Daniel Borkmann says:
====================
pull-request: bpf-next 2020-08-04
The following pull-request contains BPF updates for your *net-next* tree.
We've added 73 non-merge commits during the last 9 day(s) which contain
a total of 135 files changed, 4603 insertions(+), 1013 deletions(-).
The main changes are:
1) Implement bpf_link support for XDP. Also add LINK_DETACH operation for the BPF
syscall allowing processes with BPF link FD to force-detach, from Andrii Nakryiko.
2) Add BPF iterator for map elements and to iterate all BPF programs for efficient
in-kernel inspection, from Yonghong Song and Alexei Starovoitov.
3) Separate bpf_get_{stack,stackid}() helpers for perf events in BPF to avoid
unwinder errors, from Song Liu.
4) Allow cgroup local storage map to be shared between programs on the same
cgroup. Also extend BPF selftests with coverage, from YiFei Zhu.
5) Add BPF exception tables to ARM64 JIT in order to be able to JIT BPF_PROBE_MEM
load instructions, from Jean-Philippe Brucker.
6) Follow-up fixes on BPF socket lookup in combination with reuseport group
handling. Also add related BPF selftests, from Jakub Sitnicki.
7) Allow to use socket storage in BPF_PROG_TYPE_CGROUP_SOCK-typed programs for
socket create/release as well as bind functions, from Stanislav Fomichev.
8) Fix an info leak in xsk_getsockopt() when retrieving XDP stats via old struct
xdp_statistics, from Peilin Ye.
9) Fix PT_REGS_RC{,_CORE}() macros in libbpf for MIPS arch, from Jerry Crunchtime.
10) Extend BPF kernel test infra with skb->family and skb->{local,remote}_ip{4,6}
fields and allow user space to specify skb->dev via ifindex, from Dmitry Yakunin.
11) Fix a bpftool segfault due to missing program type name and make it more robust
to prevent them in future gaps, from Quentin Monnet.
12) Consolidate cgroup helper functions across selftests and fix a v6 localhost
resolver issue, from John Fastabend.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes coccicheck warning:
./drivers/net/ethernet/broadcom/bnxt/bnxt.c:3730:19-37: WARNING:
dma_alloc_coherent use in stats -> hw_stats already zeroes out
memory, so memset is not needed
dma_alloc_coherent use in status already zeroes out memory,
so memset is not needed
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Li Heng <liheng40@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now we can report all the full 64-bit CPU endian software accumulated
counters instead of the hw counters, some of which may be less than
64-bit wide. Define the necessary macros to access the software
counters.
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that we have the infrastructure in place, add the new function
bnxt_accumulate_all_stats() to periodically accumulate and check for
counter rollover of all ring stats and port stats.
A chip bug was also discovered that could cause some ring counters to
become 0 during DMA. Workaround by ignoring zeros on the affected
chips.
Some older frimware will reset port counters during ifdown. We need
to check for that and free the accumulated port counters during ifdown
to prevent bogus counter overflow detection during ifup.
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If supported by newer firmware, make the firmware call to query all
the port counter masks. If not supported, assume 40-bit port
counter masks.
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Newer firmware has a new call HWRM_FUNC_QSTATS_EXT to retrieve the
masks of all ring counters. Make this call when supported to
initialize the hardware masks of all ring counters. If the call
is not available, assume 48-bit ring counter masks on P5 chips.
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some of these DMAed hardware counters are not full 64-bit counters and
so we need to accumulate them as they overflow. Allocate copies of these
DMA statistics memory blocks with the same size for accumulation. The
hardware counter widths are also counter specific so we allocate
memory for masks that correspond to each counter.
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver manages multiple statistics structures of different sizes.
They are all allocated, freed, and handled practically the same. Define
a new bnxt_stats_mem structure and common allocation and free functions
for all staistics memory blocks.
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Main changes are 200G support and fixing the definitions of discard and
error counters to match the hardware definitions.
