000 (16:24 +0100)]
net/smc: do not delete lgr from list twice
When 2 callers call smc_lgr_terminate() at the same time
for the same lgr, one gets the lgr_lock and deletes the lgr from the
list and releases the lock. Then the second caller gets the lock and
tries to delete it again.
In smc_lgr_terminate() add a check if the link group lgr is already
deleted from the link group list and prevent to try to delete it a
second time.
And add a check if the lgr is marked as freeing, which means that a
termination is already pending.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Ursula Braun <ubraun@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Karsten Graul [Mon, 17 Feb 2020 15:24:51 +0000 (16:24 +0100)]
net/smc: use termination worker under send_lock
smc_tx_rdma_write() is called under the send_lock and should not call
smc_lgr_terminate() directly. Call smc_lgr_terminate_sched() instead
which schedules a worker.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Ursula Braun <ubraun@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Karsten Graul [Mon, 17 Feb 2020 15:24:50 +0000 (16:24 +0100)]
net/smc: improve smc_lgr_cleanup()
smc_lgr_cleanup() is called during termination processing, there is no
need to send a DELETE_LINK at that time. A DELETE_LINK should have been
sent before the termination is initiated, if needed.
And remove the extra call to wake_up(&lnk->wr_reg_wait) because
smc_llc_link_inactive() already calls the related helper function
smc_wr_wakeup_reg_wait().
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Ursula Braun <ubraun@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
mlxsw: Reduce dependency between bridge and router code
This patch set reduces the dependency between the bridge and the router
code in preparation for RTNL removal from the route insertion path in
mlxsw.
The motivation and solution are explained in detail in patch #3. The
main idea is that we need to stop special-casing the VXLAN devices with
regards to the reference counting of the FIDs. Otherwise, we can bump
into the situation described in patch #3, where the routing code calls
into the bridge code which calls back into the routing code. After
adding a mutex to protect router data structures to remove RTNL
dependency, this can result in an AA deadlock.
Patches #1 and #2 are preparations. They convert the FIDs to use
'refcount_t' for reference counting in order to catch over/under flows
and add extack to the bridge creation function.
Patches #3-#5 reduce the dependency between the bridge and the router
code. First, by having the VXLAN device take a reference on the FID in
patch #3 and then by removing unnecessary code following the change in
patch #3.
Patches #6-#10 adjust existing selftests and add new ones to exercise
the new code paths.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:40 +0000 (16:29 +0200)]
selftests: mlxsw: vxlan: Add test for error path
Test that when two VXLAN tunnels with conflicting configurations (i.e.,
different TTL) are enslaved to the same VLAN-aware bridge, then the
enslavement of a port to the bridge is denied.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:39 +0000 (16:29 +0200)]
selftests: mlxsw: vxlan: Adjust test to recent changes
After recent changes, the VXLAN tunnel will be offloaded regardless if
any local ports are member in the FID or not. Adjust the test to make
sure the tunnel is offloaded in this case.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:37 +0000 (16:29 +0200)]
selftests: mlxsw: extack: Test bridge creation with VXLAN
Test that creation of a bridge (both VLAN-aware and VLAN-unaware) fails
with an extack when a VXLAN device with an unsupported configuration is
already enslaved to it.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:36 +0000 (16:29 +0200)]
selftests: mlxsw: Remove deprecated test
The addition of a VLAN on a bridge slave prompts the driver to have the
local port in question join the FID corresponding to this VLAN.
Before recent changes, the operation of joining the FID would also mean
that the driver would enable VXLAN tunneling if a VXLAN device was also
member in the VLAN. In case the configuration of the VXLAN tunnel was
not supported, an extack error would be returned.
Since the operation of joining the FID no longer means that VXLAN
tunneling is potentially enabled, the test is no longer relevant. Remove
it.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:35 +0000 (16:29 +0200)]
mlxsw: spectrum: Reduce dependency between bridge and router code
Commit f40be47a3e40 ("mlxsw: spectrum_router: Do not force specific
configuration order") added a call from the routing code to the bridge
code in order to handle the case where VNI should be set on a FID
following the joining of the router port to the FID.
This is no longer required, as previous patches made VXLAN devices
explicitly take a reference on the FID and set VNI on it.
Therefore, remove the unnecessary call and simply have the RIF take a
reference on the FID without checking if VNI should also be set on it.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:34 +0000 (16:29 +0200)]
mlxsw: spectrum_switchdev: Remove VXLAN checks during FID membership
As explained in previous patch, VXLAN devices now take a reference on
the FID and not only local ports. Therefore, there is no need for local
ports to check if they need to set a VNI on the FID when they join the
FID.
Remove these unnecessary checks.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:33 +0000 (16:29 +0200)]
mlxsw: spectrum_switchdev: Have VXLAN device take reference on FID
Up until now only local ports and the router port (which is also a local
port) took a reference on the corresponding FID (Filtering Identifier)
when joining a bridge. For example:
In this case the reference count of the FID will be '2'. Since the VXLAN
device does not take a reference on the FID, whenever a local port joins
the bridge it needs to check if a VXLAN device is already enslaved. If
the VXLAN device should be mapped to the FID in question, then the VXLAN
device's VNI is set on the FID.
Beside the fact that this scheme special-cases the VXLAN device, it also
creates an unnecessary dependency between the routing and bridge code:
1. [R] IP address is added on 'br0', which prompts the creation of a RIF
and a backing FID
2. [B] VNI is enabled on backing FID
3. [R] Host route corresponding to VXLAN device's source address is
promoted to perform NVE decapsulation
[R] - Routing code
[B] - Bridge code
This back and forth dependency will become problematic when a lock is
added in the routing code instead of relying on RTNL, as it will result
in an AA deadlock.
