Commit dd82364c3a ("mlxsw: Flip to the new dev walk API") did some
small changes in mlxsw code, but it did not respect the naming
conventions. So fix this now.
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the structure of the LPM tree changes (f.e., due to the addition of
a new prefix), we unbind the old tree and then bind the new one. This
may result in temporary packet loss.
Instead, overwrite the old binding with the new one.
Fixes: 6b75c4807d ("mlxsw: spectrum_router: Add virtual router management")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The cited commit makes a great job of finding optimal shift/multiplier
values assuming a 10 seconds wrap around, but forgot to change the
overflow_period computation.
It overflows in cyclecounter_cyc2ns(), and the final result is 804 ms,
which is silly.
Lets simply use 5 seconds, no need to recompute this, given how it is
supposed to work.
Later, we will use a timer instead of a work queue, since the new RX
allocation schem will no longer need mlx4_en_recover_from_oom() and the
service_task firing every 250 ms.
Fixes: 31c128b66e ("net/mlx4_en: Choose time-stamping shift value according to HW frequency")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Cc: Eugenia Emantayev <eugenia@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking fixes from David Miller:
1) Some 'const'ing in qlogic networking drivers, from Bhumika Goyal.
2) Fix scheduling while atomic in l2tp network namespace exit by
deferring the work to the workqueue. From Ridge Kennedy.
3) Fix use after free in dccp timewait handling, from Andrey Ryabinin.
4) mlx5e CQE compression engine not initialized properly, from Tariq
Toukan.
5) Some UAPI header fixes from Dmitry V. Levin.
6) Don't overwrite module parameter value in mlx4 driver, from Majd
Dibbiny.
7) Fix divide by zero in xt_hashlimit netfilter module, from Alban
Browaeys.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (35 commits)
bpf: Fix bpf_xdp_event_output
net/mlx4_en: Use __skb_fill_page_desc()
net/mlx4_core: Use cq quota in SRIOV when creating completion EQs
net/mlx4_core: Fix VF overwrite of module param which disables DMFS on new probed PFs
net/mlx4: Spoofcheck and zero MAC can't coexist
net/mlx4: Change ENOTSUPP to EOPNOTSUPP
uapi: fix linux/rds.h userspace compilation errors
uapi: fix linux/seg6.h and linux/seg6_iptunnel.h userspace compilation errors
lib: Remove string from parman config selection
forcedeth: Remove return from a void function
bpf: fix spelling mistake: "proccessed" -> "processed"
uapi: fix linux/llc.h userspace compilation error
uapi: fix linux/ip6_tunnel.h userspace compilation errors
net/mlx5e: Fix wrong CQE decompression
net/mlx5e: Update MPWQE stride size when modifying CQE compress state
net/mlx5e: Fix broken CQE compression initialization
net/mlx5e: Do not reduce LRO WQE size when not using build_skb
net/mlx5e: Register/unregister vport representors on interface attach/detach
net/mlx5e: s390 system compilation fix
tcp: account for ts offset only if tsecr not zero
...
Because the Mellanox code required being based on a net-next tree,
I keept it separate from the remainder of the RDMA stack submission
that is based on 4.10-rc3.
This branch contains:
- Various mlx4 and mlx5 fixes and minor changes
- Support for adding a tag match rule to flow specs
- Support for cvlan offload operation for raw ethernet QPs
- A change to the core IB code to recognize raw eth capabilities and
enumerate them (touches non-Mellanox code)
- Implicit On-Demand Paging memory registration support
-----BEGIN PGP SIGNATURE-----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=rspQ
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Pull Mellanox rdma updates from Doug Ledford:
"Mellanox specific updates for 4.11 merge window
Because the Mellanox code required being based on a net-next tree, I
keept it separate from the remainder of the RDMA stack submission that
is based on 4.10-rc3.
