This repo will contain code for search backend of eBPF projects annotations. Current implementation will be based on elasticsearch as the backend.
- Copy all annotated dbs that need to be searched into the
dbsfolder. - Run elastic_loader.py to load annotated dbs into an elastic search backend as follows
python3 elastic_loader.py --help
usage: elastic_loader.py [-h] --index INDEX
Indexing elasticsearch documents.
options:
-h, --help show this help message and exit
--index INDEX Elasticsearch index name.
-
List/count all functions which use a user-input bpf_helper
-
List/count all maps used in a user-input repository
-
List/count all functions which read from a user-input map in a repository
-
List/count all functions which write to a user-input map in a repository
-
List the function call graph for a user-input function in a repository
-
List/count all functions which have a user-input call_depth in a repository
-
List/count all functions with a user-input value for a chosen user-input field (say startline == 35 etc.)
-
For a input function find all maps which are only read (and not written) across its FCG.
-
For a input function find all maps which are only written (and not read) across its FCG.
-
For a input function find all maps which are read as well as written across its FCG.
-
For a given FCG, for all maps in that FCG, (Identify producer-consumer relation among functions.)
a. find functions which write to a map, and any of its successor function which reads the same map.
b. find functions which read a map, and any of its successor function which writes the same map.
- Create a consolidated explainability for a function, by identifying for a user input function's FCG:
a. All compatible hookpoints. (this will be a intersection of compatible hook-points for all functions).
b. All bpf_helpers called through FCG (function call graph).
c. All human extracted comments. (please think of a good data-structure, which encompasses comments (human, AI, developer) as well as FCG.
- Avg FCG size for root functions -- should be average computation on results of [5]