Because the HWRM_PORT_PHY_QCFG message size has now exceeded the max.
encapsulated response message size of 96 bytes from the PF to the VF,
we now need to cap this message to 96 bytes for forwarding. The forwarded
response only needs to contain the basic link status and speed information
and can be capped without adding the new information.
v2: Fix bnxt_re compile error.
Cc: Selvin Xavier <selvin.xavier@broadcom.com>
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove PCIe non-counters display from ethtool statistics, as
they are not simple counters but register dump. The next few
patches will add logic to detect counter roll-over and it won't
work with these PCIe non-counters.
There will be a follow up patch to get PCIe information via
ethtool register dump.
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that BPF program/link management is centralized in generic net_device
code, kernel code never queries program id from drivers, so
XDP_QUERY_PROG/XDP_QUERY_PROG_HW commands are unnecessary.
This patch removes all the implementations of those commands in kernel, along
the xdp_attachment_query().
This patch was compile-tested on allyesconfig.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200722064603.3350758-10-andriin@fb.com
The UDP reuseport conflict was a little bit tricky.
The net-next code, via bpf-next, extracted the reuseport handling
into a helper so that the BPF sk lookup code could invoke it.
At the same time, the logic for reuseport handling of unconnected
sockets changed via commit efc6b6f6c3
which changed the logic to carry on the reuseport result into the
rest of the lookup loop if we do not return immediately.
This requires moving the reuseport_has_conns() logic into the callers.
While we are here, get rid of inline directives as they do not belong
in foo.c files.
The other changes were cases of more straightforward overlapping
modifications.
Signed-off-by: David S. Miller <davem@davemloft.net>
The fact that NETIF_F_HW_TC is not set should be a sufficient
indication to the user that TC offloads are not supported.
No need to bother users of older firmware versions with
pointless warnings on every boot.
Also, since the support is optional, bnxt_init_tc() should not
return an error in case FW is old, similarly to how error
is not returned when CONFIG_BNXT_FLOWER_OFFLOAD is not set.
With that we can add an error message to the caller, to warn
about actual unexpected failures.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previously, shared blocks were only relevant for the pseudo-qdiscs ingress
and clsact. Recently, a qevent facility was introduced, which allows to
bind blocks to well-defined slots of a qdisc instance. RED in particular
got two qevents: early_drop and mark. Drivers that wish to offload these
blocks will be sent the usual notification, and need to know which qdisc it
is related to.
To that end, extend flow_block_offload with a "sch" pointer, and initialize
as appropriate. This prompts changes in the indirect block facility, which
now tracks the scheduler in addition to the netdevice. Update signatures of
several functions similarly.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current completion ring sizing formula is wrong with TPA enabled.
The formula assumes that the number of TPA completions are bound by the
RX ring size, but that's not true. TPA_START completions are immediately
recycled so they are not bound by the RX ring size. We must add
bp->max_tpa to the worst case maximum RX and TPA completions.
The completion ring can overflow because of this mistake. This will
cause hardware to disable the completion ring when this happens,
leading to RX and TX traffic to stall on that ring. This issue is
generally exposed only when the RX ring size is set very small.
Fix the formula by adding bp->max_tpa to the number of RX completions
if TPA is enabled.
Fixes: c0c050c58d ("bnxt_en: New Broadcom ethernet driver.");
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In a shared port PHY configuration, async event is received when any of the
port modifies the configuration. Ethtool link settings should be
initialised after updated PHY configuration from firmware.
Fixes: b1613e78e9 ("bnxt_en: Add async. event logic for PHY configuration changes.")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver was modified to not rely on rtnl lock to protect link
settings about 2 years ago. The pause setting was missed when
making that change. Fix it by acquiring link_lock mutex before
calling bnxt_hwrm_set_pause().
