Rename dsa_get_dst to dsa_tree_find since it doesn't increment the
reference counter, rename dsa_add_dst to dsa_tree_alloc for symmetry
with dsa_tree_free, and provide a convenient dsa_tree_touch function to
find or allocate a new tree.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Provide convenient dsa_tree_get and dsa_tree_put functions scoping a DSA
tree used to increment and decrement its reference counter, instead of
poking directly its kref structure.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
DSA trees have a refcount used to automatically free the dsa_switch_tree
structure once there is no switch devices inside of it.
The refcount is incremented when a switch is added to the tree, and
decremented when it is removed from it.
But because of kref_init, the refcount is also incremented at
initialization, and when looking up the tree from the list for symmetry.
Thus the current code stores the number of switches plus one, and makes
the switch registration more complex.
To simplify the switch registration function, we reset the refcount to
zero after initialization and don't increment it when looking up a tree.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Similarly to a DSA switch and port, rename the tree index from "tree" to
"index" and make it an unsigned int because it isn't supposed to be less
than 0.
u32 is an OF specific data used to retrieve the value and has no need to
be propagated up to the tree index.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define the DSA switch index as an unsigned int, because it will never be
less than 0.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
netdev->rx_dropped was including packets dropped by napi_gro_receive.
If a packet is dropped by network stack, it should not be counted under
driver Rx dropped.
Made necessary changes to not include network stack drops under
netdev->rx_dropped.
Signed-off-by: Intiyaz Basha <intiyaz.basha@cavium.com>
Signed-off-by: Satanand Burla <satananda.burla@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The two functions were declared as static inline in a header file. There
is no particular reason why they should be inlined, they just happened to
remain in the same header file when they were turned from macros to
functions in a precious commit.
Make them non-inlined functions and move them to common.c file instead.
Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jakub Kicinski says:
====================
bpf: add offload as a first class citizen
This series is my stab at what was discussed at a recent IOvisor
bi-weekly call. The idea is to make the device translator run at
the program load time. This makes the offload more explicit to
the user space. It also makes it easy for the device translator
to insert information into the original verifier log.
v2:
- include linux/bug.h instead of asm/bug.h;
- rebased on top of Craig's verifier fix (no changes, the last patch
just removes more code now). I checked the set doesn't conflict
with Jiri's, Josef's or Roman's patches, but missed Craig's fix :(
v1:
- rename the ifindex member on load;
- improve commit messages;
- split nfp patches more.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Thanks to the ability to load a program for a specific device,
running verifier twice is no longer needed.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Following steps are taken in the driver to offload an XDP program:
XDP_SETUP_PROG:
* prepare:
- allocate program state;
- run verifier (bpf_analyzer());
- run translation;
* load:
- stop old program if needed;
- load program;
- enable BPF if not enabled;
* clean up:
- free program image.
With new infrastructure the flow will look like this:
BPF_OFFLOAD_VERIFIER_PREP:
- allocate program state;
BPF_OFFLOAD_TRANSLATE:
- run translation;
XDP_SETUP_PROG:
- stop old program if needed;
- load program;
- enable BPF if not enabled;
BPF_OFFLOAD_DESTROY:
- free program image.
Take advantage of the new infrastructure. Allocation of driver
metadata has to be moved from jit.c to offload.c since it's now
done at a different stage. Since there is no separate driver
private data for verification step, move temporary nfp_meta
pointer into nfp_prog. We will now use user space context
offsets.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct nfp_prog is currently only used internally by the translator.
This means there is a lot of parameter passing going on, between
the translator and different stages of offload. Simplify things
by allocating nfp_prog in offload.c already.
We will now use kmalloc() to allocate the program area and only
DMA map it for the time of loading (instead of allocating DMA
coherent memory upfront).
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most of offload/translation prepare logic will be moved to
offload.c. To help git generate more reasonable diffs
move nfp_prog_prepare() and nfp_prog_free() functions
there as a first step.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Firmware supports live replacement of programs for quite some
time now. Remove the software-fallback related logic and
depend on the FW for program replace. Seamless reload will
become a requirement if maps are present, anyway.