This branch contains:
- Various mlx4 and mlx5 fixes and minor changes
- Support for adding a tag match rule to flow specs
- Support for cvlan offload operation for raw ethernet QPs
- A change to the core IB code to recognize raw eth capabilities and
enumerate them (touches non-Mellanox code)
- Implicit On-Demand Paging memory registration support"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (40 commits)
IB/mlx5: Fix configuration of port capabilities
IB/mlx4: Take source GID by index from HW GID table
IB/mlx5: Fix blue flame buffer size calculation
IB/mlx4: Remove unused variable from function declaration
IB: Query ports via the core instead of direct into the driver
IB: Add protocol for USNIC
IB/mlx4: Support raw packet protocol
IB/mlx5: Support raw packet protocol
IB/core: Add raw packet protocol
IB/mlx5: Add implicit MR support
IB/mlx5: Expose MR cache for mlx5_ib
IB/mlx5: Add null_mkey access
IB/umem: Indicate that process is being terminated
IB/umem: Update on demand page (ODP) support
IB/core: Add implicit MR flag
IB/mlx5: Support creation of a WQ with scatter FCS offload
IB/mlx5: Enable QP creation with cvlan offload
IB/mlx5: Enable WQ creation and modification with cvlan offload
IB/mlx5: Expose vlan offloads capabilities
IB/uverbs: Enable QP creation with cvlan offload
...
Or we might miss the fact that a page was allocated from memory reserves.
Fixes: dceeab0e52 ("mlx4: support __GFP_MEMALLOC for rx")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When creating EQs to handle CQ completion events for the PF
or for VFs, we create enough EQE entries to handle completions
for the max number of CQs that can use that EQ.
When SRIOV is activated, the max number of CQs a VF (or the PF) can
obtain is its CQ quota (determined by the Hypervisor resource tracker).
Therefore, when creating an EQ, the number of EQE entries that the VF
should request for that EQ is the CQ quota value (and not the total
number of CQs available in the FW).
Under SRIOV, the PF, also must use its CQ quota, because
the resource tracker also controls how many CQs the PF can obtain.
Using the FW total CQs instead of the CQ quota when creating EQs resulted
wasting MTT entries, due to allocating more EQEs than were needed.
Fixes: 5a0d0a6161 ("mlx4: Structures and init/teardown for VF resource quotas")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Reported-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the VF driver, module parameter mlx4_log_num_mgm_entry_size was
mistakenly overwritten -- and in a manner which overrode the
device-managed flow steering option encoded in the parameter.
log_num_mgm_entry_size is a global module parameter which
affects all ConnectX-3 PFs installed on that host.
If a VF changes log_num_mgm_entry_size, this will affect all PFs
which are probed subsequent to the change (by disabling DMFS for
those PFs).
Fixes: 3c439b5586 ("mlx4_core: Allow choosing flow steering mode")
Signed-off-by: Majd Dibbiny <majd@mellanox.com>
Reviewed-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Spoofcheck can't be enabled if VF MAC is zero.
Vice versa, can't zero MAC if spoofcheck is on.
Fixes: 8f7ba3ca12 ('net/mlx4: Add set VF mac address support')
Signed-off-by: Eugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As ENOTSUPP is specific to NFS, change the return error value to
EOPNOTSUPP in various places in the mlx4 driver.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Suggested-by: Yotam Gigi <yotamg@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In cqe compression with striding RQ, the decompression of the CQE field
wqe_counter was done with a wrong wraparound value.
This caused handling cqes with a wrong pointer to wqe (rx descriptor)
and creating SKBs with wrong data, pointing to wrong (and already consumed)
strides/pages.
The meaning of the CQE field wqe_counter in striding RQ holds the
stride index instead of the WQE index. Hence, when decompressing
a CQE, wqe_counter should have wrapped-around the number of strides
in a single multi-packet WQE.
We dropped this wrap-around mask at all in CQE decompression of striding
RQ. It is not needed as in such cases the CQE compression session would
break because of different value of wqe_id field, starting a new
compression session.
Tested:
ethtool -K ethxx lro off/on
ethtool --set-priv-flags ethxx rx_cqe_compress on
super_netperf 16 {ipv4,ipv6} -t TCP_STREAM -m 50 -D
verified no csum errors and no page refcount issues.
