On Mon, Oct 17, 2022 at 8:39 PM Zhengchao Shao <shaozhengc...@huawei.com> wrote:
>
> When the default qdisc is fq_codel, if the qdisc of dev_queue fails to be
> inited during mqprio_init(), fq_codel_reset() is invoked to clear
> resources. In this case, the flow is NULL, and it will cause gpf issue.
>
> The process is as follows:
> qdisc_create_dflt()
>         fq_codel_init()
>                 ...
>                 q->flows_cnt = 1024;
>                 ...
>                 q->flows = kvcalloc(...)      --->failed, q->flows is NULL
>                 ...
>         ...
>         qdisc_put()
>                 ...
>                 fq_codel_reset()
>                         ...
>                         flow = q->flows + i   --->q->flows is NULL
>
> The following is the Call Trace information:
> general protection fault, probably for non-canonical address
> 0xdffffc0000000001: 0000 [#1] PREEMPT SMP KASAN
> KASAN: null-ptr-deref in range [0x0000000000000008-0x000000000000000f]
> RIP: 0010:fq_codel_reset+0x14d/0x350
> Call Trace:
> <TASK>
> qdisc_reset+0xed/0x6f0
> qdisc_destroy+0x82/0x4c0
> qdisc_put+0x9e/0xb0
> qdisc_create_dflt+0x2c3/0x4a0
> mqprio_init+0xa71/0x1760
> qdisc_create+0x3eb/0x1000
> tc_modify_qdisc+0x408/0x1720
> rtnetlink_rcv_msg+0x38e/0xac0
> netlink_rcv_skb+0x12d/0x3a0
> netlink_unicast+0x4a2/0x740
> netlink_sendmsg+0x826/0xcc0
> sock_sendmsg+0xc5/0x100
> ____sys_sendmsg+0x583/0x690
> ___sys_sendmsg+0xe8/0x160
> __sys_sendmsg+0xbf/0x160
> do_syscall_64+0x35/0x80
> entry_SYSCALL_64_after_hwframe+0x46/0xb0
> RIP: 0033:0x7fd272b22d04
> </TASK>
>
> Fixes: 494f5063b86c ("net: sched: fq_codel: remove redundant resource cleanup 
> in fq_codel_init()")
> Signed-off-by: Zhengchao Shao <shaozhengc...@huawei.com>
> ---

I vote for a revert, previous code was much cleaner.
_______________________________________________
Cake mailing list
Cake@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cake

Reply via email to