Dear Maintainer,
please update ppp to version 2.5.3

*** Reporter, please consider answering these questions, where appropriate ***

   * What led up to the situation?

  /etc/ppp/peers/tng had defaultroute6. Before installing the IPv6 default
route, pppd's defaultroute6_exists() reads /proc/net/ipv6_route line-by-line -
a seq_file where every 4 KB read re-walks the kernel FIB trie across all
tables. Since ~May 22, bird carries a full BGP IPv6 table (~249k routes, kernel
table 100). After each reconnect, bird re-converges and re-inserts routes at
the same time pppd scans - a mutual livelock (verified live: pppd spinning in
userspace state=R/empty kernel stack, bird at 52 % CPU with maxed 212 KB
netlink Recv-Q, table growing ~120 routes/s). The scan took up to 15 minutes;
pppd couldn't answer the BRAS's LCP echoes meanwhile, so the BRAS killed the
session, and the cycle repeated - each round shorter as the table shrank -
until one attempt completed in time. That's why reestablishment "took a long
time" since May 23.

   * What exactly did you do (or not do) that was effective (or
     ineffective)?

Workaround:

1. Removed defaultroute6 from /etc/ppp/peers/tng (backup: tng.bak-20260604)
2. Added /etc/ppp/ipv6-up.d/00defaultroute: ip -6 route replace default dev "$IFACE" metric 1024 - pure netlink, instant at any table size, produces the identical route, auto-removed when ppp0 drops
3. Restarted pppd (it only reads the peers file at startup) - reconnect now completes in ~3 seconds, ConfAck→LL-address in the same second, bird unaffected, default route present

Fix: Update to version 2.5.3

*** End of the template - remove these template lines ***