In commit fc922bb0dd ("mlxsw: spectrum_router: Use one LPM tree for
all virtual routers") I increased the scale of supported VRFs by having
all of them share the same LPM tree.
In order to avoid look-ups for prefix lengths that don't exist, each
route removal would trigger an aggregation across all the active virtual
routers to see which prefix lengths are in use and which aren't and
structure the tree accordingly.
With the way the data structures are currently laid out, this is a very
expensive operation. When preformed repeatedly - due to the invocation
of the abort mechanism - and with enough VRFs, this can result in a hung
task.
For now, avoid this optimization until it can be properly re-added in
net-next.
Fixes: fc922bb0dd ("mlxsw: spectrum_router: Use one LPM tree for all virtual routers")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reported-by: David Ahern <dsa@cumulusnetworks.com>
Tested-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add constants and callback functions for the dwmac on rk3128 soc.
As can be seen, the base structure is the same, only registers
and the bits in them moved slightly.
Signed-off-by: David Wu <david.wu@rock-chips.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When considering whether to set RTNH_F_OFFLOAD flag on an IPv6 route,
mlxsw_sp_fib6_entry_offload_set() looks up the mlxsw_sp_nexthop
corresponding to a given route, and decides based on whether the next
hop's offloaded flag was set. When looking for the matching next hop, it
also takes into account the device of the route, which must match next
hop's RIF.
IPIP next hops however hitherto didn't set the RIF. As a result, IPv6
routes forwarding traffic to IP-in-IP netdevices are never marked as
offloaded, even when they actually are.
Thus track RIF of IPIP next hops the same way as that of ETHERNET next
hops.
Fixes: 8f28a30976 ("mlxsw: spectrum_router: Support IPv6 overlay encap")
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>
When creating a new RIF, bumping RIF count of the containing VR is the
last thing to be done. Symmetrically, when destroying a RIF, RIF count
is first dropped and only then the rest of the cleanup proceeds.
That's a problem for loopback RIFs. Those hold two VR references: one
for overlay and one for underlay. mlxsw_sp_rif_destroy() releases the
overlay one, and the deconfigure() callback the underlay one. But if
both overlay and underlay are the same, and if there are no other
artifacts holding the VR alive, this put actually destroys the VR. Later
on, when mlxsw_sp_rif_destroy() calls mlxsw_sp_vr_put() for the same VR,
the VR will already have been released and the kernel crashes with NULL
pointer dereference.
The underlying problem is that the RIF under destruction ends up
referencing the overlay VR much longer than it claims: all the way until
the call to mlxsw_sp_vr_put(). So line up the reference counting
properly to reflect this. Make corresponding changes in
mlxsw_sp_rif_create() as well for symmetry.
Fixes: 6ddb7426a7 ("mlxsw: spectrum_router: Introduce loopback RIFs")
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>
If this sanity check fails, we must free 'rss_indir'. Otherwise there is a
memory leak.
'goto err' as done in the other error handling paths to fix it.
Fixes: 46a3df9f97 ("net: hns3: Fix for setting rss_size incorrectly")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
On Armada 7K/8K we need to explicitly enable the bus clock. The bus clock
is optional because not all the SoCs need them but at least for Armada
7K/8K it is actually mandatory.
The binding documentation is updating accordingly.
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Re-enable the MAC receiver by setting CONFIG_RE before powering down,
as instructed in section 6.3.5.1 of [1]. Without this the MAC fails
to receive WoL packets and never wakes up.
[1] DWC Ethernet QoS Databook 4.10a October 2014
Signed-off-by: Ed Blake <ed.blake@sondrel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add hook to stmmac_pltfr_pm_ops for suspend / resume handling.
Signed-off-by: Ed Blake <ed.blake@sondrel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch stops the internal MAC Tx clock from being enabled as the
internal clock isn't used. The definition used for the bit controlling
this behaviour is renamed as well as it was wrongly named (bit 4 of
GMAC_CTRL_2_REG).
