DRM-managed resources / devm_drm_dev_alloc leaking resources

  Hi,

After trying to get more kunit tests for KMS, I found out that the
recent kunit helpers we merged to create a DRM device [1] are broken and
won't free their device-managed and DRM-managed resources.

With some help from Thomas, we've dug into this and it turns out that if
we allocate a device with root_device_register, initialise our drm
device with devm_drm_dev_alloc(), register it using drm_dev_register(),
unregister it using drm_dev_unregister/drm_dev_unplug and then remove
the parent device, neither the device managed nor the DRM managed
actions are run.

root_device_register initializes the device by eventually calling
device_initialize() which sets the initial reference count of the root
device to 1 [2]. devm_drm_dev_alloc() then comes in, drm_dev_init() will
increase the root device refcount [3] and initialize our DRM device to 1
[4]. drm_dev_register(), through drm_minor_register() and device_add(),
will increase the root device refcount [5].

When unrolling things, drm_dev_unregister(), through
drm_minor_unregister() and device_del(), will give up its reference [6].
root_device_unregister(), through device_unregister(), will also give up
its own [7].

So we end up with this for the reference counts:

+------------------------+-------------+------------+
|         funcs          | root device | DRM device |
+------------------------+-------------+------------+
| root_device_register   |           1 | N/A        |
| devm_drm_dev_alloc     |           2 | 1          |
| drm_dev_register       |           3 | 1          |
| drm_dev_unregister     |           2 | 1          |
| root_device_unregister |           1 | 1          |
+------------------------+-------------+------------+

If we go back to the list of reference taken, the root device reference
and the initial drm_device reference, both taken by devm_drm_dev_alloc()
through drm_dev_init(), haven't been put back.

If we look at the drm_dev_init code(), we can see that it sets up a
DRM-managed action [8] that will put back the device reference [9]. The
DRM-managed code is executed by the drm_managed_cleanup() function, that
is executed as part of a release hook [10] executed once we give up the
final reference to the DRM device [11].

If we go back a little, the final reference to the DRM device is
actually the initial one setup by devm_drm_dev_alloc(). This function
has superseded drm_dev_alloc(), with the documentation that we do need a
final drm_dev_put() to put back our final reference [12].

devm_drm_dev_alloc() is a more convenient variant that has been
introduced explicitly to not require that drm_dev_put(), and states it
as such in the documentation [13]. It does so by adding a device-managed
action that will call drm_dev_put() [14].

Device-managed actions are ran as part devres_release_all() that is
called by device_release() [15], itself being run when the last
reference on the device is put back [16][17][18].

So if we sum things up, the DRM device will only give its last root
device reference when the last DRM device reference will be put back,
and the last DRM device reference will be put back when the last device
reference will be put back, which sounds very circular to me, with both
ending up in a deadlock scenario.

I've added two kunit tests that demonstrate the issue: we register a
device, allocate and register a DRM device, register a DRM-managed
action, remove the DRM device and the parent device, and wait for the
action to execute. drm_register_unregister_with_devm_test() uses the
broken(?) devm_drm_dev_alloc and is failing.
drm_register_unregister_test uses the deprecated drm_dev_alloc() that
requires an explicit call to drm_dev_put() which works fine.

It's also worth noting that Thomas tested with simpledrm and it seems to
work fine. Using a platform_device instead of the root_device doesn't
change anything to the outcome in my tests, so there might be a more
subtle behaviour involved.

