tetragon: Add event config map #53
Conversation
olsajiri
force-pushed
the
kill_btf
branch
10 times, most recently
from
c481f4b to
bbbaced
Compare
|
I'm now suddenly getting this panic: need to figure this out, moving to draft again |
must be my setup, increased the timeout and all's passed |
Did a
so there does not seem to be a big difference between the two. |
|
I've seen local test timeouts like this when I've done a lot of |
I had several failures due to context timeout with this test when running on loaded machine. The test is loading 4 kprobes so the load/verification can cross 20 second limit, increasing it to 60s for this test. Signed-off-by: Jiri Olsa <[email protected]>
We should display ArgM type in here. Signed-off-by: Jiri Olsa <[email protected]>
Add map with config data so it can replace BTF config data completely. Starting with func_id, other fields will come in following changes. The new internal config_map map is loaded for both tracepoints and kprobes and loaded with config data at zero index. Signed-off-by: Jiri Olsa <[email protected]>
Adding arg types to the config map and using them in bpf code instead of the argX BTF fields. Signed-off-by: Jiri Olsa <[email protected]>
Adding arg meta values to the config map and using them in bpf code instead of the argXm BTF fields. Signed-off-by: Jiri Olsa <[email protected]>
Adding tp ctx offset values to the config map and using them in bpf code instead of the t_argX_ctx_off BTF fields. Signed-off-by: Jiri Olsa <[email protected]>
Adding sigkill value to the config map and using it in bpf code instead of the sigkill BTF field. Signed-off-by: Jiri Olsa <[email protected]>
Adding syscall value to the config map and using it in bpf code instead of the syscall BTF field. Signed-off-by: Jiri Olsa <[email protected]>
Adding argreturncopy type to the config map and using it in bpf code instead of the argreturncopy BTF field. Signed-off-by: Jiri Olsa <[email protected]>
Adding argreturn type to the config map and using it in bpf code instead of the argreturn BTF field. Signed-off-by: Jiri Olsa <[email protected]>
Removing BTF enums and code which is no longer needed, because we have all the data in the config map. Signed-off-by: Jiri Olsa <[email protected]>
right, that could easily happened |
olsajiri
changed the title
Getting rid of extra configuration inside BTF and moving it to map.
The new internal config_map map is loaded for both tracepoints
and kprobes and loaded with config data at zero index.
Signed-off-by: Jiri Olsa [email protected]