Fixes: 3919357fb0 ("net: mvpp2: initialize the GMAC when using a port")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The private port_list array has a list of pointers to mvpp2_port
instances. This list is allocated given the number of ports enabled in
the device tree, but the pointers are set using the port-id property. If
on a single port is enabled, the port_list array will be of size 1, but
when registering the port, if its id is not 0 the driver will crash.
Other crashes were encountered in various situations.
This fixes the issue by using an index not equal to the value of the
port-id property.
Fixes: 3f518509de ("ethernet: Add new driver for Marvell Armada 375 network unit")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: Yan Markman <ymarkman@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Parsing fragmentation detection failed due to wrong configured
parser TCAM entry's. Some traffic was marked as fragmented in RX
descriptor, even it wasn't IP fragmented. The hardware also failed to
calculate checksums which lead to use software checksum and caused
performance degradation.
Fixes: 3f518509de ("ethernet: Add new driver for Marvell Armada 375 network unit")
Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Signed-off-by: Stefan Chulski <stefanc@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of calculating the offloads counters, count them explicitly.
The calculations done for these counters would result in bugs in some
cases, for example:
When running TCP traffic over a VXLAN tunnel with TSO enabled the following
counters would increase:
tx_csum_partial: 1,333,284
tx_csum_partial_inner: 29,286
tx4_csum_partial_inner: 384
tx7_csum_partial_inner: 8
tx9_csum_partial_inner: 34
tx10_csum_partial_inner: 26,807
tx11_csum_partial_inner: 287
tx12_csum_partial_inner: 27
tx16_csum_partial_inner: 6
tx25_csum_partial_inner: 1,733
Seems like tx_csum_partial increased out of nowhere.
The issue is in the following calculation in mlx5e_update_sw_counters:
s->tx_csum_partial = s->tx_packets - tx_offload_none - s->tx_csum_partial_inner;
While tx_packets increases by the number of GSO segments for each SKB,
tx_csum_partial_inner will only increase by one, resulting in wrong
tx_csum_partial counter.
Fixes: bfe6d8d1d4 ("net/mlx5e: Reorganize ethtool statistics")
Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Toggling of C-tag VLAN filter should not affect the "any S-tag" steering rule.
Fixes: 8a271746a2 ("net/mlx5e: Receive s-tagged packets in promiscuous mode")
Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Use the correct formatting for netdev features.
Fixes: 0e405443e8 ("net/mlx5e: Improve set features ndo resiliency")
Signed-off-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Encap entries cached by the driver could be invalidated due to
tunnel destination neighbour state changes.
When attempting to offload a flow that uses a cached encap entry,
we must check the entry validity and defer the offloading
if the entry exists but not valid.
When EAGAIN is returned, the flow offloading to hardware takes place
by the neigh update code when the tunnel destination neighbour
becomes connected.
Fixes: 232c001398 ("net/mlx5e: Add support to neighbour update flow")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
When offloading header re-write, the HW may need to adjust checksums along
the packet. For IP traffic, and a case where we are asked to modify fields in
the IP header, current HW supports that only for TCP and UDP. Enforce it, in
this case fail the offloading attempt for non TCP/UDP packets.
Fixes: d7e75a325c ('net/mlx5e: Add offloading of E-Switch TC pedit (header re-write) actions')
Fixes: 2f4fe4cab0 ('net/mlx5e: Add offloading of NIC TC pedit (header re-write) actions')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
In case the neighbour for the tunnel destination isn't valid,
we send a neighbour update request but we free the encap
header buffer. This is wrong, because we still need it for
allocating a HW encap entry once the neighbour is available.
Fix that by skipping freeing it if we wait for neighbour.
Fixes: 232c001398 ('net/mlx5e: Add support to neighbour update flow')
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Added check for the maximal number of flow counters attached
to rule (FTE).
Fixes: bd5251dbf1 ('net/mlx5_core: Introduce flow steering destination of type counter')
Signed-off-by: Raed Salem <raeds@mellanox.com>
Reviewed-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Currently, FPGA capability is located in (mdev)->caps.hca_cur,
change the location to be (mdev)->caps.fpga,
since hca_cur is reserved for HCA device capabilities.
Fixes: e29341fb3a ("net/mlx5: FPGA, Add basic support for Innova")
Signed-off-by: Inbar Karmy <inbark@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
When cleaning rdma netdevice we need to save the mdev pointer
because priv is released when we release netdev.
