From mboxrd@z Thu Jan 1 00:00:00 1970 From: Subash Abhinov Kasiviswanathan Subject: Crash due to NULL dereference in tcp_rearm_rto Date: Fri, 13 Apr 2018 18:39:19 -0600 Message-ID: <0e7dbdef0ecf0ee0562a144050165453@codeaurora.org> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII; format=flowed Content-Transfer-Encoding: 7bit To: netdev@vger.kernel.org, soheil@google.com, eric.dumazet@gmail.com Return-path: Received: from smtp.codeaurora.org ([198.145.29.96]:40702 "EHLO smtp.codeaurora.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750839AbeDNAjU (ORCPT ); Fri, 13 Apr 2018 20:39:20 -0400 Sender: netdev-owner@vger.kernel.org List-ID: We are seeing a warning followed by a crash on an ARM64 device with Android 4.14 based kernel. It looks like both sk->sk_write_queue and sk->sk_send_head are NULL. Since the sk->sk_write_queue is NULL and is dereferenced in tcp_rto_delta_us() to get the skb->skb_mstamp, there is crash observed. Since this is 4.14.32, it already has commit ("tcp: reset sk_send_head in tcp_write_queue_purge") https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?h=v4.14.34&id=dbbf2d1e4077bab0c65ece2765d3fc69cf7d610f 12876.013077: <6> WARNING: CPU: 5 PID: 14828 at net/ipv4/tcp_output.c:2469 tcp_send_loss_probe+0x198/0x1b8 12876.038939: <6> task: ffffffe73f7e5a80 task.stack: ffffff801b068000 12876.038941: <2> PC is at tcp_send_loss_probe+0x198/0x1b8 12876.038942: <2> LR is at tcp_send_loss_probe+0x28/0x1b8 12876.038944: <2> pc : [] lr : [] pstate: 60400145 12876.038944: <2> sp : ffffff800802bd30 12876.038945: <2> x29: ffffff800802bd50 x28: ffffff8dc9d83eb8 12876.038948: <2> x27: ffffff800802be08 x26: ffffff8dc9737000 12876.038950: <2> x25: 0000000000000001 x24: ffffffe744ea1728 12876.038952: <2> x23: ffffffe73f7e5a80 x22: 0000000000000558 12876.038954: <2> x21: 0000000000000000 x20: 0000000001080020 12876.038956: <2> x19: ffffffe73d06e440 x18: 0000000000000020 12876.038958: <2> x17: 0000000000000014 x16: 0000000000000030 12876.038960: <2> x15: 0000000000000000 x14: 0000000000000000 12876.038962: <2> x13: 0000000013af314c x12: 0000002773f8a550 12876.038965: <2> x11: 0000000000000538 x10: 0000000000000000 12876.038967: <2> x9 : 0000000000000020 x8 : ffffffe73d06e5f0 12876.038969: <2> x7 : 0000000000924278 x6 : ffffffe76fe9ed80 12876.038971: <2> x5 : ffffffe76fe9ed80 x4 : 000000000f500458 12876.038973: <2> x3 : ffffff800802bce0 x2 : ffffff800802bce8 12876.038975: <2> x1 : 0000000000000000 x0 : 0000000000000558 12876.039082: <2> [] tcp_send_loss_probe+0x198/0x1b8 12876.039084: <2> [] tcp_write_timer_handler+0xf8/0x1c4 12876.039086: <2> [] tcp_write_timer+0x5c/0x98 12876.039089: <2> [] call_timer_fn+0xc0/0x1b4 12876.039091: <2> [] run_timer_softirq+0x230/0x850 12876.039094: <2> [] __do_softirq+0x1dc/0x3a4 12876.039096: <2> [] irq_exit+0xc8/0xd4 12876.039098: <2> [] __handle_domain_irq+0x8c/0xc4 12876.039099: <2> [] gic_handle_irq+0x164/0x1bc [net/ipv4/tcp_output.c] void tcp_send_loss_probe(struct sock *sk) { struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb; int pcount; int mss = tcp_current_mss(sk); ... /* Retransmit last segment. */ if (WARN_ON(!skb)) goto rearm_timer; 12876.043967: <6> Unable to handle kernel NULL pointer dereference at virtual address 00000010 12876.091600: <6> Internal error: Oops: 96000005 [#1] PREEMPT SMP 12876.152597: <2> PC is at tcp_rearm_rto+0x48/0x90 12876.156979: <2> LR is at tcp_send_loss_probe+0x178/0x1b8 12876.162077: <2> pc : [] lr : [] pstate: 60400145 12876.169657: <2> sp : ffffff800802bd10 12876.173056: <2> x29: ffffff800802bd20 x28: ffffff8dc9d83eb8 12876.178511: <2> x27: ffffff800802be08 x26: ffffff8dc9737000 12876.183967: <2> x25: 0000000000000001 x24: ffffffe744ea1728 12876.189418: <2> x23: ffffffe73f7e5a80 x22: 0000000000000558 12876.194863: <2> x21: 0000000000000000 x20: 0000000001080020 12876.200312: <2> x19: ffffffe73d06e440 x18: 0000000000000020 12876.205758: <2> x17: 0000000000000014 x16: 0000000000000030 12876.211212: <2> x15: 0000000000000000 x14: 0000000000000000 12876.216660: <2> x13: 0000000013af314c x12: 0000002773f8a550 12876.222108: <2> x11: 0000000000000538 x10: 0000000000000000 12876.227561: <2> x9 : 0000000000000000 x8 : ffffffe73d06e5f0 12876.233008: <2> x7 : 0000000000924278 x6 : ffffffe76fe9ed80 12876.238455: <2> x5 : ffffffe76fe9ed80 x4 : 000000000f500458 12876.243907: <2> x3 : ffffff800802bce0 x2 : ffffff800802bce8 12876.249360: <2> x1 : 0000000000000000 x0 : 0000000000000867 12876.473522: <2> [] tcp_rearm_rto+0x48/0x90 12876.478971: <2> [] tcp_send_loss_probe+0x178/0x1b8 12876.485131: <2> [] tcp_write_timer_handler+0xf8/0x1c4 12876.491557: <2> [] tcp_write_timer+0x5c/0x98 12876.497189: <2> [] call_timer_fn+0xc0/0x1b4 12876.502731: <2> [] run_timer_softirq+0x230/0x850 12876.508716: <2> [] __do_softirq+0x1dc/0x3a4 12876.514260: <2> [] irq_exit+0xc8/0xd4 12876.519261: <2> [] __handle_domain_irq+0x8c/0xc4 12876.525245: <2> [] gic_handle_irq+0x164/0x1bc [net/ipv4/tcp_input.c] void tcp_rearm_rto(struct sock *sk) { ... inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS); } else { u32 rto = inet_csk(sk)->icsk_rto; /* Offset the time elapsed after installing regular RTO */ if (icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT || icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) { s64 delta_us = tcp_rto_delta_us(sk); /* delta_us may not be positive if the socket is locked * when the retrans timer fires and is rescheduled. */ rto = usecs_to_jiffies(max_t(int, delta_us, 1)); [include/net/tcp.h] static inline s64 tcp_rto_delta_us(const struct sock *sk) { const struct sk_buff *skb = tcp_write_queue_head(sk); u32 rto = inet_csk(sk)->icsk_rto; u64 rto_time_stamp_us = skb->skb_mstamp + jiffies_to_usecs(rto); return rto_time_stamp_us - tcp_sk(sk)->tcp_mstamp; } -- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project