Thanks,
Maxime

--------- 8< -----------
--------- 8< -----------

1: https://cgit.freedesktop.org/drm-misc/tree/drivers/gpu/drm/tests/drm_kunit_helpers.c
2: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L2979
3: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L597
4: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L596
5: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L3437
6: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L201
7: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L3737
8: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L618
9: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L570
10: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L751
11: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L785
12: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L259
13: https://elixir.bootlin.com/linux/latest/source/include/drm/drm_drv.h#L505
14: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/drm_drv.c#L682
15: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L2321
16: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L2357
17: https://elixir.bootlin.com/linux/latest/source/lib/kobject.c#L721
18: https://elixir.bootlin.com/linux/latest/source/lib/kobject.c#L647
  

Hello Maxime,

On 11/17/22 17:53, Maxime Ripard wrote:
> Hi,
> 
> After trying to get more kunit tests for KMS, I found out that the
> recent kunit helpers we merged to create a DRM device [1] are broken and
> won't free their device-managed and DRM-managed resources.
> 
> With some help from Thomas, we've dug into this and it turns out that if
> we allocate a device with root_device_register, initialise our drm
> device with devm_drm_dev_alloc(), register it using drm_dev_register(),
> unregister it using drm_dev_unregister/drm_dev_unplug and then remove
> the parent device, neither the device managed nor the DRM managed
> actions are run.
> 
> root_device_register initializes the device by eventually calling
> device_initialize() which sets the initial reference count of the root
> device to 1 [2]. devm_drm_dev_alloc() then comes in, drm_dev_init() will
> increase the root device refcount [3] and initialize our DRM device to 1
> [4]. drm_dev_register(), through drm_minor_register() and device_add(),
> will increase the root device refcount [5].
> 
> When unrolling things, drm_dev_unregister(), through
> drm_minor_unregister() and device_del(), will give up its reference [6].
> root_device_unregister(), through device_unregister(), will also give up
> its own [7].
> 
> So we end up with this for the reference counts:
> 
> +------------------------+-------------+------------+
> |         funcs          | root device | DRM device |
> +------------------------+-------------+------------+
> | root_device_register   |           1 | N/A        |
> | devm_drm_dev_alloc     |           2 | 1          |
> | drm_dev_register       |           3 | 1          |
> | drm_dev_unregister     |           2 | 1          |
> | root_device_unregister |           1 | 1          |
> +------------------------+-------------+------------+
> 
> If we go back to the list of reference taken, the root device reference
> and the initial drm_device reference, both taken by devm_drm_dev_alloc()
> through drm_dev_init(), haven't been put back.
> 
> If we look at the drm_dev_init code(), we can see that it sets up a
> DRM-managed action [8] that will put back the device reference [9]. The
> DRM-managed code is executed by the drm_managed_cleanup() function, that
> is executed as part of a release hook [10] executed once we give up the
> final reference to the DRM device [11].
> 
> If we go back a little, the final reference to the DRM device is
> actually the initial one setup by devm_drm_dev_alloc(). This function
> has superseded drm_dev_alloc(), with the documentation that we do need a
> final drm_dev_put() to put back our final reference [12].
> 
> devm_drm_dev_alloc() is a more convenient variant that has been
> introduced explicitly to not require that drm_dev_put(), and states it
> as such in the documentation [13]. It does so by adding a device-managed
> action that will call drm_dev_put() [14].
> 
> Device-managed actions are ran as part devres_release_all() that is
> called by device_release() [15], itself being run when the last
> reference on the device is put back [16][17][18].
>

Thanks a lot for the write up. It was very informative and detailed.

> So if we sum things up, the DRM device will only give its last root
> device reference when the last DRM device reference will be put back,
> and the last DRM device reference will be put back when the last device
> reference will be put back, which sounds very circular to me, with both
> ending up in a deadlock scenario.
>

So my conclusion after looking at this is the same than your, that the
drivers would need to signal somehow to the DRM core when a DRM device
won't be used anymore and drop the final reference to the DRM device.

That is, I don't think we can get away of drivers not calling either
drm_dev_put().

I think that we should try to simplify the DRM register and release API
and make very clear in the documentation what should be used. Right now
for example as you mentioned we have both drm_dev_unregister() and
drm_dev_unplug() but AFAICT the only difference is that the latter does
a sync to protect critical sections during drm_dev_{enter,exit}().

The drawback is that the DRM device will be marked as unplugged before
drm_atomic_helper_shutdown(), but is this really a problem in practice?