Fixes: e2dc9b6e38 ("bnxt_en: Don't use rtnl lock to protect link change logic in workqueue.")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert to new infra, taking advantage of sleeping in callbacks.
v2:
- use bp->*_fw_dst_port_id != INVALID_HW_RING_ID as indication
that the offload is active.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, devlink_port_attrs_set accepts a long list of parameters,
that most of them are devlink port's attributes.
Use the devlink_port_attrs struct to replace the relevant parameters.
Signed-off-by: Danielle Ratson <danieller@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bare-metal use cases require giving firmware and the embedded
application processor control over VLAN offloads. The driver should
not attempt to override or utilize this feature in such scenarios
since it will not work as expected.
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The hardware VLAN offload feature on our NIC does not have separate
knobs for handling customer and service tags on RX. Either offloading
of both must be enabled or both must be disabled. Introduce definitions
for the combined feature set in order to clean up the code and make
this constraint more clear. Technically these features can be separately
enabled on TX, however, since the default is to turn both on, the
combined TX feature set is also introduced for code consistency.
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With the new infrastructure in place, we can now support the setting of
the indirection table from ethtool.
When changing channels, in a rare case that firmware cannot reserve the
rings that were promised, we will still try to keep the RSS map and only
revert to default when absolutely necessary.
v4: Revert RSS map to default during ring change only when absolutely
necessary.
v3: Add warning messages when firmware cannot reserve the requested RX
rings, and when the RSS table entries have to change to default.
v2: When changing channels, if the RSS table size changes and RSS map
is non-default, return error.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that we have the logical indirection table, we can return these
proper logical indices directly to ethtool -x instead of the physical
IDs.
Reported-by: Jakub Kicinski <kicinski@fb.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that we have the logical table, we can fill the HW RSS table
using the logical table's entries and converting them to the HW
specific format. Re-initialize the logical table to standard
distribution if the number of RX rings changes during ring reservation.
v4: Use bnxt_get_rxfh_indir_size() to get the RSS table size.
v2: Use ALIGN() to roundup the RSS table size.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On some chips, this varies based on the number of RX rings. Add this
helper function and refactor the existing code to use it.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver currently does not keep track of the logical RSS indirection
table. The hardware RSS table is set up with standard default ring
distribution when initializing the chip. This makes it difficult to
support user sepcified indirection table entries. As a first step, add
the logical table in the main bnxt structure and allocate it according
to chip specific table size. Add a function that sets up default
RSS distribution based on the number of RX rings.
v4: Use bnxt_get_rxfh_indir_size() for the current RSS table size.
v2: Use kmalloc_array() since we init. all entries afterwards.
Use ALIGN() to roundup the RSS table size.
Use ethtool_rxfh_indir_default() to init. the default entries.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix up bnxt_get_rxfh_indir_size() to return the proper current RSS
table size for P5 chips. Change it to non-static so that bnxt.c
can use it to get the table size.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, we allocate one page for the hardware DMA RSS indirection
table. While the size is currently big enough for all chips, future
chip variations may support bigger sizes, so it is better to calculate
and store the chip specific size and allocate accordingly.
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Minor overlapping changes in xfrm_device.c, between the double
ESP trailing bug fix setting the XFRM_INIT flag and the changes
in net-next preparing for bonding encryption support.
Signed-off-by: David S. Miller <davem@davemloft.net>
Virtual functions does not have VPD information. This patch modifies
calling bnxt_read_vpd_info() only for PFs and avoids an unnecessary
error log.
Fixes: a0d0fd70fe ("bnxt_en: Read partno and serialno of the board from VPD")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On older firmware, the hardware statistics are not cleared when the
driver frees the hardware stats contexts during ifdown. The driver
expects these stats to be cleared and saves a copy before freeing
the stats contexts. During the next ifup, the driver will likely
allocate the same hardware stats contexts and this will cause a big
increase in the counters as the old counters are added back to the
saved counters.
