Add printf for key/leaf and fix trace_printk bug

There was an issue where the rewrite of bpf_trace_printk combined with
conversion of function argument to ctx->$reg was mangling the text. Fix
up this case and add a test.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

This extends upon the sscanf reader functionality, with the intent of
providing key/leaf pretty printing
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Fix recursive loop in parsing pointer to self struct

Issue occurs in the description visitor class, when the struct used in a
map key is a pointer to self, as in:

struct node;
struct node {
  struct node *next;
};

Avoid this in the desc by using "unsigned long long" for all pointers.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Fixing more ctrl-c problem

add connectivity validation in distributed bridge examples

Add one line desc output for simple_tc

Signed-off-by: Wei-Chun Chao <weichunc@plumgrid.com>

Signed-off-by: Yonghong Song <yhs@plumgrid.com>

Signed-off-by: Yonghong Song <yhs@plumgrid.com>

  o Also suggest task_switch.py to try with libbcc package
Signed-off-by: Yonghong Song <yhs@plumgrid.com>

Patch series: reorganize bpf_module and add sscanf feature for fuse

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

This is the culmination of the previous patches. It adds an api that can
input map data in a string format, but validating the numbers and
locations of data fields. The use case is for fuse file input/output. A
printf api may follow.

Take the table with key/leaf of:
  struct Key { int a; int b; };
  struct Leaf { int a; int b; int c; struct SubLeaf { int x; int y; } s; };

One would input to this table using:
  update_table(table_name, "{1 2}", "{1 2 -3 {9 0xa}}");

The implementation uses a JITed function for each unique type, that is
invoked to run sscanf on behalf of the caller. The input must have the
exact right number of arguments. Bit fields are supported, but the
caller must be aware of the collapse of those bitfields into an aligned
field, as well as endianness.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

It was bothering me that a whole bunch of accesses to the tables were
based on a string hash lookup, in addition to the vectors that were kept
to convert ids to names. Switch those around, as well as the api
internally.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

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

After the modules have been created, create a helper function for each
table leaf/key type.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

This whole project is about bpf, prefixing everything with the same
acronym is redundant.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

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

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

Add cmake checks for required test programs

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

Set as warning, since the build will still work fine and be able to
create a binary package.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Move helper.h prototype for tail call into 4.2 section

With #115, bpf_tail_call was incorrectly disabled in 4.2 builds.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

Add INSTALL.md instructions for using Ubuntu binary build

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

Flags for 4.3 were accidentally committed, revert

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

release IPDB in host ns early

We need it for setting vxlan flow based flags currently. Release
it once the interfaces are set up. This does not help Ctrl-C on
main.py though.
Signed-off-by: Wei-Chun Chao <weichunc@plumgrid.com>

Suggest Release and add X86 target in readme

The default CMAKE_BUILD_TYPE is Debug when cloning from git, suggest
Release instead.

Also, there are some worrisome warnings during cmake when X86 target is
omitted. Additionally, I may want X86 target support for a future
feature, so lets suggest it here.
Signed-off-by: Brenden Blanco <bblanco@plumgrid.com>

llvm 3.7+ deb is broken

https://llvm.org/bugs/show_bug.cgi?id=24154

Build from source until it's resolved. Another plan is to provide
binary package.
Signed-off-by: Wei-Chun Chao <weichunc@plumgrid.com>

better clean up of created namespace/interfaces if ctrl-c is pressed

fix tunnel_mesh.py

The IP assignment was wrong.
Signed-off-by: Wei-Chun Chao <weichunc@plumgrid.com>

  o There are still some cases when ctrl-c is pressed, the program does not exit
    gracefully, with a however lower probability. In my random manual experiments,
    one in ten cases it still hangs and needs further ctrl-z.
Signed-off-by: Yonghong Song <yhs@plumgrid.com>