{"id":51497,"date":"2026-04-07T18:53:33","date_gmt":"2026-04-07T22:53:33","guid":{"rendered":"https:\/\/mjtsai.com\/blog\/?p=51497"},"modified":"2026-04-10T13:54:26","modified_gmt":"2026-04-10T17:54:26","slug":"tahoe-tcp-overflow-bug","status":"publish","type":"post","link":"https:\/\/mjtsai.com\/blog\/2026\/04\/07\/tahoe-tcp-overflow-bug\/","title":{"rendered":"Tahoe TCP Overflow Bug"},"content":{"rendered":"

Photon<\/a> (Hacker News<\/a>):<\/p>\n

\n

After exactly 49 days, 17 hours, 2 minutes, and 47 seconds of continuous uptime, a 32-bit unsigned integer overflow in Apple’s XNU kernel freezes the internal TCP timestamp clock. Once frozen, TIME_WAIT connections never expire, ephemeral ports slowly exhaust, and eventually no new TCP connections can be established at all. ICMP (ping) keeps working. Everything else dies. The only fix most people know is a reboot.<\/p>\n

[…]<\/p>\n

This is a 32-bit unsigned integer timer wraparound<\/strong> bug in the TCP subsystem, specifically a TCP timestamp counter overflow<\/strong>. The counter in question, tcp_now<\/code>, is the kernel’s internal TCP clock. When it stops ticking, every timer in the TCP stack that depends on it stops working.<\/p>\n<\/blockquote>\n\n

They suggest that the bug may have been around since Catalina, but I’ve had a Mac server running from the Catalina days all the way through Sequoia, with months of uptime, and haven’t seen this problem. I’ve not updated the server to Tahoe yet.<\/p>\n\n

Previously:<\/p>\n