Fixes: 7219ab34f1 ("net/mlx5e: CQE compression")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Reported-by: Tom Herbert <tom@herbertland.com>
Cc: kernel-team@fb.com
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the admin enables/disables cqe compression, updating
mpwqe stride size is required:
CQE compress ON ==> stride size = 256B
CQE compress OFF ==> stride size = 64B
This is already done on driver load via mlx5e_set_rq_type_params, all we
need is just to call it on arbitrary admin changes of cqe compression
state via priv flags or when changing timestamping state
(as it is mutually exclusive with cqe compression).
This bug introduces no functional damage, it only makes cqe compression
occur less often, since in ConnectX4-LX CQE compression is performed
only on packets smaller than stride size.
Tested:
ethtool --set-priv-flags ethxx rx_cqe_compress on
pktgen with 64 < pkt size < 256 and netperf TCP_STREAM (IPv4/IPv6)
verify `ethtool -S ethxx | grep compress` are advancing more often
(rapidly)
Fixes: 7219ab34f1 ("net/mlx5e: CQE compression")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Cc: kernel-team@fb.com
Signed-off-by: David S. Miller <davem@davemloft.net>
Some of RQ type parameters are derived from CQE compression state flag,
CQE compression flag was initialized only after RQ type parameters
setup. This leads to load RQ with stride size smaller than what we
want for when CQE compression is on.
This bug introduces no functional damage, it only makes CQE compression
occur less often, since in ConnectX4-LX CQE compression is performed
only on packets smaller than stride size.
Fix this by marking default status of CQE compression in PFLAG prior to
calling mlx5e_set_rq_priv_params(), as it inits some fields based on it.
Tested:
load driver on systems where rx CQE compress will be on (MH)
pktgen with 64 < pkt size < 256 and netperf TCP_STREAM (IPv4/IPv6)
verify `ethtool -S ethxx | grep compress` are advancing more often
(rapidly)
Fixes: 2fc4bfb725 ("net/mlx5e: Dynamic RQ type infrastructure")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Cc: kernel-team@fb.com
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When rq_type is Striding RQ, no room of SKB_RESERVE is needed
as SKB allocation is not done via build_skb.
Fixes: e4b8550807 ("net/mlx5e: Slightly reduce hardware LRO size")
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently vport representors are added only on driver load and removed on
driver unload. Apparently we forgot to handle them when we added the
seamless reset flow feature. This caused to leave the representors
netdevs alive and active with open HW resources on pci shutdown and on
error reset flows.
To overcome this we move their handling to interface attach/detach, so
they would be cleaned up on shutdown and recreated on reset flows.
Fixes: 26e59d8077 ("net/mlx5e: Implement mlx5e interface attach/detach callbacks")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Reviewed-by: Hadar Hen Zion <hadarh@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since mlx4 NIC are used on PowerPC with 64K pages, we need to adapt
MLX4_EN_ALLOC_PREFER_ORDER definition.
Otherwise, a fragment sitting in an out of order TCP queue can hold
0.5 Mbytes and it is a serious OOM risk.
Fixes: 51151a16a6 ("mlx4: allow order-0 memory allocations in RX path")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1) In the case where rate == priv->pkt_rate_low == priv->pkt_rate_high,
mlx4_en_auto_moderation() does a divide by zero.
2) We want to properly change the moderation parameters if rx_frames
was changed (like in ethtool -C eth0 rx-frames 16)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All rx and rx netdev interrupts are handled by respectively
by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
But mlx4_eq_int() also fires a tasklet to service all items that were
queued via mlx4_add_cq_to_tasklet(), but this handler was not called
unless user cqe was handled.
This is very confusing, as "mpstat -I SCPU ..." show huge number of
tasklet invocations.
This patch saves this overhead, by carefully firing the tasklet directly
from mlx4_add_cq_to_tasklet(), removing four atomic operations per IRQ.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Cc: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current behaviour of "mirred redirect" action (forward) offload is a bit
odd. For matched packets the action forwards them to the desired
destination, but it also lets the packet duplicates to go the original
way down (bridge, router, etc). That is more like "mirred mirror".
