CVE-2021-46936

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> net: fix use-after-free in tw_timer_handler<br /> <br /> A real world panic issue was found as follow in Linux 5.4.<br /> <br /> BUG: unable to handle page fault for address: ffffde49a863de28<br /> PGD 7e6fe62067 P4D 7e6fe62067 PUD 7e6fe63067 PMD f51e064067 PTE 0<br /> RIP: 0010:tw_timer_handler+0x20/0x40<br /> Call Trace:<br /> <br /> call_timer_fn+0x2b/0x120<br /> run_timer_softirq+0x1ef/0x450<br /> __do_softirq+0x10d/0x2b8<br /> irq_exit+0xc7/0xd0<br /> smp_apic_timer_interrupt+0x68/0x120<br /> apic_timer_interrupt+0xf/0x20<br /> <br /> This issue was also reported since 2017 in the thread [1],<br /> unfortunately, the issue was still can be reproduced after fixing<br /> DCCP.<br /> <br /> The ipv4_mib_exit_net is called before tcp_sk_exit_batch when a net<br /> namespace is destroyed since tcp_sk_ops is registered befrore<br /> ipv4_mib_ops, which means tcp_sk_ops is in the front of ipv4_mib_ops<br /> in the list of pernet_list. There will be a use-after-free on<br /> net-&gt;mib.net_statistics in tw_timer_handler after ipv4_mib_exit_net<br /> if there are some inflight time-wait timers.<br /> <br /> This bug is not introduced by commit f2bf415cfed7 ("mib: add net to<br /> NET_ADD_STATS_BH") since the net_statistics is a global variable<br /> instead of dynamic allocation and freeing. Actually, commit<br /> 61a7e26028b9 ("mib: put net statistics on struct net") introduces<br /> the bug since it put net statistics on struct net and free it when<br /> net namespace is destroyed.<br /> <br /> Moving init_ipv4_mibs() to the front of tcp_init() to fix this bug<br /> and replace pr_crit() with panic() since continuing is meaningless<br /> when init_ipv4_mibs() fails.<br /> <br /> [1] https://groups.google.com/g/syzkaller/c/p1tn-_Kc6l4/m/smuL_FMAAgAJ?pli=1