Gargoyle 1.15.x OpenWrt 24.10 - 2026-06-13
Moderator: Moderators
Re: Gargoyle 1.15.x OpenWrt 24.10 - 2026-06-13
I'm not sure of Lantis plans
Re: Gargoyle 1.15.x OpenWrt 24.10 - 2026-06-13
I will find time to merge them upstream and compile, but I now have a backlog of changes to review first.
I had a large influx of changes to review in a single 48 hour window and more have followed.
It takes time.
I had a large influx of changes to review in a single 48 hour window and more have followed.
It takes time.
https://lantisproject.com/downloads/gargoylebuilds for the latest releases
Please be respectful when posting. I do this in my free time on a volunteer basis.
https://lantisproject.com/blog
Please be respectful when posting. I do this in my free time on a volunteer basis.
https://lantisproject.com/blog
Re: Gargoyle 1.15.x OpenWrt 24.10 - 2026-06-13
It was a mistakeI had a large influx of changes to review in a single 48 hour window and more have followed.
I blame AI, its the devil i say
Re: Gargoyle 1.15.x OpenWrt 24.10 - 2026-06-13
Thanks a lot for taking the time to fix these ispyisail and for generally helping Lantis, I wish I had you skills so that I could be of further assistance besides testing and logging!ispyisail wrote: ↑Thu Jun 25, 2026 3:02 am@fiye68
Thanks for the detailed log — it made the bugs straightforward to track down.
There are two separate problems here:
1. nftables script bugs (now fixed)
Found five issues across three files in the gargoyle-tor package, all nftables migration problems:
Also fixed the CPU spike: update_tor_nftset was calling nft add element once per relay IP (~10,000 times per minute). Replaced with a single batched nft -f - call.
PR submitted: https://github.com/lantis1008/gargoyle/pull/59
2. RAM (hardware limit — not fixable)
The OOM kill at 50% bootstrap is a separate problem:
Tor needs ~50 MB just to load relay descriptors. With Gargoyle's full stack running on 128 MB that leaves no headroom. OpenSSL 3 has increased Tor's memory footprint and this appears to be the new floor. The WRT3200ACM won't be able toCode: Select all
tor pid 6631 — 60484kB virtual, 51380kB RSS → killed by OOM
run Tor reliably regardless of the script fixes — 256 MB is the practical minimum now.
If anyone has confirmed working hardware on current Gargoyle + Tor I'd be interested to know the model.
If the fixes were in part aided by automation tools, I am sure that they were thoroughly checked before merging them to ensure the continuous quality of gargoyle's code.
When there is time and suitable hardware, I will try to test Tor on routers with faster CPU and larger RAM and report back.