netdev - Re: [RFC net-next 5/5] eventpoll: Control irq suspension for prefer_busy

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <ZrpuodWa6cKh0sPk@mini-arch>
Date: Mon, 12 Aug 2024 13:20:49 -0700
From: Stanislav Fomichev <sdf@...ichev.me>
To: Joe Damato <jdamato@...tly.com>
Cc: netdev@...r.kernel.org, mkarsten@...terloo.ca,
	amritha.nambiar@...el.com, sridhar.samudrala@...el.com,
	Alexander Viro <viro@...iv.linux.org.uk>,
	Christian Brauner <brauner@...nel.org>, Jan Kara <jack@...e.cz>,
	"open list:FILESYSTEMS (VFS and infrastructure)" <linux-fsdevel@...r.kernel.org>,
	open list <linux-kernel@...r.kernel.org>
Subject: Re: [RFC net-next 5/5] eventpoll: Control irq suspension for
 prefer_busy_poll

On 08/12, Joe Damato wrote:
> From: Martin Karsten <mkarsten@...terloo.ca>
> 
> When events are reported to userland and prefer_busy_poll is set, irqs are
> temporarily suspended using napi_suspend_irqs.
> 
> If no events are found and ep_poll would go to sleep, irq suspension is
> cancelled using napi_resume_irqs.
> 
> Signed-off-by: Martin Karsten <mkarsten@...terloo.ca>
> Co-developed-by: Joe Damato <jdamato@...tly.com>
> Signed-off-by: Joe Damato <jdamato@...tly.com>
> Tested-by: Joe Damato <jdamato@...tly.com>
> Tested-by: Martin Karsten <mkarsten@...terloo.ca>
> ---
>  fs/eventpoll.c | 22 +++++++++++++++++++++-
>  1 file changed, 21 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/eventpoll.c b/fs/eventpoll.c
> index cc47f72005ed..d74b5b9c1f51 100644
> --- a/fs/eventpoll.c
> +++ b/fs/eventpoll.c
> @@ -457,6 +457,8 @@ static bool ep_busy_loop(struct eventpoll *ep, int nonblock)
>  		 * it back in when we have moved a socket with a valid NAPI
>  		 * ID onto the ready list.
>  		 */
> +		if (prefer_busy_poll)
> +			napi_resume_irqs(napi_id);
>  		ep->napi_id = 0;
>  		return false;
>  	}
> @@ -540,6 +542,14 @@ static long ep_eventpoll_bp_ioctl(struct file *file, unsigned int cmd,
>  	}
>  }
>  
> +static void ep_suspend_napi_irqs(struct eventpoll *ep)
> +{
> +	unsigned int napi_id = READ_ONCE(ep->napi_id);
> +
> +	if (napi_id >= MIN_NAPI_ID && READ_ONCE(ep->prefer_busy_poll))
> +		napi_suspend_irqs(napi_id);
> +}
> +
>  #else
>  
>  static inline bool ep_busy_loop(struct eventpoll *ep, int nonblock)
> @@ -557,6 +567,10 @@ static long ep_eventpoll_bp_ioctl(struct file *file, unsigned int cmd,
>  	return -EOPNOTSUPP;
>  }
>  
> +static void ep_suspend_napi_irqs(struct eventpoll *ep)
> +{
> +}
> +
>  #endif /* CONFIG_NET_RX_BUSY_POLL */
>  
>  /*
> @@ -788,6 +802,10 @@ static bool ep_refcount_dec_and_test(struct eventpoll *ep)
>  
>  static void ep_free(struct eventpoll *ep)
>  {
> +	unsigned int napi_id = READ_ONCE(ep->napi_id);
> +
> +	if (napi_id >= MIN_NAPI_ID && READ_ONCE(ep->prefer_busy_poll))
> +		napi_resume_irqs(napi_id);
>  	mutex_destroy(&ep->mtx);
>  	free_uid(ep->user);
>  	wakeup_source_unregister(ep->ws);
> @@ -2005,8 +2023,10 @@ static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events,
>  			 * trying again in search of more luck.
>  			 */
>  			res = ep_send_events(ep, events, maxevents);
> -			if (res)
> +			if (res) {
> +				ep_suspend_napi_irqs(ep);

Aren't we already doing defer in the busy_poll_stop? (or in napi_poll
when it's complete/done). Why do we need another rearming here?