linux-kernel - Re: [RFC V1 10/13] vdpa_sim: flush workers on suspend

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]

Date: Wed, 17 Jan 2024 15:31:06 -0500
From: Steven Sistare <steven.sistare@...cle.com>
To: Eugenio Perez Martin <eperezma@...hat.com>
Cc: virtualization@...ts.linux-foundation.org, linux-kernel@...r.kernel.org,
        "Michael S. Tsirkin" <mst@...hat.com>,
        Jason Wang <jasowang@...hat.com>, Si-Wei Liu <si-wei.liu@...cle.com>,
        Xuan Zhuo <xuanzhuo@...ux.alibaba.com>,
        Dragos Tatulea <dtatulea@...dia.com>, Eli Cohen <elic@...dia.com>,
        Xie Yongji <xieyongji@...edance.com>
Subject: Re: [RFC V1 10/13] vdpa_sim: flush workers on suspend

On 1/16/2024 1:57 PM, Eugenio Perez Martin wrote:
> On Wed, Jan 10, 2024 at 9:40 PM Steve Sistare <steven.sistare@...cle.com> wrote:
>>
>> To pass ownership of a live vdpa device to a new process, the user
>> suspends the device, calls VHOST_NEW_OWNER to change the mm, and calls
>> VHOST_IOTLB_REMAP to change the user virtual addresses to match the new
>> mm.  Flush workers in suspend to guarantee that no worker sees the new
>> mm and old VA in between.
> 
> The worker should already be stopped by the end of the suspend ioctl,
> so maybe we can consider this a fix?

Do you mean: the current behavior is a bug, independently of my new use case,
so I should submit this patch as a separate bug fix?  If yes, then will do.

>> Signed-off-by: Steve Sistare <steven.sistare@...cle.com>
>> ---
>>  drivers/vdpa/vdpa_sim/vdpa_sim.c | 13 +++++++++++++
>>  1 file changed, 13 insertions(+)
>>
>> diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim.c b/drivers/vdpa/vdpa_sim/vdpa_sim.c
>> index 6304cb0b4770..8734834983cb 100644
>> --- a/drivers/vdpa/vdpa_sim/vdpa_sim.c
>> +++ b/drivers/vdpa/vdpa_sim/vdpa_sim.c
>> @@ -74,6 +74,17 @@ static void vdpasim_worker_change_mm_sync(struct vdpasim *vdpasim,
>>         kthread_flush_work(work);
>>  }
>>
>> +static void flush_work_fn(struct kthread_work *work) {}
>> +
>> +static void vdpasim_flush_work(struct vdpasim *vdpasim)
>> +{
>> +       struct kthread_work work;
>> +
>> +       kthread_init_work(&work, flush_work_fn);
>> +       kthread_queue_work(vdpasim->worker, &work);
>> +       kthread_flush_work(&work);
> 
> Wouldn't it be better to cancel the work with kthread_cancel_work_sync here?

I believe that does not guarantee that currently executing work completes:

  static bool __kthread_cancel_work_sync()
    if (worker->current_work != work)
        goto out_fast;

- Steve

>> +}
>> +
>>  static struct vdpasim *vdpa_to_sim(struct vdpa_device *vdpa)
>>  {
>>         return container_of(vdpa, struct vdpasim, vdpa);
>> @@ -512,6 +523,8 @@ static int vdpasim_suspend(struct vdpa_device *vdpa)
>>         vdpasim->running = false;
>>         mutex_unlock(&vdpasim->mutex);
>>
>> +       vdpasim_flush_work(vdpasim);
>> +
>>         return 0;
>>  }
>>
>> --
>> 2.39.3
>>
>