JdeBP
@JdeBP@mastodonapp.uk
This is a general, but non-politics, account covering everything from computer programming to supermarket meal deals; any politics will be taken up by @JdeBP. For #senryu and #SlowLife tasks, see @JdeBP. For the command-line and system tools (including #nosh and #djbwares), specifically, see @JdeBP.
mastodonapp.uk
JdeBP
@JdeBP@mastodonapp.uk
This is a general, but non-politics, account covering everything from computer programming to supermarket meal deals; any politics will be taken up by @JdeBP. For #senryu and #SlowLife tasks, see @JdeBP. For the command-line and system tools (including #nosh and #djbwares), specifically, see @JdeBP.
mastodonapp.uk
@JdeBP@mastodonapp.uk
·
Apr 09, 2026
Here are some things that one can add to the analysis of the MacOS TCP timeout clock freeze bug.
The code for calculate_tcp_clock() in XNU was changed in May 2025. Older versions of this function (e.g. in xnu-11417) worked quite differently and wouldn't have stopped ticking the clock at 32-bit unsigned integer wraparound.
None of #NetBSD, #FreeBSD, nor #OpenBSD share this exact way of doing TCP timeout processing with #XNU.
FreeBSD does not have a tcp_now and works off the global 32-bit ticks variable. OpenBSD effectively works off the kernel's system clock, too, but with a randomized offset, and does 64-bit unsigned modular arithmetic. NetBSD uses a distinct 32-bit unsigned tcp_now counter that it simply increments by 1 at regular intervals, and does modular arithmetic subtraction.
https://photon.codes/blog/we-found-a-ticking-time-bomb-in-macos-tcp-networking
#TCP #MacOS
0
0
0
You've seen all posts