Re: [PATCH net] tcp: ensure proper barriers in lockless contexts

2015-11-16 Thread David Miller
From: Paul Gortmaker Date: Mon, 16 Nov 2015 20:45:25 -0500 > Just a heads up that this breaks all arm64 builds in linux-next from > Monday; bisect says: > > 00fd38d938db3f1ab1c486549afc450cb7e751b1 is the first bad commit > commit 00fd38d938db3f1ab1c486549afc450cb7e751b1 > Author: Eric Dumazet >

Re: [PATCH net] tcp: ensure proper barriers in lockless contexts

2015-11-16 Thread Paul Gortmaker
On Sun, Nov 15, 2015 at 6:37 PM, David Miller wrote: > From: Eric Dumazet > Date: Thu, 12 Nov 2015 08:43:18 -0800 > >> From: Eric Dumazet >> >> Some functions access TCP sockets without holding a lock and >> might output non consistent data, depending on compiler and or >> architecture. >> >> tc

Re: [PATCH net] tcp: ensure proper barriers in lockless contexts

2015-11-15 Thread David Miller
From: Eric Dumazet Date: Thu, 12 Nov 2015 08:43:18 -0800 > From: Eric Dumazet > > Some functions access TCP sockets without holding a lock and > might output non consistent data, depending on compiler and or > architecture. > > tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() .

Re: [PATCH net] tcp: ensure proper barriers in lockless contexts

2015-11-14 Thread Herbert Xu
On Sat, Nov 14, 2015 at 07:21:41PM -0800, Eric Dumazet wrote: > > This patch makes sure sk_ack_backlog is written before sk_state is set > to TCP_LISTEN, otherwise one could see a '0' listener backlog. > > We had a spurious kernel log for a similar issue that was solved in > commit f985c65c908f6b2

Re: [PATCH net] tcp: ensure proper barriers in lockless contexts

2015-11-14 Thread Eric Dumazet
On Sun, 2015-11-15 at 09:20 +0800, Herbert Xu wrote: > Eric Dumazet wrote: > > From: Eric Dumazet > > > > Some functions access TCP sockets without holding a lock and > > might output non consistent data, depending on compiler and or > > architecture. > > > > tcp_diag_get_info(), tcp_get_info()

Re: [PATCH net] tcp: ensure proper barriers in lockless contexts

2015-11-14 Thread Herbert Xu
Eric Dumazet wrote: > From: Eric Dumazet > > Some functions access TCP sockets without holding a lock and > might output non consistent data, depending on compiler and or > architecture. > > tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ... For the information gathering ones

[PATCH net] tcp: ensure proper barriers in lockless contexts

2015-11-12 Thread Eric Dumazet
From: Eric Dumazet Some functions access TCP sockets without holding a lock and might output non consistent data, depending on compiler and or architecture. tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ... Introduce sk_state_load() and sk_state_store() to fix the issues, and