fddev quic-trace #3462
fddev quic-trace #3462
Conversation
Introduces scaffold to allow other processes to join fdctl quic tile data structures. - Add fd_quic_tile.h - Add a magic number to fd_quic_ctx_t - Rename tiles/fd_quic.c => tiles/fd_quic_tile.c to fix object name collision with waltz/fd_quic.c.
ripatel-fd
force-pushed
the
ripatel/quic-trace
branch
from
2132a19 to
653ff7c
Compare
| #define STEM_CALLBACK_AFTER_FRAG after_frag | ||
| #include "../../../disco/stem/fd_stem.c" | ||
|
|
||
| fd_topo_run_tile_t fd_tile_quic_trace_rx = { |
There was a problem hiding this comment.
I don't think you need this? Not used anywhere.
There was a problem hiding this comment.
@mmcgee-jump I'd like to keep it. I want this to work as a real time eventually
| if( !metrics ) FD_LOG_ERR(( "out of memory" )); | ||
| fd_metrics_register( metrics ); | ||
|
|
||
| /* ... redirect fseq updates */ |
There was a problem hiding this comment.
I don't think this does anything? The QUIC tile has already. booted once you do this, so wouldn't read these fseq pointers again?
There was a problem hiding this comment.
Oh lol I see how this works ... kind of an interesting strategy to just pretend it's a QUIC tile, but I guess works for now.
There was a problem hiding this comment.
@mmcgee-jump I don't see any other way. We need to consume net frags but we also need decryption keys.
If we skip ahead of the real QUIC tile then the packets we see might not have encryption keys available.
There was a problem hiding this comment.
Oh I thought your comment was about something else.
Eventually, I want to make it less hacky and make it a real tile. But I wasn't going to bother with these fdctl APIs just for a debugging tool
Relates to #3461