There are two files in the tree called libbpf.h which is becoming
problematic.  Most samples don't actually need the local libbpf.h
they simply include it to get to bpf/bpf.h.  Include bpf/bpf.h
directly instead.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Daniel Borkmann says:

====================
This series follows up mostly with with some minor cleanups on top
of 'Move ld_abs/ld_ind to native BPF' as well as implements better
32/64 bit immediate load into register and saves tail call init on
cBPF for the arm64 JIT. Last but not least we add a couple of test
cases. For details please see individual patches. Thanks!

v1 -> v2:
  - Minor fix in i64_i16_blocks() to remove 24 shift.
  - Added last two patches.
  - Added Acks from prior round.
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Add test cases where we combine semi-random imm values, mainly for testing
JITs when they have different encoding options for 64 bit immediates in
order to reduce resulting image size.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

We can trivially save 4 bytes in prologue for cBPF since tail calls
can never be used from there. The register push/pop is pairwise,
here, x25 (fp) and x26 (tcc), so no point in changing that, only
reset to zero is not needed.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Improve the JIT to emit 64 and 32 bit immediates, the current
algorithm is not optimal and we often emit more instructions
than actually needed. arm64 has movz, movn, movk variants but
for the current 64 bit immediates we only use movz with a
series of movk when needed.

For example loading ffffffffffffabab emits the following 4
instructions in the JIT today:

  * movz: abab, shift:  0, result: 000000000000abab
  * movk: ffff, shift: 16, result: 00000000ffffabab
  * movk: ffff, shift: 32, result: 0000ffffffffabab
  * movk: ffff, shift: 48, result: ffffffffffffabab

Whereas after the patch the same load only needs a single
instruction:

  * movn: 5454, shift:  0, result: ffffffffffffabab

Another example where two extra instructions can be saved:

  * movz: abab, shift:  0, result: 000000000000abab
  * movk: 1f2f, shift: 16, result: 000000001f2fabab
  * movk: ffff, shift: 32, result: 0000ffff1f2fabab
  * movk: ffff, shift: 48, result: ffffffff1f2fabab

After the patch:

  * movn: e0d0, shift: 16, result: ffffffff1f2fffff
  * movk: abab, shift:  0, result: ffffffff1f2fabab

Another example with movz, before:

  * movz: 0000, shift:  0, result: 0000000000000000
  * movk: fea0, shift: 32, result: 0000fea000000000

After:

  * movz: fea0, shift: 32, result: 0000fea000000000

Moreover, reuse emit_a64_mov_i() for 32 bit immediates that
are loaded via emit_a64_mov_i64() which is a similar optimization
as done in 6fe8b9c1

 ("bpf, x64: save several bytes by using
mov over movabsq when possible"). On arm64, the latter allows to
use a single instruction with movn due to zero extension where
otherwise two would be needed. And last but not least add a
missing optimization in emit_a64_mov_i() where movn is used but
the subsequent movk not needed. With some of the Cilium programs
in use, this shrinks the needed instructions by about three
percent. Tested on Cavium ThunderX CN8890.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Follow-up to 816d9ef3

 ("bpf, arm64: remove ld_abs/ld_ind") in
that the extra 4 byte JIT scratchpad is not needed anymore since it
was in ld_abs/ld_ind as stack buffer for bpf_load_pointer().

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

The extra skb_copy_bits() buffer is not used anymore, therefore
remove the extra 4 byte stack space requirement.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Make the RETPOLINE_{RA,ED}X_BPF_JIT() a bit more readable by
cleaning up the macro, aligning comments and spacing.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Since fe83963b

 ("bpf, sparc64: remove ld_abs/ld_ind") it's not
used anymore therefore remove it.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

The ool_skb_header_pointer() and size_to_len() is unused same as
tmp_offset, therefore remove all of them.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Updated optstring parameter for getopt_long() to accept short options.
Also updated usage() function.

Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Song Liu says:
====================
Changes v2 -> v3:
  Improve syntax based on suggestion by Tobin C. Harding.

Changes v1 -> v2:
  1. Rename some variables to (hopefully) reduce confusion;
  2. Check irq_work status with IRQ_WORK_BUSY (instead of work->sem);
  3. In Kconfig, let BPF_SYSCALL select IRQ_WORK;
  4. Add static to DEFINE_PER_CPU();
   5. Remove pr_info() in stack_map_init().
====================

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

This new test captures stackmap with build_id with hardware event
PERF_COUNT_HW_CPU_CYCLES.

Because we only support one ips-to-build_id lookup per cpu in NMI
context, stack_amap will not be able to do the lookup in this test.
Therefore, we didn't do compare_stack_ips(), as it will alwasy fail.

urandom_read.c is extended to run configurable cycles so that it can be
caught by the perf event.

Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Currently, we cannot parse build_id in nmi context because of
up_read(&current->mm->mmap_sem), this makes stackmap with build_id
less useful. This patch enables parsing build_id in nmi by putting
the up_read() call in irq_work. To avoid memory allocation in nmi
context, we use per cpu variable for the irq_work. As a result, only
one irq_work per cpu is allowed. If the irq_work is in-use, we
fallback to only report ips.

Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Jakub Kicinski says:

====================
This series started out as a follow up to the bpftool perf event dumping
patches.

As suggested by Daniel patch 1 makes use of PERF_SAMPLE_TIME to simplify
code and improve accuracy of timestamps.

Remaining patches are trying to move perf event loop into libbpf as
suggested by Alexei.  One user for this new function is bpftool which
links with libbpf nicely, the other, unfortunately, is in samples/bpf.
Remaining patches make samples/bpf link against full libbpf.a (not just
a handful of objects).  Once we have full power of libbpf at our disposal
we can convert some of XDP samples to use libbpf loader instead of
bpf_load.c.  My understanding is that this is the desired direction,
at least for networking code.
====================

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Now that we can use full powers of libbpf in BPF samples, we
should perhaps make the simplest XDP programs not depend on
bpf_load helpers.  This way newcomers will be exposed to the
recommended library from the start.

Use of bpf_prog_load_xattr() will also make it trivial to later
on request offload of the programs by simply adding ifindex to
the xattr.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

BPF programs only have to specify the target kernel version for
tracing related hooks, in networking world that requirement does
not really apply.  Loosen the checks in libbpf to reflect that.

bpf_object__open() users will continue to see the error for backward
compatibility (and because prog_type is not available there).

Error code for NULL file name is changed from ENOENT to EINVAL,
as it seems more appropriate, hopefully, that's an OK change.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Fix spelling mistakes, improve and clarify the language of comments
in libbpf.h.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

There are two copies of event reading loop - in bpftool and
trace_helpers "library".  Consolidate them and move the code
to libbpf.  Return codes from trace_helpers are kept, but
renamed to include LIBBPF prefix.

Suggested-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

samples/bpf currently cherry-picks object files from tools/lib/bpf
to link against.  Just compile the full library and link statically
against it.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Both tools/lib/bpf/libbpf.h and samples/bpf/bpf_load.h define their
own version of struct bpf_map_def.  The version in bpf_load.h has
more fields.  libbpf does not support inner maps and its definition
of struct bpf_map_def lacks the related fields.  Rename the definition
in bpf_load.h (samples/bpf) to avoid conflicts.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Ask the kernel to include sample time in each even instead of
reading the clock.  This is also more accurate because our
clock reading was done when user space would dump the buffer,
not when sample was produced.

Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Sync the header from include/uapi/linux/bpf.h which was updated to add
fib lookup helper function. This fixes selftests/bpf build failure.

Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

'|& ...' is a bash 4.0+ construct which is not guaranteed to be available
when using '$(shell ...)' in a Makefile. Fall back to the more portable
'2>&1 | ...'.

Fixes the following warning during compilation:

	/bin/sh: 1: Syntax error: "&" unexpected

Signed-off-by: Joe Stringer <joe@wand.net.nz>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

David Ahern says:

====================
Provide a helper for doing a FIB and neighbor lookup in the kernel
tables from an XDP program. The helper provides a fastpath for forwarding
packets. If the packet is a local delivery or for any reason is not a
simple lookup and forward, the packet is expected to continue up the stack
for full processing.

The response from a FIB and neighbor lookup is either the egress index
with the bpf_fib_lookup struct filled in with dmac and gateway or
0 meaning the packet should continue up the stack. In time we can
revisit this to return the FIB lookup result errno if it is one of the
special RTN_'s such as RTN_BLACKHOLE (-EINVAL) so that the XDP
programs can do an early drop if desired.

Patches 1-6 do some more refactoring to IPv6 with the end goal of
extracting a FIB lookup function that aligns with fib_lookup for IPv4,
basically returning a fib6_info without creating a dst based entry.

Patch 7 adds lookup functions to the ipv6 stub. These are needed since
bpf is built into the kernel and ipv6 may not be built or loaded.

Patch 8 adds the bpf helper and 9 adds a sample program.

v3
- remove ETH_ALEN and in6_addr from uapi header

v2
- removed pkt_access from bpf_func_proto as noticed by Daniel
- added check in that IPv6 forwarding is enabled
- added DaveM's ack on patches 1-7 and 9 based on v1 response and
  fact that no changes were made to them in v2

v1
- updated commit messages and cover letter
- added comment to sample program noting lack of verification on
  egress device supporting XDP

RFC v2
- fixed use of foward helper from cls_act as noted by Daniel
- in patch 1 rename fib6_lookup_1 as well for consistency
====================

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Simple example of fast-path forwarding. It has a serious flaw
in not verifying the egress device index supports XDP forwarding.
If the egress device does not packets are dropped.

Take this only as a simple example of fast-path forwarding.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Provide a helper for doing a FIB and neighbor lookup in the kernel
tables from an XDP program. The helper provides a fastpath for forwarding
packets. If the packet is a local delivery or for any reason is not a
simple lookup and forward, the packet continues up the stack.