Instead, have the VXLAN device take a reference on the FID just like all
the other netdev members of the bridge. In order to correctly handle the
case where VXLAN devices are already enslaved to the bridge when it is
offloaded, walk the bridge's slaves and replay the configuration.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Reviewed-by: Petr Machata <petrm@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ido Schimmel [Mon, 17 Feb 2020 14:29:31 +0000 (16:29 +0200)]
mlxsw: spectrum_fid: Use 'refcount_t' for FID reference counting
'refcount_t' is very useful for catching over/under flows. Convert the
FID (Filtering Identifier) objects to use it instead of 'unsigned int'
for their reference count.
A subsequent patch in the series will change the way VXLAN devices hold
/ release the FID reference, which is why the conversion is made now.
Signed-off-by: Ido Schimmel <idosch@mellanox.com> Acked-by: Jiri Pirko <jiri@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Julian Wiedmann [Mon, 17 Feb 2020 13:45:01 +0000 (14:45 +0100)]
net: bridge: teach ndo_dflt_bridge_getlink() more brport flags
This enables ndo_dflt_bridge_getlink() to report a bridge port's
offload settings for multicast and broadcast flooding.
CC: Roopa Prabhu <roopa@cumulusnetworks.com> CC: Nikolay Aleksandrov <nikolay@cumulusnetworks.com> Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Edward Cree [Mon, 17 Feb 2020 13:43:28 +0000 (13:43 +0000)]
sfc: move some ARFS code out of headers
efx_filter_rfs_expire() is a work-function, so it being inline makes no
sense. It's only ever used in efx_channels.c, so move it there.
While we're at it, clean out some related unused cruft.
Signed-off-by: Edward Cree <ecree@solarflare.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Edward Cree [Mon, 17 Feb 2020 13:43:10 +0000 (13:43 +0000)]
sfc: only schedule asynchronous filter work if needed
Prevent excessive CPU time spent running a workitem with nothing to do.
We avoid any races by keeping the same check in efx_filter_rfs_expire().
Suggested-by: Martin Habets <mhabets@solarflare.com> Signed-off-by: Edward Cree <ecree@solarflare.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Julian Wiedmann [Mon, 17 Feb 2020 12:27:58 +0000 (13:27 +0100)]
net: vlan: suppress "failed to kill vid" warnings
When a real dev unregisters, vlan_device_event() also unregisters all
of its vlan interfaces. For each VID this ends up in __vlan_vid_del(),
which attempts to remove the VID from the real dev's VLAN filter.
But the unregistering real dev might no longer be able to issue the
required IOs, and return an error. Subsequently we raise a noisy warning
msg that is not appropriate for this situation: the real dev is being
torn down anyway, there shouldn't be any worry about cleanly releasing
all of its HW-internal resources.
So to avoid scaring innocent users, suppress this warning when the
failed deletion happens on an unregistering device.
While at it also convert the raw pr_warn() to a more fitting
netdev_warn().
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
Remove rtnl lock dependency from flow_action infra
Currently, TC flow_action infrastructure code obtain rtnl lock before
accessing action state in tc_setup_flow_action() function and releases
it afterwards. This behavior is not supposed to impact TC filter
insertion rate because filling flow_action representation is only a
small part of creating new filter and expensive operations (hardware
offload callbacks, classifiers, cls API code that creates chains and
classifiers instances) already support unlocked execution. However,
typical vswitch implementation might need to also dump TC filters
concurrently, for example to age out unused flows or update flow
counters. TC dump is fully serialized and holds rtnl lock during its
whole execution in kernel space. As such, it can significantly impact
concurrent tasks that try to intermittently obtain rtnl lock when
filling intermediate representation for new filter offload (performance
evaluation at the end of this mail).
Refactor flow_action cls API infrastructure and its dependencies to not
rely on rtnl lock for synchronization. Patch set overview:
- Refactor tc_setup_flow_action() to obtain action tcf_lock when
accessing action state. Fix its dependencies to not obtain tcf_lock
themselves and assume that caller already holds it (needs to be done
in same patch to prevent deadlock) and not to call sleeping functions
(needs to be done in same patch to prevent "sleeping while atomic"
dmesg warnings).
- Refactor action helper functions to require tcf_lock instead of rtnl.
Internally, all of the actions already use tcf_lock for
synchronization to accommodate unlocked classifier API, so this change
relies on already existing functionality.
- Remove rtnl lock and "rtnl_held" argument from tc_setup_flow_action()
function.
To test the change, multiple concurrent TC instances are invoked with
following command:
time ls add* | xargs -n 1 -P 100 sudo tc -b
Ten batch files with following typical rules (100k each) are used:
filter add dev ens1f0_0 protocol ip ingress prio 1 handle 1 flower
src_mac e4:11:0:0:0:0 dst_mac e4:12:0:0:0:0 src_ip 192.168.111.1
dst_ip 192.168.111.2 ip_proto udp dst_port 1 src_port 1 action
tunnel_key set id 1 src_ip 2.2.2.2 dst_ip 2.2.2.3 dst_port 4789
no_percpu action mirred egress redirect dev vxlan1 no_percpu
TC dump of same device is called in infinite loop from five concurrent
instances:
while true do tc -s filter show dev $NIC ingress >/dev/null done
Results obtained on current net-next commit 9f68e3655aae ("Merge tag
'drm-next-2020-01-30' of git://anongit.freedesktop.org/drm/drm"):
Test results confirm significant impact of concurrent TC dump. The
impact is almost fully mitigated by proposed change (differences can be
attributed to contention for chain and tp locks between add and dump TC
instances).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
projects / mirror_ubuntu-kernels.git / log