You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Make the userspace agent set the BPF map sizes based on configured
options. This allows more suitable map sizes to be used during run
time than the static limits set in the BPF programs. This avoids
wasting memory by using unnecessary large maps, and might slightly
improve hashmap lookup performance by sizing them based on the
expected number of entries (small enough that many entries may fit in
cache, large enough to avoid excessive hash collisions).
Scale the histogram map based on the expected number of histograms,
the PID and ifindex filter maps to fit the largest key they need to
include and the cgroup filter map to fit all tracked cgroups.
This also fixes a bug where the maximum allowed PID (PID_MAX_LIMIT)
and ifindex (IFINDEX_MAX) did not fit in their corresponding filter
maps (off-by-one error).
Signed-off-by: Simon Sundberg <[email protected]>
0 commit comments