Maybe we can just rename drm_dev_unplug() to drm_dev_unregister() and
drm_dev_unregister() to __drm_dev_unregister(). That way, the register
path could always be:

   devm_drm_dev_alloc()
   drm_dev_register()

and then in the release path:

   drm_dev_unregister()
   drm_dev_put()

making both DRM-managed and device-managed resources to always work.

> I've added two kunit tests that demonstrate the issue: we register a
> device, allocate and register a DRM device, register a DRM-managed
> action, remove the DRM device and the parent device, and wait for the
> action to execute. drm_register_unregister_with_devm_test() uses the
> broken(?) devm_drm_dev_alloc and is failing.
> drm_register_unregister_test uses the deprecated drm_dev_alloc() that
> requires an explicit call to drm_dev_put() which works fine.
>

Great. In my opinion we should add these Kunit tests even when they are
exposing an issue in the devm_drm_dev_alloc() helper.
 
> It's also worth noting that Thomas tested with simpledrm and it seems to
> work fine. Using a platform_device instead of the root_device doesn't
> change anything to the outcome in my tests, so there might be a more
> subtle behaviour involved.
>

That's strange because AFAICT simpledrm is basically doing the same than
your failing tests. I tried to look at the differences but couldn't spot
anything evident...
 
> Thanks,
> Maxime
>
  

Hi,

Javier and I looked into it some more today, and I think we have a
better idea of what is going on.

On Thu, Nov 17, 2022 at 05:53:11PM +0100, Maxime Ripard wrote:
> After trying to get more kunit tests for KMS, I found out that the
> recent kunit helpers we merged to create a DRM device [1] are broken and
> won't free their device-managed and DRM-managed resources.
> 
> With some help from Thomas, we've dug into this and it turns out that if
> we allocate a device with root_device_register, initialise our drm
> device with devm_drm_dev_alloc(), register it using drm_dev_register(),
> unregister it using drm_dev_unregister/drm_dev_unplug and then remove
> the parent device, neither the device managed nor the DRM managed
> actions are run.
> 
> root_device_register initializes the device by eventually calling
> device_initialize() which sets the initial reference count of the root
> device to 1 [2]. devm_drm_dev_alloc() then comes in, drm_dev_init() will
> increase the root device refcount [3] and initialize our DRM device to 1
> [4]. drm_dev_register(), through drm_minor_register() and device_add(),
> will increase the root device refcount [5].
> 
> When unrolling things, drm_dev_unregister(), through
> drm_minor_unregister() and device_del(), will give up its reference [6].
> root_device_unregister(), through device_unregister(), will also give up
> its own [7].
> 
> So we end up with this for the reference counts:
> 
> +------------------------+-------------+------------+
> |         funcs          | root device | DRM device |
> +------------------------+-------------+------------+
> | root_device_register   |           1 | N/A        |
> | devm_drm_dev_alloc     |           2 | 1          |
> | drm_dev_register       |           3 | 1          |
> | drm_dev_unregister     |           2 | 1          |
> | root_device_unregister |           1 | 1          |
> +------------------------+-------------+------------+
> 
> If we go back to the list of reference taken, the root device reference
> and the initial drm_device reference, both taken by devm_drm_dev_alloc()
> through drm_dev_init(), haven't been put back.
> 
> If we look at the drm_dev_init code(), we can see that it sets up a
> DRM-managed action [8] that will put back the device reference [9]. The
> DRM-managed code is executed by the drm_managed_cleanup() function, that
> is executed as part of a release hook [10] executed once we give up the
> final reference to the DRM device [11].
> 
> If we go back a little, the final reference to the DRM device is
> actually the initial one setup by devm_drm_dev_alloc(). This function
> has superseded drm_dev_alloc(), with the documentation that we do need a
> final drm_dev_put() to put back our final reference [12].
> 
> devm_drm_dev_alloc() is a more convenient variant that has been
> introduced explicitly to not require that drm_dev_put(), and states it
> as such in the documentation [13]. It does so by adding a device-managed
> action that will call drm_dev_put() [14].
> 
> Device-managed actions are ran as part devres_release_all() that is
> called by device_release() [15], itself being run when the last
> reference on the device is put back [16][17][18].
> 
> So if we sum things up, the DRM device will only give its last root
> device reference when the last DRM device reference will be put back,
> and the last DRM device reference will be put back when the last device
> reference will be put back, which sounds very circular to me, with both
> ending up in a deadlock scenario.
> 
> I've added two kunit tests that demonstrate the issue: we register a
> device, allocate and register a DRM device, register a DRM-managed
> action, remove the DRM device and the parent device, and wait for the
> action to execute. drm_register_unregister_with_devm_test() uses the
> broken(?) devm_drm_dev_alloc and is failing.
> drm_register_unregister_test uses the deprecated drm_dev_alloc() that
> requires an explicit call to drm_dev_put() which works fine.
> 
> It's also worth noting that Thomas tested with simpledrm and it seems to
> work fine.

