Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH V3 7/9] fork: Add worker flag to ignore signals
To:     Mike Christie <michael.christie@oracle.com>, geert@linux-m68k.org,
        vverma@digitalocean.com, hdanton@sina.com, hch@infradead.org,
        stefanha@redhat.com, jasowang@redhat.com, mst@redhat.com,
        sgarzare@redhat.com, virtualization@lists.linux-foundation.org,
        christian.brauner@ubuntu.com, linux-kernel@vger.kernel.org
References: <20211004192128.381453-1-michael.christie@oracle.com>
 <20211004192128.381453-8-michael.christie@oracle.com>
From:   Jens Axboe <axboe@kernel.dk>
Message-ID: <e3840acc-09c0-01bd-ad64-54e9d2dfb888@kernel.dk>
Date:   Mon, 4 Oct 2021 14:04:12 -0600
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.10.0
MIME-Version: 1.0
In-Reply-To: <20211004192128.381453-8-michael.christie@oracle.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 10/4/21 1:21 PM, Mike Christie wrote:
> The kthread API creates threads that ignore all signals by default so
> modules like vhost that will move from that API to kernel_worker will
> not be expecting them. This patch adds a worker flag that tells
> kernel_worker to setup the task to ignore signals.
> 
> Signed-off-by: Mike Christie <michael.christie@oracle.com>
> Acked-by: Christian Brauner <christian.brauner@ubuntu.com>
> ---
>  include/linux/sched/task.h |  1 +
>  kernel/fork.c              | 11 ++++++++++-
>  2 files changed, 11 insertions(+), 1 deletion(-)
> 
> diff --git a/include/linux/sched/task.h b/include/linux/sched/task.h
> index 781abbc1c288..aefa0d221b57 100644
> --- a/include/linux/sched/task.h
> +++ b/include/linux/sched/task.h
> @@ -21,6 +21,7 @@ struct css_set;
>  #define KERN_WORKER_IO		BIT(0)
>  #define KERN_WORKER_USER	BIT(1)
>  #define KERN_WORKER_NO_FILES	BIT(2)
> +#define KERN_WORKER_NO_SIGS	BIT(3)
>  
>  struct kernel_clone_args {
>  	u64 flags;
> diff --git a/kernel/fork.c b/kernel/fork.c
> index 3f3fcabffa5f..34d3dca70cfb 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -2555,6 +2555,8 @@ struct task_struct *create_io_thread(int (*fn)(void *), void *arg, int node)
>  struct task_struct *kernel_worker(int (*fn)(void *), void *arg, int node,
>  				  unsigned long clone_flags, u32 worker_flags)
>  {
> +	struct task_struct *tsk;
> +
>  	struct kernel_clone_args args = {
>  		.flags		= ((lower_32_bits(clone_flags) | CLONE_VM |
>  				   CLONE_UNTRACED) & ~CSIGNAL),
> @@ -2564,7 +2566,14 @@ struct task_struct *kernel_worker(int (*fn)(void *), void *arg, int node,
>  		.worker_flags	= KERN_WORKER_USER | worker_flags,
>  	};
>  
> -	return copy_process(NULL, 0, node, &args);
> +	tsk = copy_process(NULL, 0, node, &args);
> +	if (IS_ERR(tsk))
> +		return tsk;
> +
> +	if (worker_flags & KERN_WORKER_NO_SIGS)
> +		ignore_signals(tsk);
> +
> +	return tsk;

When I originally did it this way, Eric (correctly) pointed out that
it's racy. See where it's currently done as part of copy_process(), not
after.

-- 
Jens Axboe