Özgür Kesim
a9f0f27ee2
main.go: - reading /proc - iteration over entries in NNN/maps - filter glob-search for "*python3*" in pathname - find symbol and its offset in pathnanme - calculate offset in memory - add pid and offset to map TODO: encapsulating this into a module ebpf.go: - added type MapFD int, changing all function on a FD to methods This allows us to enrich the data type going forward - added bpf_update_elem() from the manpage ebpf2. .updateElement() is the verbatim wrapper to it. - added .Add/.Change/.Set methods, which call .updateElement with specific flags TODO: re-implement ebpf.go with pure go, using direct syscalls. |
||
---|---|---|
GetRuntimeAddresses | ||
README.md | ||
runforever.py | ||
scratchpad.txt | ||
TODO.md |
Overview
The goal of this exercise is to construct a Go program that populates an eBPF
map with the address in memory of the _PyRuntime
global variable for all live
Python 3.7 processes. This variable is typically found within libpython
, which
is dynamically loaded. The target system can be assumed to be a 64-bit Linux OS
running a 5.x kernel.
As an example, say there are two live processes that have mapped
libpython3.7m.so.1.0
(on your OS libpython
may have a different name) into
their address spaces, with process IDs 100 and 200. In process 100 the library
is mapped at address 0x10000, and in process 200 the library is mapped at
address 0x20000. Furthermore, assume that _PyRuntime
is at offset 256
within libpython3.7m.so.1.0
.
Your solution should populate an eBPF map with the following two entries: (100 -> 0x10256), (200 -> 0x20256).
Your Solution
Treat your solution as if this was code you would have to live with for a while, rather than just a throw-away proof-of-concept. i.e. ideally it should be modular, with a reasonable breakdown of functionality into packages and functions as required, documented and formatted as per the Go standards, with any tests as may be necessary (aiming for full test coverage isn't necessary. Use your judgement as to what may need a test and what may not).
When you have completed your solution, create a pull request for it to the master branch.
Task 1
Enumerate all live processes that contain a Python 3.7 interpreter. Note that this includes both Python REPL processes that have Python as their main binary, and other processes that have libpython.*.so loaded in their address space.
Task 2
For all processes containing a Python interpreter determine the address of the
_PyRuntime
global variable within each process. In Python 3.7 this variable is
in the dynamic symbols of the Python library and the symbol information will be
available regardless of whether or not debug symbols are available. Note that
due to address space layout randomization the actual address of the variable
within each process will differ.
Task 3
For all live processes containing a Python interpreter, add a mapping from
process ID to the address of _PyRuntime
to an eBPF map. The skeleton solution
provides an ebpf
package with a function for creating a map, as well as a
function for reading a map. You need to add a function for writing to the map.
Getting Started
You will need to following in order to complete the exercise:
-
A functioning Go installation. See https://golang.org for information on how to get started with this if you have not already done so.
-
A functioning Python 3.7 install. Run
python3 --version
to see what the default installed version of Python is on your system. It is also necessary for your Python install to be dynamically linked againstlibpython
. To check if this is the case runldd /path/to/python3.7
and check for alibpython
entry in the output. If there is no such entry the easiest solution is to download the Python 3.7 source code from https://www.python.org/downloads/release/python-376/ and compile/install it via./configure --prefix=`pwd`/install --enable-shared && make && make install
. Once that completes you will have a Python interpreter at./install/bin/python3.7
that you can use. -
A running Linux kernel with eBPF enabled. It is enabled by default on most modern kernel configurations. To check if this is the case for yours, search the config of your running kernel for
CONFIG_BPF=y
. On Fedora this can be done viacat /boot/config-$(uname -r)| grep -i "CONFIG_BPF=y"
.
Clone this repository into $GOPATH/src/github.com/optimyze-interviews
. A
skeleton Go program is provided in the GetRuntimeAddresses
sub-directory. From
within GetRuntimeAddresses
you can build the program via go build
. This will
create a binary called ./GetRuntimeAddresses
. In its current state it will
simply create an eBPF map, attempt to print the contents of this map, and then
exit. The output would look something like the following:
$ ./GetRuntimeAddresses
Created eBPF map (FD: 3)
Printing contents of map 3
Finally, a Python script called runforever.py
is also provided. When run via
python3.7 ./runforever.py
this script simply prints the ID of the Python process and
enters an infinite loop. You can run a few of these simulataneously to test your
solution, if you wish.