CVE-2024-41007

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> tcp: avoid too many retransmit packets<br /> <br /> If a TCP socket is using TCP_USER_TIMEOUT, and the other peer<br /> retracted its window to zero, tcp_retransmit_timer() can<br /> retransmit a packet every two jiffies (2 ms for HZ=1000),<br /> for about 4 minutes after TCP_USER_TIMEOUT has &amp;#39;expired&amp;#39;.<br /> <br /> The fix is to make sure tcp_rtx_probe0_timed_out() takes<br /> icsk-&gt;icsk_user_timeout into account.<br /> <br /> Before blamed commit, the socket would not timeout after<br /> icsk-&gt;icsk_user_timeout, but would use standard exponential<br /> backoff for the retransmits.<br /> <br /> Also worth noting that before commit e89688e3e978 ("net: tcp:<br /> fix unexcepted socket die when snd_wnd is 0"), the issue<br /> would last 2 minutes instead of 4.