Re: [GIT] Networking

[Sowmini Varadhan]

On (07/09/17 11:49), Linus Torvalds wrote:
> 
> On Sat, Jul 8, 2017 at 3:36 AM, David Miller <davem@xxxxxxxxxxxxx> wrote:
> >
> > 8) Fix socket leak on accept() in RDS, from Sowmini Varadhan.  Also
> >    add a WARN_ON() to sock_graft() so other protocol stacks don't trip
> >    over this as well.
> 
> Hmm. This one triggers for me on both my desktop and laptop at bootup.
> Bog-standard machines, running F25 and F24 respectively.
> 
> The warning doesn't seem particularly useful, although maybe that
> "alg_accept()" gives people who know this code enough of a clue.

My initial question was whether sock_graft() should do a sock_put()
before cutting loose the existing parent->sk and assigning a new
parent->sk (https://www.spinics.net/lists/netdev/msg442191.html)

It look like PF_ALG sets up a ->sk in alg_create() (but this
would get over-written in alg_accept()?) 

Cc'ing Herbert to see if this is expected behavior (and PF_ALG
somehow does the right thing with the refcount for the ->sk
set up in alg_create) in which case I suppose we should drop the 
WARN_ON. 

--Sowmini

> ------------[ cut here ]------------
> WARNING: CPU: 1 PID: 492 at ./include/net/sock.h:1700 af_alg_accept+0x1bf/0x1f0
> CPU: 1 PID: 492 Comm: systemd-cryptse Not tainted 4.12.0-09010-g2b976203417c #1
> Hardware name: System manufacturer System Product Name/Z170-K, BIOS
> 1803 05/06/2016
> RIP: 0010:af_alg_accept+0x1bf/0x1f0
> Call Trace:
>  alg_accept+0x15/0x20
>  SYSC_accept4+0x105/0x210
>  ? getnstimeofday64+0xe/0x20
>  ? __audit_syscall_entry+0xb1/0xf0
>  ? syscall_trace_enter+0x1bd/0x2d0
>  ? __audit_syscall_exit+0x1a5/0x2a0
>  SyS_accept+0x10/0x20
>  do_syscall_64+0x61/0x140
>  entry_SYSCALL64_slow_path+0x25/0x25
> ---[ end trace a35e5baea85df269 ]---