Very easy to use, procfs parsing library in C++.
Build
Run cmake . && make
Currently supported CMake configuration flags:
BUILD_SHARED_LIBS=<ON|OFF>: ON to compile a shared library. OFF to compile a static library (DEFAULT: ON)
You can pass any number of those to the cmake command: cmake -D<CONFIG_FLAG>=<VALUE> .
Build using clang
If you prefer using clang, just configure the compiler while running cmake:
CXX=<clang++> CC=<clang> cmake .
After that, just use make as always.
Integrate
- Compile as a shared or static library.
- Add the contents of
/libinto your link directories - Add the contents of
/includeinto your include directories. That's it, you are good to go.
Features
- Parsing system-wide information from files directly under
/procfs. Seeprocfs.hppfor all the supported files. - Parsing per-task (processes and threads) information from files under
/procfs/[task-id]/. Seetask.hppfor all the supported files. - Parsing network information from files under
/procfs/net(which is an alias to/procfs/self/netnowadays)
Requirements
- The library requires C++11 or newer
- The library aims to support Linux kernel versions >= 2.6.32.
Notes
- All APIs and function calls might throw
std::bad_allocexceptions when allocations of standard containers such asstd::stringfail. - APIs are thread-safe. There are no internal states/members/caches that might be affected by simultaneous calls.
- Objects do NOT handle data caching. All the APIs are pure getters that always(!) fetch the information from the filesystem.
- The location of the procfs filesystem is configurable. Just create the
procfsobject with the right path for your machine.
Samples
The directory sample contains a full blown application that calls all(!) the supported APIs and prints all the information gathered. When compiling the library, the sample applications is compiled as well.
Anyway, here are some cool examples:
Example 1: Iterater over all the loaded unsigned or out-of-tree kernel modules
auto pfs = pfs::procfs();
auto modules = pfs.get_modules();
for (const auto& module : modules)
{
if (module.is_out_of_tree || module.is_unsigned)
{
... do your work ...
}
}
Example 2: Find all the memory maps for task 1234 (This can be both a process or a thread) that start with an ELFs header
auto task = pfs::procfs().get_task(1234);
auto mem = task.get_mem();
for (auto& map : task.get_maps())
{
if (!map.perm.can_read)
{
continue;
}
static const std::vector<uint8_t> ELF_HEADER = { 0x7F, 0x45, 0x4C, 0x46 };
if (mem.read(map.start_address, ELF_HEADER.size()) == ELF_HEADER)
{
... do your work ...
}
}
(You can either create pfs every time or once and keep it, the overhead is really small)
Example 3: Iterate over all the IPv4 TCP sockets currently in listening state (in my current network namespace):
// Same as pfs::procfs().get_task().get_net().get_tcp()
for (auto& socket : pfs::procfs().get_net().get_tcp())
{
if (socket.current_state == socket::state::listen)
{
... do your work ...
}
}
(API behaves similar to the procfs, where /proc/net is a soft link to /proc/self/net)
Example 4: Get all the IPv6 UDP sockets in the root network namespace belonging to a specific user ID:
for (auto& socket : pfs::procfs().get_task(1).get_net().get_udp6())
{
if (socket.uid == <some-uid-value>)
{
... do your work ...
}
}
Example 5: Check if the process catches SIGSTOP signals
auto status = pfs::procfs().get_task(1234).get_status();
bool handles_sigstop = status.sig_cgt.is_set(pfs::signal::sigstop);
674 Dec 30, 2022
1.1k Jan 3, 2023
106 Dec 22, 2022
9 Nov 8, 2021
28 Dec 20, 2022
1.2k Jan 1, 2023
13 Sep 13, 2022
13 Feb 7, 2022
6.6k Jan 8, 2023
1.3k Jan 8, 2023
400 Dec 25, 2022
526 Jan 2, 2023
84 Dec 18, 2022
305 Dec 13, 2022
16.3k Dec 27, 2022
2.1k Jan 1, 2023
260 Dec 28, 2022
2.1k Jan 6, 2023
1.6k Jan 8, 2023
831 Dec 28, 2022