Re: [PATCH v2 14/32] liveupdate: luo_files: add infrastructure for FDs

[Pasha Tatashin <pasha.tatashin@soleen.com>]

> > +struct liveupdate_file_ops {
> > +     int (*prepare)(struct file *file, void *arg, u64 *data);
> > +     int (*freeze)(struct file *file, void *arg, u64 *data);
> > +     void (*cancel)(struct file *file, void *arg, u64 data);
> > +     void (*finish)(struct file *file, void *arg, u64 data, bool reclaimed);
> > +     int (*retrieve)(void *arg, u64 data, struct file **file);
> > +     bool (*can_preserve)(struct file *file, void *arg);
> > +};
>
> ops structures often have an owner = THIS_MODULE

Added here, and to subsystems.

>
> It wouldn't hurt to add it here too, and some appropriate module_get's
> though I didn't try to figure what happens if userspace races a module
> unload with other luo operations.

I added try_module_get()/module_put() to register/unregister functions.

> > +
> > +/**
> > + * struct liveupdate_file_handler - Represents a handler for a live-updatable
> > + * file type.
> > + * @ops:           Callback functions
> > + * @compatible:    The compatibility string (e.g., "memfd-v1", "vfiofd-v1")
> > + *                 that uniquely identifies the file type this handler supports.
> > + *                 This is matched against the compatible string associated with
> > + *                 individual &struct liveupdate_file instances.
> > + * @arg:           An opaque pointer to implementation-specific context data
> > + *                 associated with this file handler registration.
>
> Why? This is not the normal way, if you want context data then
> allocate a struct driver_liveupdate_file_handler and embed a normal
> struct liveupdate_file_handler inside it, then use container_of.

Good point. I removed arg, and added handler as an argument to the
callback functions.

> > +     fdt_for_each_subnode(file_node_offset, luo_file_fdt_in, 0) {
> > +             bool handler_found = false;
> > +             u64 token;
> > +
> > +             node_name = fdt_get_name(luo_file_fdt_in, file_node_offset,
> > +                                      NULL);
> > +             if (!node_name) {
> > +                     panic("FDT subnode at offset %d: Cannot get name\n",
> > +                           file_node_offset);
>
> I think this approach will raise lots of questions..
>
> I'd introduce a new function "luo_deserialize_failure" that does panic
> internally.
>
> Only called by places that are parsing the FDT & related but run into
> trouble that cannot be savely recovered from.

Agreed. I added a new macro in luo_internal.h:

 11 /*
 12  * Handles a deserialization failure: devices and memory is in
unpredictable
 13  * state.
 14  *
 15  * Continuing the boot process after a failure is dangerous
because it could
 16  * lead to leaks of private data.
 17  */
 18 #define luo_restore_fail(__fmt, ...) panic(__fmt, ##__VA_ARGS__)

And use it in places where we panic during deserialization.

Pasha