Fix this by using PBS type which behaves exactly like "mirred redirect".
Note that PBS does not support loopback mode.
Fixes: 4cda7d8d70 ("mlxsw: core: Introduce flexible actions support")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using a reader-writer lock in fast path is silly, when we can
instead use RCU or a seqlock.
For mlx4 hwstamp clock, a seqlock is the way to go, removing
two atomic operations and false sharing.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Point back the unregister IPv6 mc table to the bc table.
It is done since IPv6 mcast snooping is not supported for Spectrum yet.
Reported-by: Jiri Pirko <jiri@mellanox.com>
Fixes: 71c365bdc4 ("mlxsw: spectrum: Separate bc and mc floods")
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Tested-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When called by HW offloading drivers, the TC action (e.g
net/sched/act_mirred.c) code uses this_cpu logic, e.g
_bstats_cpu_update(this_cpu_ptr(a->cpu_bstats), bytes, packets)
per the kernel documention, preemption should be disabled, add that.
Before the fix, when running with CONFIG_PREEMPT set, we get a
BUG: using smp_processor_id() in preemptible [00000000] code: tc/3793
asserion from the TC action (mirred) stats_update callback.
Fixes: aad7e08d39 ('net/mlx5e: Hardware offloaded flower filter statistics support')
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Flow rules with same match criteria and value should be mapped to
the same flow table entry regardless the flow tag identifier.
Flow tag is part of flow table entry context and not of the
destination, therefore we should return error when we try to add
destination to flow table entry with different flow tag.
Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
HW does not understand ETH_P_ALL. So treat this special case differently
and translate to 0/0 key/mask. That will allow HW to match all ethertypes.
Fixes: 7aa0f5aa90 ("mlxsw: spectrum: Implement TC flower offload")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a function to update mc_disabled from switchdev attr
SWITCHDEV_ATTR_ID_BRIDGE_MC_DISABLED
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function mlxsw_sp_port_orig_get returns the vport from the physical
port if needed, based on the original device.
This patch addresses the case where the original device is a bridge.
If it is vlan unaware bridge, it returns the matching vport. If it is vlan
aware bridge, there is no matching vport, and it returns the original port.
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The decision whether to flood a multicast packet to a port dependent
on three flags: mc_disabled, mc_router_port, mc_flood.
If mc_disabled is on, the port will be flooded according to mc_flood,
otherwise, according to mc_router_port. To accomplish that, add those
flags into the mlxsw_sp_port struct and update the mc flood table
accordingly.
Update mc_router_port by switchdev attribute
SWITCHDEV_ATTR_ID_PORT_MC_ROUTER_PORT.
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Break the bm (broadcast-multicast) into two tables, one for broadcast
(and link local multicast that behaves like bc) and one for unknown
multicasts.
Add a bool into mlxsw_sp_port named mc_flood that reflect the value this
port should have in the mc flood table (currently, always 1);
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A user that wants many bridges will use 1.Q bridge which are scalable.
One can have as many 1.Q bridges as vfids.
This patch sets their number to 1k, which is a reasonably large number.
This change is done here because the next patches will add a new flood
table, and without it, it will increase the overall size of the flood
tables dramatically.
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, there is a per port flood update function only for the UC
table. Make the function more generic by changing the table type to be
an input.
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the flood set function can't operate on only one table, but
sets both uc_flood and mb_flood together.
This patch creates a function that sets the flood state per table.
Signed-off-by: Nogah Frankel <nogahf@mellanox.com>
Signed-off-by: Yotam Gigi <yotamg@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Upon the reception of an ENTRY_REPLACE notification, resolve the FIB
node corresponding to the prefix and length and insert the new route
before the first matching entry.
Since the notification also signals the deletion of the replaced route,
delete it from the driver's cache.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a new route is appended, it's placed after existing routes sharing
the same parameters (prefix, length, table ID, TOS and priority).
