[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <a53bfd1b-c54d-49de-96fb-b687e6e97533@linux.ibm.com>
Date: Wed, 6 Mar 2024 11:36:57 +0100
From: Jan Karcher <jaka@...ux.ibm.com>
To: Dmitry Antipov <dmantipov@...dex.ru>, Wen Gu <guwen@...ux.alibaba.com>
Cc: Wenjia Zhang <wenjia@...ux.ibm.com>,
"D. Wythe"
<alibuda@...ux.alibaba.com>, linux-s390@...r.kernel.org,
netdev@...r.kernel.org, lvc-project@...uxtesting.org
Subject: Re: Reaching official SMC maintainers
On 05/03/2024 17:39, Dmitry Antipov wrote:
> On 3/4/24 13:51, Wen Gu wrote:
>
>> IMHO, if we want to address the problem of fasync_struct entries being
>> incorrectly inserted to old socket, we may have to change the general
>> code.
>
> BTW what about using shared wait queue? Just to illustrate an idea:
I'm sorry but could we please clean up the e-mail threads?
This one here is a question if we are still alive: Yes, we are.
The other one i currently treat as an RFC and gracefully ignore the
PATCH tag. If you want to post it as an patch please come up with a
solution, clean it up and re-post it.
See patchwork errors for example:
https://patchwork.kernel.org/project/netdevbpf/patch/20240221051608.43241-1-dmantipov@yandex.ru/
For the general RFC discussion we are going to comment on it as soon as
we have something to say about it.
Feel free to re-post your idea regarding a shared wait queue there.
Thank you for your interest in smc and the ideas!
- Jan
>
> diff --git a/include/linux/net.h b/include/linux/net.h
> index c9b4a63791a4..02df64747db7 100644
> --- a/include/linux/net.h
> +++ b/include/linux/net.h
> @@ -126,6 +126,7 @@ struct socket {
> const struct proto_ops *ops; /* Might change with IPV6_ADDRFORM
> or MPTCP. */
>
> struct socket_wq wq;
> + struct socket_wq *shared_wq;
> };
>
> /*
> diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
> index 0f53a5c6fd9d..f04d61e316b2 100644
> --- a/net/smc/af_smc.c
> +++ b/net/smc/af_smc.c
> @@ -3360,6 +3360,9 @@ static int __smc_create(struct net *net, struct
> socket *sock, int protocol,
> smc->clcsock = clcsock;
> }
>
> + sock->shared_wq = &smc->shared_wq;
> + smc->clcsock->shared_wq = &smc->shared_wq;
> +
> out:
> return rc;
> }
> diff --git a/net/smc/smc.h b/net/smc/smc.h
> index df64efd2dee8..26e66c289d4f 100644
> --- a/net/smc/smc.h
> +++ b/net/smc/smc.h
> @@ -287,6 +287,7 @@ struct smc_sock { /* smc sock
> container */
> /* protects clcsock of a listen
> * socket
> * */
> + struct socket_wq shared_wq;
> };
>
> #define smc_sk(ptr) container_of_const(ptr, struct smc_sock, sk)
> diff --git a/net/socket.c b/net/socket.c
> index ed3df2f749bf..9b9e6932906f 100644
> --- a/net/socket.c
> +++ b/net/socket.c
> @@ -1437,7 +1437,8 @@ static int sock_fasync(int fd, struct file *filp,
> int on)
> {
> struct socket *sock = filp->private_data;
> struct sock *sk = sock->sk;
> - struct socket_wq *wq = &sock->wq;
> + struct socket_wq *wq = (unlikely(sock->shared_wq) ?
> + sock->shared_wq : &sock->wq);
>
> if (sk == NULL)
> return -EINVAL;
>
> Dmitry
>
Powered by blists - more mailing lists