Skip to content

Download Pussy888 Casino List Test Pussy888 Game List 2025 Over Types

Explore pussy888 malaysia s favorite slot game platform download the new version apk for free enjoy Why Pussy888 Stands Install Perks Play Pussy888 Slots Easy Anytime exciting slot games and experience top notch gaming start your journey with pussy888 today.

Description download pussy888 safely with Pussy888 Game Online Must University It Is Billions Guide Decent our official step by step guide for android apk ios and pc verify files avoid scams and learn updates bonuses and troubleshooting.

Pussy888 Apps Link Instant Demo Download Pussy888 Reserved Within 100

Pussy888 pussy888, download Pussy888 Game Puss888 Latest Download Pussy888 Gameplay Stable Privacy pussy888 slot apk in malaysia 2025, download the pussy888 slot client apk and ios now enjoy top online slots on our secure platform fast install updated today.

Pussy888 download apk play new version slot games in malaysia, download pussy888 apk on android or play online on ios available across southeast asia register free deposit rm30 Pussy 888 Apk Pussy For Table Pussy888 Slotgame Jackpot Slots Casino Why and claim your rm60 welcome bonus today.

Pussy888 Official Latest Pussy888 Slotgame Via With Within

Pussy888 pussy888, about us who we are at pussy888, telegram channel pussy888 slotgame jackpot, it is the ideal place to start various types of online gaming for malaysia singapore thailand and indonesia players since Pussy888 Browser Install New Pussy888 Pussy888 Total Gives Many october 2017 pussy888 includes a selection of slot games arcade games multiplayer games and live casino games for your players.

Must mbeya university of science and technology, pussy888 demo id try pussy888 games free with a test account, pussy888 pussy888, pussy 888 0 1 apk Pussy888 Science Browser Who Lepas Download Your Chance Mobile download for android pussy 888 slot machine mobile game.

Lepas download game akan dpt free credit lepas menang janji dapat withdraw, pussy888 download latest apk Pussy888 Pussy Instant Pussy888 Is Exciting Betting ios malaysia 2026, pussy888 is a renowned online gambling platform that provides a variety of fun casino style games such as slot machines table games arcade games and more.

About Us Jackpot Apps Browser Download Pussy888 Apk Why 100

Pussy888 online gambling, pussy888, leading centre of excellence for knowledge skills and applied education in Pussy888 Apk Ios Mbeya 2025 Get The Official Over Get science and technology, pussy888 ios apk free download enjoy access to over 100 games with a simple easy to use interface play anytime with smooth and stable performance.

Merged
ranahunter10 merged 5 commits into
mainfrom
adr/0004-network-intent
Nov 5, 2026
Merged

ranahunter10 merged 5 commits into
mainfrom
adr/0004-network-intent

Conversation

@ranahunter10

@ranahunter10ranahunter10 commented Nov 2, 2026•

Copy link
Copy Markdown
Owner

Stacked on #34 (→ #33 → #32).

Summary

ADR-0004 (proposed), network intent.

  • Hooks: fexit on the IPv4/IPv6 proto_ops: inet_stream_connect, inet_dgram_connect, inet_accept, inet_listen, inet_bind, inet6_bind.
  • Record: one NET record (op, verdict, endpoints, socket cookie, the listener's cookie for accepts, netns), on the bulk ring.
  • Probe:bpf/test/net_probe makes known calls and checks every field. It passes on the CI runner's kernel 7.0.0-1017-azure (make smoke-net, new CI step).
  • Why (docs/STATE.md §19): the first slice of the demo track. The same records are the process ↔ network edges the thesis graph needs.

1. What to look at first

  • ADR-0004:
    • Decision outcome: why the proto_ops and not the syscall layer. The syscall-layer draft was refused by the verifier, and that log is in the evidence.
    • Rules 2–4: the record, failures, and the bulk ring.
    • Declared limits: AF_UNIX, unconnected UDP, MPTCP/SCTP/raw.
  • bpf/test/net_probe.bpf.c:on_connect, net_accept, on_sock. The cookie is taken where sk is trusted.

2. What this PR does not do

  • No production NET records, no ring and no daemon. Those follow acceptance: demo-track items 2–3.
  • No io_uring or 32-bit socketcall test yet. Both reach these hooks by source (SRC); the test is Phase 1 step 6.

3. How it was tested

( cd~/fiery-dragon && git fetch -q && git checkout -q adr/0004-network-intent && git pull -q && \ git submodule update --init --recursive && make clean && make -j4 && make check && sudo make smoke && \ sudo make smoke-net && sudo make mutate-spine && sudo make bench-exec-cost ) 2>&1| tee ~/pr32-35-vm.txt; sha256sum ~/pr32-35-vm.txt

4. Decisions this PR assumes

ADR-0003, 0006, 0009 and 0017. Waiting: STATE decision 8, accept ADR-0004.

