Re: [PATCH 00/11] Use copy_process in vhost layer

[Mike Christie <michael.christie@oracle.com>]

On 3/11/23 11:21 AM, Linus Torvalds wrote:
> On Fri, Mar 10, 2023 at 2:04 PM Mike Christie
> <michael.christie@oracle.com> wrote:
>>
>> The following patches were made over Linus's tree and apply over next. They
>> allow the vhost layer to use copy_process instead of using
>> workqueue_structs to create worker threads for VM's devices.
> 
> Ok, all these patches looked fine to me from a quick scan - nothing
> that I reacted to as objectionable, and several of them looked like
> nice cleanups.
> 
> The only one I went "Why do you do it that way" for was in 10/11
> (entirely internal to vhost, so I don't feel too strongly about this)
> how you made "struct vhost_worker" be a pointer in "struct vhost_dev".
> 
> It _looks_ to me like it could just have been an embedded structure
> rather than a separate allocation.
> 
> IOW, why do
> 
>    vhost_dev->worker
> 
> instead of doing
> 
>   vhost_dev.worker
> 
> and just having it all in the same allocation?
> 
> Not a big deal. Maybe you wanted the 'test if worker pointer is NULL'
> code to stay around, and basically use that pointer as a flag too. Or
> maybe there is some other reason you want to keep that separate..
> 

There were 2 reasons:
1. Yeah, we needed a flag to indicate that the worker was not setup
for the cases like where userspace just opens the dev then closes it
without doing the IOCTL that does vhost_dev_set_owner.

2. I could have handled #1 by embedding the worker in the vhost_dev
and then just testing worker.vtsk. However, I have a followup patchset
that allows us to create multiple worker threads per device. For
that patchset I then do:

- if (vhost_dev->worker)

+ if (vhost_dev->workers)

so I think it just saved me some typing.