While the device supports only one route with the same prefix and length
in a single table, it's important to correctly place the appended route
in the driver's cache, as when a route is deleted the next one is
programmed into the device.
Following the reception of an ENTRY_APPEND notification, resolve the
FIB node corresponding to the prefix and length and correctly place the
new entry in its entry list.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the device, routes are indexed in a routing table based on the prefix
and its length. This is in contrast to the kernel's FIB where several
FIB aliases can exist with these parameters being identical. In such
cases, the routes will be sorted by table ID (LOCAL first, then MAIN),
TOS and finally priority (metric).
During lookup, these routes will be evaluated in order. In case the
packet's TOS field is non-zero and a FIB alias with a matching TOS is
found, then it's selected. Otherwise, the lookup defaults to the route
with TOS 0 (if it exists). However, if the requested scope is narrower
than the one found, then the lookup continues.
To best reflect the kernel's datapath we should take the above into
account. Given a prefix and its length, the reflected route will always
be the first one in the FIB alias list. However, if the route has a
non-zero TOS then its action will be converted to trap instead of
forward, since we currently don't support TOS-based routing. If this
turns out to be a real issue, we can add support for that using
policy-based switching.
The route's scope can be effectively ignored as any packet being routed
by the device would've been looked-up using the widest scope (UNIVERSE).
To achieve that we need to do two changes. Firstly, we need to create
another struct (FIB node) that will hold the list of FIB entries sharing
the same prefix and length. This struct will be hashed using these two
parameters.
Secondly, we need to change the route reflection to match the above
logic, so that the first FIB entry in the list will be programmed into
the device while the rest will remain in the driver's cache in case of
subsequent changes.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The kernel resolves the nexthops for a given route using
FIB_LOOKUP_IGNORE_LINKSTATE which means a notification can be sent for a
route with one of its nexthops being LINKDOWN.
In case IGNORE_ROUTES_WITH_LINKDOWN is set for the nexthop netdev, then
we shouldn't reflect the nexthop to the device's table.
Once the nexthop netdev's carrier goes up we'll be notified using NH_ADD
and reflect it to the device.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the last IP address is removed from a netdev, its RIF is deleted.
However, if user didn't first remove neighbours and nexthops using this
interface, then they would still be present in the device's tables.
Therefore, whenever a RIF is deleted, make sure all the neighbours and
nexthops (adjacency entries) using it are removed from the relevant
tables as well.
The action associated with any route using this RIF would be refreshed,
most likely to trap. If the kernel decides to remove the route (f.e.,
because all the nexthops are now DEAD), then an event would be sent,
causing the route to be removed from the device.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a packet hits a multipath route in the device's routing table, a
hash is computed over its headers, which is then used to select the
appropriate nexthop from the device's adjacency table.
There are situations in which the kernel removes a nexthop from a
multipath route (e.g., no carrier) and the device should do the same.
Upon the reception of NH_{ADD,DEL} events, add or remove a nexthop from
the device's adjacency table and refresh all the routes using the
nexthop group. If all the nexthops of a multipath route are invalid,
then any packet hitting the route would be trapped to the CPU for
forwarding.
If all the nexthops are DEAD, then the kernel would remove the route
entirely. On the other hand, if all the nexthops are merely LINKDOWN,
then the kernel would keep the route and forward any incoming packet
using a different route.
While the last case might sound like a problem, it's expected that a
routing daemon running in user space would remove such a route from the
FIB as it's dumped with the DEAD flag set.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The device can have one of three actions associated with a route:
1) Remote - packets continue to the adjacency table
2) Local - packets continue to the neighbour table
3) Trap - packets continue to the CPU
The first two actions can also trap packets to the CPU, but they do so
using a different trap ID, which has a lower traffic class and less
allotted bandwidth.
We currently use the third action for both RTN_{LOCAL,BROADCAST} routes
and RTN_UNICAST routes not pointing to the switch ports.
However, packets that merely need to be forwarded by the switch are
likely not control packets and can be therefore scheduled towards the
CPU using a lower traffic class.
