Jesper Dangaard Brouer

I need to figure out which is the best Kernel timestamping API to use, when we need to implement jitter measurements.

Implement monitoring of the TTL (Time-To-Live) value in the UDP packets. This is useful for detecting when a stream changes path in the network. This involves: 1. Small kernel module...

Increment operator cause unnecessary data races by always calling bpf_map_update_elem

2

### What reproduces the bug? Provide code if possible. The simple increment operator `@++` cause unnecessary data races by always calling `bpf_map_update_elem`. This simple bpftrace oneliner: `sudo bpftrace -e 'kprobe:cgroup_rstat_flush_locked...

Minor fixes to areas/latency/cgroup_rstat_tracepoint.bt

1

The bpftrace increment operator is both slow and contain data races. Fix comment, where I thought slowness was due to it being an atomic operation. Upstream bpftrace issues are in...

bpftrace isn't handling bool from tracepoint correctly

9

### What reproduces the bug? Provide code if possible. In my script `cgroup_rstat_tracepoint.bt` I notice that tracepoint decoding `bool` was returning wrong values (e.g. 129). - See [this commit](https://github.com/xdp-project/xdp-project/pull/102/commits/46b330e6dd600f16) for...

This benchmark module depend on a kernel feature (currently) called "traits" Tested on top of kernel tree and branch: - https://github.com/arthurfabre/linux/tree/afabre/traits-002-bounds-inline I had to export the following functions via `EXPORT_SYMBOL_GPL()`...

The feature under test is (currently) called "traits" - https://github.com/arthurfabre/linux/tree/afabre/traits-002-bounds-inline It is a compressed key-value store, that live in the top of the XDP packet data frame, just after the...