Indeed, the transition from simpledrm to a full-blown DRM driver handles
this properly. It's using a platform_device_unregister() [1] and
eventually device_del() [2][3]. That part is handled just like
root_device_unregister() [4][5]. Basically, both will call device_del(),
and then device_put(), so device_del() is called while holding a
reference.

As we've seen before, at this point the DRM driver still holds a
reference to the device as well.

device_del() will call bus_remove_device() [6], which will be skipped
for the root device since it doesn't have a bus [7].

It will then call device_release_driver() [8], which is basically forwarded
to __device_release_driver() [9][10], that will call device_unbind_cleanup() [11].

device_unbind_cleanup() calls devres_release_all() directly [12], that
runs all the device-managed actions [13]. And it does so WHILE THERE'S
STILL A REFCOUNT OF 2!

I would expect the call to devres_release_all to happen only in
device_release, once all the device reference have been put back. Not 4
functions in as a side effect, and while there's still some active
references.

> Using a platform_device instead of the root_device doesn't
> change anything to the outcome in my tests, so there might be a more
> subtle behaviour involved.

This one has the same symptom but a different cause. I was just
registering a platform_device but it wasn't bound to any driver. While
it's valid to do so according to that comment [13], it doesn't have any
driver so the check for the driver in device_release_driver() [8], and
never hits device_unbind_cleanup().

Thanks again to Thomas and Javier for their help
Maxime

1: https://elixir.bootlin.com/linux/latest/source/drivers/video/aperture.c#L199
2: https://elixir.bootlin.com/linux/latest/source/drivers/base/platform.c#L793
3: https://elixir.bootlin.com/linux/latest/source/drivers/base/platform.c#L751
4: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L4153
5: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L3733
6: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L3704
7: https://elixir.bootlin.com/linux/latest/source/drivers/base/bus.c#L511
8: https://elixir.bootlin.com/linux/latest/source/drivers/base/bus.c#L529
9: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L1298
10: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L1275
11: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L1255
12: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L530
13: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L2312
  

