net: sleeping function called from invalid context in net_enable_timestamp · Dmitry Vyukov <dvyukov@google.com> · 2017-03-01
Re: net: sleeping function called from invalid context in net_enable_timestamp · Eric Dumazet <edumazet@google.com> · 2017-03-01
Re: net: sleeping function called from invalid context in net_enable_timestamp · Eric Dumazet <hidden> · 2017-03-01
Re: net: sleeping function called from invalid context in net_enable_timestamp · Eric Dumazet <hidden> · 2017-03-01
[PATCH net] net: net_enable_timestamp() can be called from irq contexts · Eric Dumazet <hidden> · 2017-03-02
Re: [PATCH net] net: net_enable_timestamp() can be called from irq contexts · David Miller <davem@davemloft.net> · 2017-03-02

Re: net: sleeping function called from invalid context in net_enable_timestamp

From: Eric Dumazet <hidden>
Date: 2017-03-01 21:51:53
Also in: lkml

On Wed, 2017-03-01 at 12:07 -0800, Eric Dumazet wrote:

On Wed, 2017-03-01 at 11:59 -0800, Eric Dumazet wrote:

quoted

On Wed, Mar 1, 2017 at 11:51 AM, Dmitry Vyukov [off-list ref] wrote:

quoted

Hello,

I've got the following report while running syzkaller fuzzer on
e5d56efc97f8240d0b5d66c03949382b6d7e5570



Right, a listener is playing fool games.

We need to use a work queue for all net_enable_timestamp() invocations

Something like :

We need something better, I will send a patch keeping good performance
for this jump label thing.