This bug was found using the kernel address sanitizer (KASAN).
use-after-free in mlx5_rdma_netdev_free+0xe3/0x100 [mlx5_core]
Fixes: 48935bbb7a ("net/mlx5e: IPoIB, Add netdevice profile skeleton")
Signed-off-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Call skb_frag_dma_map multiple times if tx length is greater than
device max and avoid processing tx ring until entire packet has been
sent.
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Pavel Belous <pavel.belous@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Due to a bug in aquantia atlantic card firmware, it sometimes reports
invalid link speed bits. That caused driver to report link down events,
although link itself is totally fine.
This patch ignores such out of blue readings.
Signed-off-by: Pavel Belous <Pavel.Belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Driver did a poor job in managing its Tx queues: Sometimes it could stop
tx queues due to link down condition in aq_nic_xmit - but never waked up
them. That led to Tx path total suspend.
This patch fixes this and improves generic queue management:
- introduces queue restart counter
- uses generic netif_ interface to disable and enable tx path
- refactors link up/down condition and introduces dmesg log event when
link changes.
- introduces new constant for minimum descriptors count required for queue
wakeup
Signed-off-by: Pavel Belous <Pavel.Belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Although hardware is capable for almost 16K MTU, without max_mtu field
correctly set it only allows standard MTU to be used.
This patch enables max MTU, calculating it from hardware maximum frame size
of 16352 octets (including FCS).
Fixes: 5513e16421 ("net: ethernet: aquantia: Fixes for aq_ndev_change_mtu")
Signed-off-by: Pavel Belous <Pavel.Belous@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Inlining these functions creates lots of stack variables that each take
64 bytes when KASAN is enabled, leading to this warning about potential
stack overflow:
drivers/net/ethernet/rocker/rocker_ofdpa.c: In function 'ofdpa_cmd_flow_tbl_add':
drivers/net/ethernet/rocker/rocker_ofdpa.c:621:1: error: the frame size of 2752 bytes is larger than 1536 bytes [-Werror=frame-larger-than=]
gcc-8 can now consolidate the stack slots itself, but on older versions
we get the same behavior by using a temporary variable that holds a
copy of the inline function argument.
Cc: stable@vger.kernel.org
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81715
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
When mapping the RX DMA buffers, the driver was accidentally specifying
zero for the buffer length. Under normal circumstances, SWIOTLB does not
need to allocate a bounce buffer, so the address is just mapped without
checking the size field. This is why the error was not detected earlier.
Fixes: b9b17debc6 ("net: emac: emac gigabit ethernet controller driver")
Cc: stable@vger.kernel.org
Signed-off-by: Timur Tabi <timur@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
All the error handling paths 'goto error', except this one.
We should also go to error in this case, or some resources will be
leaking.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make sure (of/i2c/platform)_device_id tables are NULL terminated.
Found by coccinelle spatch "misc/of_table.cocci"
Signed-off-by: Thomas Meyer <thomas@m3y3r.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
The EMAC has the option of sending only a single pause frame when
flow control is enabled and the RX queue is full. Although sending
only one pause frame has little value, this would allow admins to
enable automatic flow control without having to worry about the EMAC
flooding nearby switches with pause frames if the kernel hangs.
The option is enabled by using the single-pause-mode private flag.
Signed-off-by: Timur Tabi <timur@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current mapping between pri and tc is one to one,
so user can't map multi priorities to the same tc.
This patch changes the mapping to many to one.
Fixes: 848440544b ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rss_size is 1, 2, 4, 8, 16, 32, 64, 128, but acutal tc queue
size can be any u16 less than 128. If tc queue size is 5, we
set the rss_size to 8, indirection table will be used to limit
the size of actual queue size.
It may cause dropping of receiving packet in hardware if
rss_size is not set correctly.
For now, each TC has the same rss size.
Fixes: 46a3df9f97 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a typo error for feild, which should be field.
Fixes: 848440544b ("net: hns3: Add support of TX Scheduler & Shaper to HNS3 driver")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When hdev doesn't support DCB, rx private buffer is not allocated,
otherwise there is not enough buffer for rx shared buffer, causing
buffer allocation process to fail.