On Fri, Nov 18, 2022 at 05:15:58PM +0100, Maxime Ripard wrote:
> Hi,
> 
> Javier and I looked into it some more today, and I think we have a
> better idea of what is going on.
> 
> On Thu, Nov 17, 2022 at 05:53:11PM +0100, Maxime Ripard wrote:
> > After trying to get more kunit tests for KMS, I found out that the
> > recent kunit helpers we merged to create a DRM device [1] are broken and
> > won't free their device-managed and DRM-managed resources.
> > 
> > With some help from Thomas, we've dug into this and it turns out that if
> > we allocate a device with root_device_register, initialise our drm
> > device with devm_drm_dev_alloc(), register it using drm_dev_register(),
> > unregister it using drm_dev_unregister/drm_dev_unplug and then remove
> > the parent device, neither the device managed nor the DRM managed
> > actions are run.
> > 
> > root_device_register initializes the device by eventually calling
> > device_initialize() which sets the initial reference count of the root
> > device to 1 [2]. devm_drm_dev_alloc() then comes in, drm_dev_init() will
> > increase the root device refcount [3] and initialize our DRM device to 1
> > [4]. drm_dev_register(), through drm_minor_register() and device_add(),
> > will increase the root device refcount [5].
> > 
> > When unrolling things, drm_dev_unregister(), through
> > drm_minor_unregister() and device_del(), will give up its reference [6].
> > root_device_unregister(), through device_unregister(), will also give up
> > its own [7].
> > 
> > So we end up with this for the reference counts:
> > 
> > +------------------------+-------------+------------+
> > |         funcs          | root device | DRM device |
> > +------------------------+-------------+------------+
> > | root_device_register   |           1 | N/A        |
> > | devm_drm_dev_alloc     |           2 | 1          |
> > | drm_dev_register       |           3 | 1          |
> > | drm_dev_unregister     |           2 | 1          |
> > | root_device_unregister |           1 | 1          |
> > +------------------------+-------------+------------+
> > 
> > If we go back to the list of reference taken, the root device reference
> > and the initial drm_device reference, both taken by devm_drm_dev_alloc()
> > through drm_dev_init(), haven't been put back.
> > 
> > If we look at the drm_dev_init code(), we can see that it sets up a
> > DRM-managed action [8] that will put back the device reference [9]. The
> > DRM-managed code is executed by the drm_managed_cleanup() function, that
> > is executed as part of a release hook [10] executed once we give up the
> > final reference to the DRM device [11].
> > 
> > If we go back a little, the final reference to the DRM device is
> > actually the initial one setup by devm_drm_dev_alloc(). This function
> > has superseded drm_dev_alloc(), with the documentation that we do need a
> > final drm_dev_put() to put back our final reference [12].
> > 
> > devm_drm_dev_alloc() is a more convenient variant that has been
> > introduced explicitly to not require that drm_dev_put(), and states it
> > as such in the documentation [13]. It does so by adding a device-managed
> > action that will call drm_dev_put() [14].
> > 
> > Device-managed actions are ran as part devres_release_all() that is
> > called by device_release() [15], itself being run when the last
> > reference on the device is put back [16][17][18].
> > 
> > So if we sum things up, the DRM device will only give its last root
> > device reference when the last DRM device reference will be put back,
> > and the last DRM device reference will be put back when the last device
> > reference will be put back, which sounds very circular to me, with both
> > ending up in a deadlock scenario.
> > 
> > I've added two kunit tests that demonstrate the issue: we register a
> > device, allocate and register a DRM device, register a DRM-managed
> > action, remove the DRM device and the parent device, and wait for the
> > action to execute. drm_register_unregister_with_devm_test() uses the
> > broken(?) devm_drm_dev_alloc and is failing.
> > drm_register_unregister_test uses the deprecated drm_dev_alloc() that
> > requires an explicit call to drm_dev_put() which works fine.
> > 
> > It's also worth noting that Thomas tested with simpledrm and it seems to
> > work fine.
> 
> Indeed, the transition from simpledrm to a full-blown DRM driver handles
> this properly. It's using a platform_device_unregister() [1] and
> eventually device_del() [2][3]. That part is handled just like
> root_device_unregister() [4][5]. Basically, both will call device_del(),
> and then device_put(), so device_del() is called while holding a
> reference.
> 
> As we've seen before, at this point the DRM driver still holds a
> reference to the device as well.
> 
> device_del() will call bus_remove_device() [6], which will be skipped
> for the root device since it doesn't have a bus [7].
> 
> It will then call device_release_driver() [8], which is basically forwarded
> to __device_release_driver() [9][10], that will call device_unbind_cleanup() [11].
> 
> device_unbind_cleanup() calls devres_release_all() directly [12], that
> runs all the device-managed actions [13]. And it does so WHILE THERE'S
> STILL A REFCOUNT OF 2!
> 
> I would expect the call to devres_release_all to happen only in
> device_release, once all the device reference have been put back. Not 4
> functions in as a side effect, and while there's still some active
> references.

