The ability of the clang rewriter to extract the type information for
some types of structs, unions, and pointers to the aforementioned was
somewhat buggy. This became exposed in a test_clang case after a user
upgraded to a newer kernel, wherein the struct definition changed. The
functionality in question is only used to pass json-ified representation
of the struct to python in order to program the Key/Leaf metaclass.

Improve support for this and other types, including unions.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

make sure LDFLAGS are propagated as well

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

Fix breakage from LLVM 3.8 API change

There is some difference in how to take a pointer from an iterator. This
code should work for both 3.7 and 3.8 APIs.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

p4: fixed bugs in control-flow generation

Honor the external C/CXX FLAGS

Signed-off-by: Gary Lin <glin@suse.com>

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Automatically bump memlock ulimit

Instead of requiring the user to bump the ulimit in their shell before
starting a bcc script, try to setrlimit automatically when a failure
occurs. Since there is no getrusage for memlock limit, unfortunately we
have to brute force setting the limit. For now, just try bpf() once and
then try to set unlimited ulimit, then try bpf() again.

Fixes: #281
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Fixup dependencies of bcc-tools package

This should depend on python-bcc, which itself depends on libbcc.

Fixes: #291
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Add bcc-tools package

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Ticket: #291
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Fixup broken symlink tests/cc/simulation.py

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

added P4 deparser synthesis

Updates to INSTALL.md for readability and ease of use

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

A stable 4.3 kernel is available, suggest users use that version
instead.

Move LLVM build instructions to the end, prefer that they pull llvm
binaries instead.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Fix unary operator handling of probe reads with parens

Testing for bpf_probe_read should not include parenethes when walking
the tree, since the inner operation will have already been rewritten.

Fixes: #289
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Edit README, link path of some example files

As some example files were reorganized into subdirecotries, some
pathes in README bring connection on 404. So it have to be new pathes.

Fixes #285 - build process assumes git use

Add bpf() syscall feature check and readme update

Fixes: #283
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Add perf_output support for high rate events

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Incoporating feedback from bgregg:
Make per-cpu nature of output buffer be hidden and implicit. As such,
incoporate some rewriting into the definition of the PERF_OUTPUT.
Create two different macros to distinguish the perf_array (hardware
counters) from the perf_output (ring buffer) use cases.

Also, rename perf_output to perf_submit.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

This adds support for the bpf_perf_event_output command. This is
intended for per-process events from bpf to userspace at high rate. The
events from the bpf program can be completely customized.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>