Load and start stages have to be split now, since replace
only needs a load, start has already been done on add.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently create a fake cls_bpf offload object when we want
to offload XDP. Simplify and clarify the code by moving the
TC/XDP specific logic out of common offload code. This is easy
now that we don't support legacy TC actions. We only need the
bpf program and state of the skip_sw flag.
Temporarily set @code to NULL in nfp_net_bpf_offload(), compilers
seem to have trouble recognizing it's always initialized. Next
patches will eliminate that variable.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
BPF offload's main header does not need to include nfp_net.h.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The register renumbering was removed and will not be coming back
in its old, naive form, given that it would be fundamentally
incompatible with calling functions. Remove the leftovers.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only support BPF_PROG_TYPE_SCHED_CLS programs in direct
action mode. This simplifies preparing the offload since
there will now be only one mode of operation for that type
of program. We need to know the attachment mode type of
cls_bpf programs, because exit codes are interpreted
differently for legacy vs DA mode.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If TC program is loaded with skip_sw flag, we should allow
the device-specific programs to be accepted.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass the netdev pointer to bpf_prog_get_type(). This way
BPF code can decide whether the device matches what the
code was loaded/translated for.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If program is bound to a device, print the name of the relevant
interface or unknown if the netdev has since been removed.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Extend struct bpf_prog_info to contain information about program
being bound to a device. Since the netdev may get destroyed while
program still exists we need a flag to indicate the program is
loaded for a device, even if the device is gone.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The fact that we don't know which device the program is going
to be used on is quite limiting in current eBPF infrastructure.
We have to reverse or limit the changes which kernel makes to
the loaded bytecode if we want it to be offloaded to a networking
device. We also have to invent new APIs for debugging and
troubleshooting support.
Make it possible to load programs for a specific netdev. This
helps us to bring the debug information closer to the core
eBPF infrastructure (e.g. we will be able to reuse the verifer
log in device JIT). It allows device JITs to perform translation
on the original bytecode.
__bpf_prog_get() when called to get a reference for an attachment
point will now refuse to give it if program has a device assigned.
Following patches will add a version of that function which passes
the expected netdev in. @type argument in __bpf_prog_get() is
renamed to attach_type to make it clearer that it's only set on
attachment.
All calls to ndo_bpf are protected by rtnl, only verifier callbacks
are not. We need a wait queue to make sure netdev doesn't get
destroyed while verifier is still running and calling its driver.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ndo_xdp is a control path callback for setting up XDP in the
driver. We can reuse it for other forms of communication
between the eBPF stack and the drivers. Rename the callback
and associated structures and definitions.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make use of the swap macro and remove unnecessary variables skb and cnt.
This makes the code easier to read and maintain.
This code was detected with the help of Coccinelle.
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.
Addresses-Coverity-ID: 114893
Addresses-Coverity-ID: 114894
Addresses-Coverity-ID: 114895
Addresses-Coverity-ID: 114896
Addresses-Coverity-ID: 114897
Addresses-Coverity-ID: 114898
Addresses-Coverity-ID: 114899
Addresses-Coverity-ID: 114900
Addresses-Coverity-ID: 114901
Addresses-Coverity-ID: 114902
Addresses-Coverity-ID: 114903
Addresses-Coverity-ID: 114904
Addresses-Coverity-ID: 114905
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_init_nondata_skb() is fed with freshly allocated skbs.
They already have a cleared csum field, no need to clear it again.
This is based on Neal review on commit 3b11775033 ("tcp: do not mangle
skb->cb[] in tcp_make_synack()"), noticing I did not clear skb->csum.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reduce one indentation level to make code more readable.
tcp_sync_mss() can be factorized.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can now build this driver on ARM, so I ran into a randconfig build
warning that presumably had existed on powerpc already.