So maybe I'm missing something, but devm is tied to the device's
existence, not to the lifetime of the struct device.

Or well, it's complicated, there's actually two cleanups, once when doing
the hotunplug/unregister, and once when the final struct device cleanup
happens. This is why devm_drm_dev_alloc absolutely must be in the former
cleanup group, otherwise you have the loop you've described.

With usual device model this should all Just Work, if you hand roll your
struct device it gets more complicated. I've reworked how this works for
vgem/vkms a while ago in an attempt to make them look more like real
devices/drivers, but it's a bit nasty.

But yeah if you don't have that unregister step, then you have a loop and
the drm managed stuff never gets freed. That is by design, because the
full lifetime dependencies are

"hw device" (devm-managed) -> "drm device" (drmm-managed) -> "struct
device sw pieces" (kref, nothing else should be attached here really)

Cheers,
Daniel

> 
> > Using a platform_device instead of the root_device doesn't
> > change anything to the outcome in my tests, so there might be a more
> > subtle behaviour involved.
> 
> This one has the same symptom but a different cause. I was just
> registering a platform_device but it wasn't bound to any driver. While
> it's valid to do so according to that comment [13], it doesn't have any
> driver so the check for the driver in device_release_driver() [8], and
> never hits device_unbind_cleanup().
> 
> Thanks again to Thomas and Javier for their help
> Maxime
> 
> 1: https://elixir.bootlin.com/linux/latest/source/drivers/video/aperture.c#L199
> 2: https://elixir.bootlin.com/linux/latest/source/drivers/base/platform.c#L793
> 3: https://elixir.bootlin.com/linux/latest/source/drivers/base/platform.c#L751
> 4: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L4153
> 5: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L3733
> 6: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L3704
> 7: https://elixir.bootlin.com/linux/latest/source/drivers/base/bus.c#L511
> 8: https://elixir.bootlin.com/linux/latest/source/drivers/base/bus.c#L529
> 9: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L1298
> 10: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L1275
> 11: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L1255
> 12: https://elixir.bootlin.com/linux/latest/source/drivers/base/dd.c#L530
> 13: https://elixir.bootlin.com/linux/latest/source/drivers/base/core.c#L2312
  