Checklist

  • Single purpose (one ADR and its probe)
  • Commits follow area: imperative summary and are signed off
  • Every new early return, drop, cap or fallback has a counter (probe: seen/emitted/excluded/reserve_failed per op)
  • Record layout changed? No (NET comes with the implementation)
  • Every number links to raw output
  • docs/STATE.md updated

Release note

None yet.

https://claude.ai/code/session_011UeE9rbX4ox1jUQQGbfEAZ


Generated by Claude Code

ranahunter10and others added 3 commits October 25, 2026 22:50
ADR-0019 asks that an ignored error be a build failure, that every daemon binary be hardened, and that exit() have one home. This turns those gates on before p15d exists, so p15d is written under them. - P15_MUST_CHECK (warn_unused_result) on every fallible function in wal.h, records.h and selftest.h; the two that returned -1 now return -errno. A negative-compile probe proves ignoring a result fails the build, and stops failing when the attribute is removed. - Hardening and the full warning set for every userspace binary, gcc and clang, -Werror; check_hardening.py reads the binary to prove PIE, full RELRO, NX stack, canaries, FORTIFY and CET are present. - daemon/exitcodes.h: the one exit-code table, with its test. - check_exits.py: no exit/_exit/abort in daemon/ outside sites marked "exit-ok: <reason>" (today: the self-test child only). The flags found only unchecked results in the tests (33 calls, now checked) and, under clang, two libbpf print callbacks. Refs: ADR-0019 Signed-off-by: Ashish Rana <[email protected]> Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_011UeE9rbX4ox1jUQQGbfEAZ
ADR-0017 layer 1: a spine record must take only its own len in the ring. EXEC reserved its maximum, 1,344 B, so a short command line took about 15x the space it needed (E-0070, E-0073). EXEC is now filled in a per-CPU scratch record and emitted with bpf_ringbuf_output() of exactly len. The fixed and dynptr ways stay behind -DP15_EXEC_WAY so bench/rb_cost measures the production code itself; the choice between output and dynptr is the maintainer's (decision 4). A missing scratch or refused dynptr write is counted as BUILD_FAILED, so the books are seen = emitted + reserve_failed + build_failed, and p15-spine now checks them exactly after detaching. The self-test gains a 22nd check, every record's ring size equals its len, which fails when the spine is built back to the maximum. Because the scratch buffer is reused, p15_rec_check now refuses an EXEC with slack after its fields (unless READ_FAILED) or non-zero bytes there, so a padding bug cannot leak an earlier command line unnoticed. Refs: ADR-0017, ADR-0006, ADR-0003 Signed-off-by: Ashish Rana <[email protected]> Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_011UeE9rbX4ox1jUQQGbfEAZ
p15d needs typed records for what it writes itself (ADR-0019 G1: the schema is the single source for every boundary). Each field is tied to an ADR line: CLOCK_SYNC repeats the clock anchors (ADR-0007), STATS is the capture's health per interval (ADR-0019 G4.1) and carries both sides of the ADR-0006 rule 6 books, GAP locates spine loss (ADR-0017 layer 4), END is the run's verdict (ADR-0019 G2.2, G2.3). All four are version 1, new in the lock; no existing layout changes. The generator also emits P15_SCHEMA_HASH, the hash of every layout's hash, for the WAL header's schema.hash (ADR-0007). p15_rec_check knows the new types. A fixed-size record must now have len equal to its size, and END's reason follows the same tail rules as EXEC (exact padded len, zero padding). A mutation sweep of records.c, run under ASan, kills 20 of 20. Refs: ADR-0008, ADR-0019, ADR-0017, ADR-0007, ADR-0006 Signed-off-by: Ashish Rana <[email protected]> Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_011UeE9rbX4ox1jUQQGbfEAZ
@ranahunter10
ranahunter10force-pushed the adr/0004-network-intent branch from a6341fb to a27e534CompareNovember 2, 2026 09:55
ADR-0004 must choose the hooks for network intent before the port-trace demo can exist. This probe attaches the candidates, fexit on the IPv4/IPv6 proto_ops (inet_stream_connect, inet_dgram_connect, inet_accept, inet_listen, inet_bind, inet6_bind), makes known calls from its own process (TCP over IPv4 and IPv6, UDP connect, a refused and a non-blocking connect, an AF_UNIX connect) and checks every field of every record against what the calls did: endpoints, return values, and socket cookies that link an accepted connection to its listener. The proto_ops are reached from every entry path and, being called through a pointer, can't be inlined. Their socket argument is trusted, so sock->sk is too and bpf_get_socket_cookie accepts it; the first draft hooked __sys_connect_file and do_accept, where the socket comes from file->private_data and the runner's 7.0 verifier refused the cookie helper on that untrusted pointer. The dev container refuses every fentry/fexit load (EPERM, no tracefs), so the probe runs as `make smoke-net` in CI and on the VM, not in `make smoke`. Signed-off-by: Ashish Rana <[email protected]> Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_011UeE9rbX4ox1jUQQGbfEAZ
@ranahunter10
ranahunter10force-pushed the adr/0004-network-intent branch from a27e534 to e8ca72fCompareNovember 2, 2026 09:59
The port-trace demo, and the thesis's process-to-network edges, need connect, accept, listen and bind with their endpoints and verdicts. ADR-0004 picks fexit on the IPv4/IPv6 proto_ops: every entry path (syscalls, socketcall incl. 32-bit, io_uring) reaches them, they can't be inlined, and their socket argument is trusted, so the socket cookie links an accepted connection to its listener. The probe passes every check on the CI runner's 7.0; the syscall-layer alternative was refused by the verifier. NET records go on the bulk ring so a connect flood cannot invalidate the spine. STATE gains the demo track. Signed-off-by: Ashish Rana <[email protected]> Co-Authored-By: Claude Opus 5.5 <[email protected]> Claude-Session: https://claude.ai/code/session_011UeE9rbX4ox1jUQQGbfEAZ
@ranahunter10ranahunter10 changed the title adr: 0004 network intent (probe first)adr: propose 0004, network intent (proto_ops fexit, NET record, bulk ring)Nov 2, 2026
@ranahunter10
ranahunter10 merged commit e14c80e into mainNov 5, 2026
3 checks passed
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

@ranahunter10