Re: [PATCH v6 05/20] liveupdate: luo_ioctl: add user interface

[David Matlack]

On Thu, Nov 20, 2025 at 11:23 AM Pasha Tatashin
<pasha.tatashin@xxxxxxxxxx> wrote:
>
> On Thu, Nov 20, 2025 at 1:38 PM David Matlack <dmatlack@xxxxxxxxxx> wrote:
> >
> > On Sat, Nov 15, 2025 at 3:34 PM Pasha Tatashin
> > <pasha.tatashin@xxxxxxxxxx> wrote:
> > > The idea is that there is going to be a single userspace agent driving
> > > the live update, therefore, only a single process can ever hold this
> > > device opened at a time.
> > ...
> > > +static int luo_open(struct inode *inodep, struct file *filep)
> > > +{
> > > +       struct luo_device_state *ldev = container_of(filep->private_data,
> > > +                                                    struct luo_device_state,
> > > +                                                    miscdev);
> > > +
> > > +       if (atomic_cmpxchg(&ldev->in_use, 0, 1))
> > > +               return -EBUSY;
> >
> > Can you remind me why the kernel needs to enforce this? What would be
> > wrong or unsafe from the kernel perspective if there were multiple
> > userspace agents holding open files for /dev/liveupdate, each with
> > their own sessions?
>
> By enforcing a singleton, we will ensure a consistent view for tooling
> like luoadm (which will track incoming/outgoing sessions, UUIDs, etc.)
> and prevent conflicting commands regarding the transition state.
>
> This is not a bottleneck because the vast majority of the work
> (preserving devicse/memory) is handled via the individual Session FDs.
> Also, since sessions persist even if /dev/liveupdate is closed, we
> allow the agent upgrade, or crashing without requiring concurrent
> access.

Yeah, I'm not concerned about bottlenecking. It just seems like an
artificial constraint to impose on userspace at this point. The only
ioctls on /dev/liveupdate are to create a session and retreive a
session. Neither of those will conflict with having multiple open
files for /dev/liveupdate.