Achieve the above by assigning the third action only to local and
broadcast routes and have any other route use either of the first two
actions, based on whether the route is gatewayed or not.
This will also allow us to refresh routes using the local action and
have them trap packets when their RIF is no longer valid following a
NH_DEL event.
One side effect of this patch is that we no longer give special
treatment to multipath routes using both switch and non-switch ports
towards their nexthops. If at least one of the nexthops can be resolved,
then the device will forward the packets instead of trapping them.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The previous patch introduced a generic function to determine whether a
route should be offloaded or not. Make use of it here.
In the future we're going to add more conditions to this test (e.g.,
whether TOS is non-zero), so it makes sense to centralize it instead of
open coding it in a few places.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently set the RTNH_F_OFFLOAD flag for all routes using remote
action, but this isn't always correct. If none of the nexthops
associated with a gatewayed route can be offloaded into the device, then
any packet hitting it would be trapped to the CPU and forwarded by the
kernel.
Solve this by pushing the setting of the offload flag to after the route
was programmed into the device, thereby allowing us to take all the
parameters into account.
This change will also help us further in the patchset, when we refresh
routes following the reception of NH_{ADD,DEL} events.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The nexthop init and de-init functions both have symmetric parts
concerned with the reflection of the neighbour entry into the device's
adjacency table, in case it's used by a gatewayed route.
These sections of code also need to be called when a nexthop is marked
as valid / invalid following NH_{ADD,DEL} events. Break these out into
appropriate functions, so that they could be invoked following the
reception of above events.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After the previous changes, the FIB info is embedded in every nexthop
group struct, which in turn is embedded in every FIB entry struct.
We can therefore safely remove the FIB info from the entry struct. This
has the added advantage of making the router-related structs more
generic and suitable for use with IPv6 offloads.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Up until now, the only FIB entries that were associated with a nexthop
group were routes to remote networks where all the nexthop devices had a
valid router interface (RIF). This is in contrast to the FIB code,
where all the routes are associated with a FIB info. The same design
choice needs to be applied to the driver's cache.
Based on the NH_{ADD,DEL} events which will be added later in the
patchset, we need to be able to change the action (forward / trap)
associated with all the routes using the nexthop group. However, if we
can't link between the nexthop and the routes using it, then the above
is impossible.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The next patch is going to generalize the way in which we store routes.
Instead of attaching a nexthop group only to gatewayed routes, one will
be attached to each route, in a similar way to the way the FIB code
stores its routes.
The above means that any function operating on a nexthop group cannot
assume the group represents only gatewayed nexthops. One such function
is the one that refreshes a nexthop group and updates the adjacency
table following nexthop changes.
For a nexthop group that doesn't represent any gateways this function
would essentially be a NOP, but it would be useful if it did update the
action associated with any route using it. This will allow us to later
consolidate code paths when a nexthop changes following NH_{ADD,DEL}
events.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently use the scope of the FIB info to distinguish between a
direct unicast route and a gatewayed one. However, the kernel is
perfectly happy to configure a route with scope UNIVERSE to a directly
connected network.
Instead, we can rely on the first nexthop's scope to check if the route
is gatewayed or not.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Later in the patchset we'll add the NH_{ADD,DEL} events which will let
us know when a nexthop is considered to be dead. Based on these events
we need to be able to add or remove the nexthop from the device's
tables.
Therefore, store the private nexthop structs in a hash table and use the
kernel's fib_nh struct as the key, so that we'll be able to easily find
them when the events are received.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, when we're notified about a new RTN_UNICAST route we perform
a lookup on the nexthop group list looking for a group with a matching
configuration to that found in the FIB info. This is quite inefficient.
Instead, we can simply rely on the kernel to consolidate several FIB
configurations into the same FIB info and use the FIB info as the key
for our private nexthop group struct.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we invalidate a nexthop we should also invalidate its neighbour
entry pointer as it might be destroyed later on. This makes the nexthop
de-init function symmetric with its init and also ensures nobody will
try to access the neighbour entry.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