This patch fixes by checking the dcb capability in
hclge_rx_buffer_calc.
Fixes: 46a3df9f97 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rx_priv_buf_alloc is used to tell hardware how much buffer is
used for rx direction, right now only the private buffer is
assigned.
For ae_dev that doesn't support DCB, private rx buffer is assigned
to zero, only shared rx buffer is used. So not setting the shared
rx buffer cause dropping of packet in SSU.
Fixes: 46a3df9f97 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When rx private buffer is disabled, there may be some case that
the rx private buffer is not set to zero, which may cause buffer
allocation process to fail.
This patch fixes this problem by setting priv->enable to 0 and
priv->buf_size to zero when rx private buffer is disabled.
Fixes: 46a3df9f97 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When ae_dev doesn't support DCB, DEFAULT_DV must be set to
a lower value, otherwise the buffer allocation process will
fail.
This patch fix it by setting it to 30K bytes.
Fixes: 46a3df9f97 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When ae_dev doesn't support DCB, rx_priv_wl_config,
common_thrd_config and tm_qs_bp_cfg can't be called, otherwise
cmd return fail, which causes the hclge module initialization
process to fail.
This patch fix it by adding a DCB capability flag to check if
the ae_dev support DCB.
Fixes: 46a3df9f97 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch add the ROCE supported flag in the driver_data
field of pci_device_id, delete roce_pci_tbl and change
HNAE_DEV_SUPPORT_ROCE_B to HNAE3_DEV_SUPPORT_ROCE_B.
This cleanup is done in order to support adding capability
in pci_device_id and to fix initialization failure when
cmd is not supported.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
fec_ptp_check_pps_event will return 1 if FEC_T_TF_MASK caused
an interrupt. Don't return IRQ_NONE in this case.
Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
Acked-by: Fugang Duan <fugang.duan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
FEC_ENET_TS_TIMER is not checked in the interrupt routine
so there is no need to enable it.
Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
Acked-by: Fugang Duan <fugang.duan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before queue 0 was always checked if any queue caused an interrupt.
It is better to just mark queue 0 if queue 0 has caused an interrupt.
Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
Acked-by: Fugang Duan <Fugang.duan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is pretty much a carbon copy of
commit 3079c65214 ("caif: Fix napi poll list corruption")
with "caif" replaced by "emac".
The commit d75b1ade56 ("net: less interrupt masking in NAPI")
breaks emac.
It is now required that if the entire budget is consumed when poll
returns, the napi poll_list must remain empty. However, like some
other drivers emac tries to do a last-ditch check and if there is
more work it will call napi_reschedule and then immediately process
some of this new work. Should the entire budget be consumed while
processing such new work then we will violate the new caller
contract.
This patch fixes this by not touching any work when we reschedule
in emac.
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When register/unregister ae_dev, ae_dev should match all client
in the client_list. Enet and roce can co-exists together so we
should continue checking for enet and roce presence together.
So break should not be there.
Above caused problems in loading and unloading of modules.
Fixes: 38eddd126772 ("net: hns3: Add support of the HNAE3 framework")
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When there is no vlan id in the packets, hardware will treat the vlan id
as 0 and look for the mac_vlan table. This patch set the default vlan id
of PF as 0. Without this config, it will fail when look for mac_vlan
table, and hardware will drop packets.
Fixes: 6427264ef330 ("net: hns3: Add HNS3 Acceleration Engine &
Compatibility Layer Support")
Signed-off-by: Mingguang Qu <qumingguang@huawei.com>
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch replaces the ethernet address copy instance with more
appropriate ether_addr_copy() function.
Fixes: 6427264ef330 ("net: hns3: Add HNS3 Acceleration Engine &
Compatibility Layer Support")
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes the initialization of MAC address, fetched from HNS3
firmware i.e. when it is not randomly generated, to the HNS3 hardware.
Fixes: ca60906d2795 ("net: hns3: Add support of HNS3 Ethernet Driver for
hip08 SoC")
Signed-off-by: Lipeng <lipeng321@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