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c: In function 'sg_fd_to_skb':
drivers/net/ethernet/freescale/dpaa/dpaa_eth.c:1712:18: error: 'skb' may be used uninitialized in this function [-Werror=maybe-uninitialized]
I'm slightly changing the logic here, to make it obvious to the
compiler that 'skb' is always initialized.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Flavio Leitner says:
====================
Allow openvswitch to query ports in another netns.
Today Open vSwitch users are moving internal ports to other namespaces and
although packets are flowing OK, the userspace daemon can't find out basic
information like if the port is UP or DOWN, for instance.
This patchset extends openvswitch API to retrieve the current netnsid of
a port. It will be used by the userspace daemon to find out in which netns
the port is located.
This patchset also extends the rtnetlink getlink call to accept and operate
on a given netnsid. More details are available in each patch.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, when an application gets netnsid from the kernel (for example as
the result of RTM_GETLINK call on one end of the veth pair), it's not much
useful. There's no reliable way to get to the netns fd from the netnsid, nor
does any kernel API accept netnsid.
Extend the RTM_GETLINK call to also accept netnsid. It will operate on the
netns with the given netnsid in such case. Of course, the calling process
needs to have enough capabilities in the target name space; for now, require
CAP_NET_ADMIN. This can be relaxed in the future.
To signal to the calling process that the kernel understood the new
IFLA_IF_NETNSID attribute in the query, it will include it in the response.
This is needed to detect older kernels, as they will just ignore
IFLA_IF_NETNSID and query in the current name space.
This patch implemetns IFLA_IF_NETNSID only for get and dump. For set
operations, this can be extended later.
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch allows reliable identification of netdevice interfaces connected
to openvswitch bridges. In particular, user space queries the netdev
interfaces belonging to the ports for statistics, up/down state, etc.
Datapath dump needs to provide enough information for the user space to be
able to do that.
Currently, only interface names are returned. This is not sufficient, as
openvswitch allows its ports to be in different name spaces and the
interface name is valid only in its name space. What is needed and generally
used in other netlink APIs, is the pair ifindex+netnsid.
The solution is addition of the ifindex+netnsid pair (or only ifindex if in
the same name space) to vport get/dump operation.
On request side, ideally the ifindex+netnsid pair could be used to
get/set/del the corresponding vport. This is not implemented by this patch
and can be added later if needed.
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
By using CQE based moderation on TX CQ we can reduce the number of TX
interrupt rate. Besides the benefit of less interrupts, this also
allows the kernel to better utilize TSO. Since TSO has some CPU overhead,
it might not aggregate when CPU is under high stress. By reducing the
interrupt rate and the CPU utilization, we can get better aggregation
and better overall throughput.
The feature is enabled by default and has a private flag in ethtool
for control.
Throughput, interrupt rate and TSO utilization improvements:
(ConnectX-4Lx 40GbE, unidirectional, 1/16 TCP streams, 64B packets)
---------------------------------------------------------
Metric | Streams | CQE Based | EQE Based | improvement
---------------------------------------------------------
BW | 1 | 2.4Gb/s | 2.15Gb/s | +11.6%
IR | 1 | 27Kips | 50.6Kips | -46.7%
TSO Util | 1 | 74.6% | 71% | +5%
BW | 16 | 29Gb/s | 25.85Gb/s | +12.2%
IR | 16 | 482Kips | 745Kips | -35.3%
TSO Util | 16 | 69.1% | 49% | +41.1%
*BW = Bandwidth, IR = Interrupt rate, ips = interrupt per second.
TSO Util = bytes in TSO sessions / all bytes transferred
Signed-off-by: Tal Gilboa <talgi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
This is needed in order to enlarge it with more members that will get
value of 0 when not set.
Signed-off-by: Rabie Loulou <rabiel@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The NIC TC offload table size was hard coded to 1k. Change it to be
min(max NIC RX table size,
min(max flow counters, 64k) * num flow groups)
where the max values are read from the firmware and the number of
flow groups is hard-coded as before this change.