We fix it by making an additional firmware call to clear the counters
before freeing the hw stats contexts when the firmware is the older
20.x firmware.
Fixes: b8875ca356 ("bnxt_en: Save ring statistics before reset.")
Reported-by: Jakub Kicinski <kicinski@fb.com>
Reviewed-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Tested-by: Jakub Kicinski <kicinski@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Older firmware may not support legacy TX push properly and may not
be disabling it. So we check certain firmware versions that may
have this problem and disable legacy TX push unconditionally.
Fixes: c0c050c58d ("bnxt_en: New Broadcom ethernet driver.")
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently only store the firmware version as a string for ethtool
and devlink info. Store it also as a version code. The next 2
patches will need to check the firmware major version to determine
some workarounds.
We also use the 16-bit firmware version fields if the firmware is newer
and provides the 16-bit fields.
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the representor is removed, then identify the indirect flow_blocks
that need to be removed by the release callback and the port representor
structure. To identify the port representor structure, a new
indr.cb_priv field needs to be introduced. The flow_block also needs to
be removed from the driver list from the cleanup path.
Fixes: 1fac52da59 ("net: flow_offload: consolidate indirect flow_block infrastructure")
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prepare fix the bug in the next patch. use flow_indr_block_cb_alloc/remove
function and remove the __flow_block_indr_binding.
Signed-off-by: wenxu <wenxu@ucloud.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds a drop frames counter to tc flower offloading.
Reporting h/w dropped frames is necessary for some actions.
Some actions like police action and the coming introduced stream gate
action would produce dropped frames which is necessary for user. Status
update shows how many filtered packets increasing and how many dropped
in those packets.
v2: Changes
- Update commit comments suggest by Jiri Pirko.
Signed-off-by: Po Liu <Po.Liu@nxp.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
AER reset should follow the same steps as suspend/resume. We need to
free context memory during AER reset and allocate new context memory
during recovery by calling bnxt_hwrm_func_qcaps(). We also need
to call bnxt_reenable_sriov() to restore the VFs.
Fixes: bae361c54f ("bnxt_en: Improve AER slot reset.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If VFs are enabled, we need to re-configure them during resume because
firmware has been reset while resuming. Otherwise, the VFs won't
work after resume.
Fixes: c16d4ee0e3 ("bnxt_en: Refactor logic to re-enable SRIOV after firmware reset detected.")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The separate steps we do in bnxt_resume() can be done more simply by
calling bnxt_hwrm_func_qcaps(). This change will add an extra
__bnxt_hwrm_func_qcaps() call which is needed anyway on older
firmware.
Fixes: f9b69d7f62 ("bnxt_en: Fix suspend/resume path on 57500 chips")
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Register ndo callback via flow_indr_dev_register() and
flow_indr_dev_unregister().
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
xdp_umem.c had overlapping changes between the 64-bit math fix
for the calculation of npgs and the removal of the zerocopy
memory type which got rid of the chunk_size_nohdr member.
The mlx5 Kconfig conflict is a case where we just take the
net-next copy of the Kconfig entry dependency as it takes on
the ESWITCH dependency by one level of indirection which is
what the 'net' conflicting change is trying to ensure.
Signed-off-by: David S. Miller <davem@davemloft.net>
The explicit mask and shift is not the appropriate way to parse fields
out of a little endian struct. The length field is internally __le16
and the strategy employed only happens to work on little endian machines
because the offset used is actually incorrect (length is at offset 6).
Also remove the related and no longer used definitions from bnxt.h.
Fixes: 845adfe40c ("bnxt_en: Improve valid bit checking in firmware response message.")
Signed-off-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When NVRAM directory is not found, return the error code
properly as per firmware command failure instead of the hardcode
-ENOBUFS.
Fixes: 3a707bed13 ("bnxt_en: Return -EAGAIN if fw command returns BUSY")
Signed-off-by: Vasundhara Volam <vasundhara-v.volam@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