If it is to be forwarded, the forwarding can be done directly if the
neighbor is already known. If the neighbor does not exist, the first
few packets go up the stack for neighbor resolution. Once resolved, the
xdp program provides the fast path.

On successful lookup the nexthop dmac, current device smac and egress
device index are returned.

The API supports IPv4, IPv6 and MPLS protocols, but only IPv4 and IPv6
are implemented in this patch. The API includes layer 4 parameters if
the XDP program chooses to do deep packet inspection to allow compare
against ACLs implemented as FIB rules.

Header rewrite is left to the XDP program.

The lookup takes 2 flags:
- BPF_FIB_LOOKUP_DIRECT to do a lookup that bypasses FIB rules and goes
  straight to the table associated with the device (expert setting for
  those looking to maximize throughput)

- BPF_FIB_LOOKUP_OUTPUT to do a lookup from the egress perspective.
  Default is an ingress lookup.

Initial performance numbers collected by Jesper, forwarded packets/sec:

       Full stack    XDP FIB lookup    XDP Direct lookup
IPv4   1,947,969       7,074,156          7,415,333
IPv6   1,728,000       6,165,504          7,262,720

These number are single CPU core forwarding on a Broadwell
E5-1650 v4 @ 3.60GHz.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Add stubs to retrieve a handle to an IPv6 FIB table, fib6_get_table,
a stub to do a lookup in a specific table, fib6_table_lookup, and
a stub for a full route lookup.

The stubs are needed for core bpf code to handle the case when the
IPv6 module is not builtin.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Similar to IPv4, IPv6 should use the FIB lookup result in the
tracepoint.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Add IPv6 equivalent to fib_lookup. Does a fib lookup, including rules,
but returns a FIB entry, fib6_info, rather than a dst based rt6_info.
fib6_lookup is any where from 140% (MULTIPLE_TABLES config disabled)
to 60% faster than any of the dst based lookup methods (without custom
rules) and 25% faster with custom rules (e.g., l3mdev rule).

Since the lookup function has a completely different signature,
fib6_rule_action is split into 2 paths: the existing one is
renamed __fib6_rule_action and a new one for the fib6_info path
is added. fib6_rule_action decides which to call based on the
lookup_ptr. If it is fib6_table_lookup then the new path is taken.

Caller must hold rcu lock as no reference is taken on the returned
fib entry.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Move source address lookup from fib6_rule_action to a helper. It will be
used in a later patch by a second variant for fib6_rule_action.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

ip6_pol_route is used for ingress and egress FIB lookups. Refactor it
moving the table lookup into a separate fib6_table_lookup that can be
invoked separately and export the new function.

ip6_pol_route now calls fib6_table_lookup and uses the result to generate
a dst based rt6_info.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Rename rt6_multipath_select to fib6_multipath_select and export it.
A later patch wants access to it similar to IPv4's fib_select_path.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Rename fib6_lookup to fib6_node_lookup to better reflect what it
returns. The fib6_lookup name will be used in a later patch for
an IPv6 equivalent to IPv4's fib_lookup.

Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

For me, as a reader whose mother language isn't English, the
old words bring a little difficulty to catch the meaning, this
patch rewords the subsection in a more clarificatory way.

This patch also add blank lines as separator at two places
to improve readability.

Signed-off-by: Wang YanQing <udknight@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Update .gitignore files.

Signed-off-by: Sirio Balmelli <sirio@b-ad.ch>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

The BPF selftests fail to build with missing headers
'asm/bitsperlong.h' and 'asm/errno.h'.

These already exist in 'tools/arch/[arch]/include';
add architecture-agnostic header files in 'tools/include/uapi'
to reference them.

Signed-off-by: Sirio Balmelli <sirio@b-ad.ch>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

i386 builds report:
  net/xdp/xdp_umem.o: In function `xdp_umem_reg':
  xdp_umem.c:(.text+0x47e): undefined reference to `__udivdi3'

This fix uses div_u64 instead of the GCC built-in.

Fixes: c0c77d8f

 ("xsk: add user memory registration support sockopt")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

Jakub Kicinski says:

====================
This small series adds a feature which extends BPF offload beyond
a pure host processing offload and firmly into the realm of
heterogeneous processing.  Allowing offloaded XDP programs to set
the RX queue index opens the door for defining fully programmable
RSS/n-tuple filter replacement.  In fact the device datapath will
skip the RSS processing completely if BPF decided on the queue
already, making the XDP program replace part of the standard NIC
datapath.

We hope some day the entire NIC datapath will be defined by BPF :)
====================

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>

BPF has access to all internal FW datapath structures.  Including
the structure containing RX queue selection.  With little coordination
with the datapath we can let the offloaded BPF select the RX queue.
We just need a way to tell the datapath that queue selection has already
been done and it shouldn't overwrite it.  Define a bit to tell datapath
BPF already selected a queue (QSEL_SET), if the selected queue is not
enabled (>= number of enabled queues) datapath will perform normal RSS.

BPF queue selection on the NIC can be used to replace standard
datapath RSS with fully programmable BPF/XDP RSS.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>