We don't know upfront the division of flows to groups (== different masks).
This setup allows each group to be of size up to the where we want to go
(when supported, all offloaded flows use counters). Thus, we don't expect
multiple occurences for a group which in turn would add steering hops.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Add debug print when changing the configuration of QoS through dcbnl.
Use ethtool -s <devname> msglvl hw on/off to toggle debug messages.
Signed-off-by: Inbar Karmy <inbark@mellanox.com>
Reviewed-by: Gal Pressman <galp@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
If the port is in DSCP trust state, packets are placed in the right
priority queue based on the dscp value. This is done by selecting
the transmit queue based on the dscp of the skb.
Until now select_queue honors priority only from the vlan header.
However that is not sufficient in cases where port trust state is DSCP
mode as packet might not even contain vlan header. Therefore if the port
is in dscp trust state and vport's min inline mode is not NONE,
copy the IP header to the eseg's inline header if the skb has it.
This is done by changing the transmit queue sq's min inline mode to L3.
Note that the min inline mode of sqs that belong to other features such
as xdpsq, icosq are not modified.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
This patch implements dcbnl hooks to set and delete DSCP to priority map
as defined by the DCB subsystem. Device maintains internal trust state
which needs to be set to DSCP state for performing DSCP to priority mapping.
When the first dscp to priority APP entry is added by the user, the
trust state is changed to dscp.
When the last dscp to priority APP entry is deleted by the user, the
trust state is changed to pcp.
If user sends multiple dscp to priority APP entries on the same dscp,
the last sent one will take effect. All the previous sent will be
deleted.
The dscp to priority APP entries are added and deleted in the net/dcb
APP database using dcb_ieee_setapp/getapp.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The QPTS register allows changing the priority trust state between pcp and
dscp. Add support to get/set trust state from device. When the port is
in pcp/dscp trust state, packet is routed by hardware to matching priority
based on its pcp/dscp value respectively.
The QPDPM register allow channing the dscp to priority mapping. Add support
to get/set dscp to priority mapping from device.
Note that to change a dscp mapping, the "e" bit of this dscp structure
must be set in the QPDPM firmware command.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
The QCAM register provides capability bit for all the QoS registers
using ACCESS_REG command.
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Reviewed-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
IN6_ADDR_HSIZE is private to addrconf.c, move it here to avoid
confusion.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
pktgen accidentally used IN6_ADDR_HSIZE, instead of using the size of an
IPv6 address.
Since IN6_ADDR_HSIZE recently was increased from 16 to 256, this old
bug is hitting us.
Fixes: 3f27fb2321 ("ipv6: addrconf: add per netns perturbation in inet6_addr_hash()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently n->flags is being operated on by a logical && operator rather
than a bitwise & operator. This looks incorrect as these should be bit
flag operations. Fix this.
Detected by CoverityScan, CID#1460398 ("Logical vs. bitwise operator")
Fixes: 245dc5121a ("net: sched: cls_u32: call block callbacks for offload")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Average RTT is 32-bit thus full 64-bit division is redundant.
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Suggested-by: Stephen Hemminger <stephen@networkplumber.org>
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some time ago Eric Dumazet suggested a "hack the IFF_XMIT_DST_RELEASE
flag on the vlan netdev". But the last comment was "does not support
properly bonding/team.(If the real_dev->privflags IFF_XMIT_DST_RELEASE
bit changes, we want to update all the vlans at the same time )"
I've extended that patch to support changes of IFF_XMIT_DST_RELEASE in
bonding/team.
Both bonding and team call netdev_change_features() after recalculation
of features including priv_flags IFF_XMIT_DST_RELEASE bit. So the only
thing needed to support is to recheck this bit in
vlan_transfer_features().
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Vadim Fedorenko <vfedorenko@yandex-team.ru>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes the following sparse warnings:
drivers/net/phy/phylink.c:570:6: warning:
symbol 'phylink_phy_change' was not declared. Should it be static?
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