Hi

On Sat, Nov 19, 2022 at 12:14:14PM +0100, Daniel Vetter wrote:
> On Fri, Nov 18, 2022 at 05:15:58PM +0100, Maxime Ripard wrote:
> > Hi,
> > 
> > Javier and I looked into it some more today, and I think we have a
> > better idea of what is going on.
> > 
> > On Thu, Nov 17, 2022 at 05:53:11PM +0100, Maxime Ripard wrote:
> > > After trying to get more kunit tests for KMS, I found out that the
> > > recent kunit helpers we merged to create a DRM device [1] are broken and
> > > won't free their device-managed and DRM-managed resources.
> > > 
> > > With some help from Thomas, we've dug into this and it turns out that if
> > > we allocate a device with root_device_register, initialise our drm
> > > device with devm_drm_dev_alloc(), register it using drm_dev_register(),
> > > unregister it using drm_dev_unregister/drm_dev_unplug and then remove
> > > the parent device, neither the device managed nor the DRM managed
> > > actions are run.
> > > 
> > > root_device_register initializes the device by eventually calling
> > > device_initialize() which sets the initial reference count of the root
> > > device to 1 [2]. devm_drm_dev_alloc() then comes in, drm_dev_init() will
> > > increase the root device refcount [3] and initialize our DRM device to 1
> > > [4]. drm_dev_register(), through drm_minor_register() and device_add(),
> > > will increase the root device refcount [5].
> > > 
> > > When unrolling things, drm_dev_unregister(), through
> > > drm_minor_unregister() and device_del(), will give up its reference [6].
> > > root_device_unregister(), through device_unregister(), will also give up
> > > its own [7].
> > > 
> > > So we end up with this for the reference counts:
> > > 
> > > +------------------------+-------------+------------+
> > > |         funcs          | root device | DRM device |
> > > +------------------------+-------------+------------+
> > > | root_device_register   |           1 | N/A        |
> > > | devm_drm_dev_alloc     |           2 | 1          |
> > > | drm_dev_register       |           3 | 1          |
> > > | drm_dev_unregister     |           2 | 1          |
> > > | root_device_unregister |           1 | 1          |
> > > +------------------------+-------------+------------+
> > > 
> > > If we go back to the list of reference taken, the root device reference
> > > and the initial drm_device reference, both taken by devm_drm_dev_alloc()
> > > through drm_dev_init(), haven't been put back.
> > > 
> > > If we look at the drm_dev_init code(), we can see that it sets up a
> > > DRM-managed action [8] that will put back the device reference [9]. The
> > > DRM-managed code is executed by the drm_managed_cleanup() function, that
> > > is executed as part of a release hook [10] executed once we give up the
> > > final reference to the DRM device [11].
> > > 
> > > If we go back a little, the final reference to the DRM device is
> > > actually the initial one setup by devm_drm_dev_alloc(). This function
> > > has superseded drm_dev_alloc(), with the documentation that we do need a
> > > final drm_dev_put() to put back our final reference [12].
> > > 
> > > devm_drm_dev_alloc() is a more convenient variant that has been
> > > introduced explicitly to not require that drm_dev_put(), and states it
> > > as such in the documentation [13]. It does so by adding a device-managed
> > > action that will call drm_dev_put() [14].
> > > 
> > > Device-managed actions are ran as part devres_release_all() that is
> > > called by device_release() [15], itself being run when the last
> > > reference on the device is put back [16][17][18].
> > > 
> > > So if we sum things up, the DRM device will only give its last root
> > > device reference when the last DRM device reference will be put back,
> > > and the last DRM device reference will be put back when the last device
> > > reference will be put back, which sounds very circular to me, with both
> > > ending up in a deadlock scenario.
> > > 
> > > I've added two kunit tests that demonstrate the issue: we register a
> > > device, allocate and register a DRM device, register a DRM-managed
> > > action, remove the DRM device and the parent device, and wait for the
> > > action to execute. drm_register_unregister_with_devm_test() uses the
> > > broken(?) devm_drm_dev_alloc and is failing.
> > > drm_register_unregister_test uses the deprecated drm_dev_alloc() that
> > > requires an explicit call to drm_dev_put() which works fine.
> > > 
> > > It's also worth noting that Thomas tested with simpledrm and it seems to
> > > work fine.
> > 
> > Indeed, the transition from simpledrm to a full-blown DRM driver handles
> > this properly. It's using a platform_device_unregister() [1] and
> > eventually device_del() [2][3]. That part is handled just like
> > root_device_unregister() [4][5]. Basically, both will call device_del(),
> > and then device_put(), so device_del() is called while holding a
> > reference.
> > 
> > As we've seen before, at this point the DRM driver still holds a
> > reference to the device as well.
> > 
> > device_del() will call bus_remove_device() [6], which will be skipped
> > for the root device since it doesn't have a bus [7].
> > 
> > It will then call device_release_driver() [8], which is basically forwarded
> > to __device_release_driver() [9][10], that will call device_unbind_cleanup() [11].
> > 
> > device_unbind_cleanup() calls devres_release_all() directly [12], that
> > runs all the device-managed actions [13]. And it does so WHILE THERE'S
> > STILL A REFCOUNT OF 2!
> > 
> > I would expect the call to devres_release_all to happen only in
> > device_release, once all the device reference have been put back. Not 4
> > functions in as a side effect, and while there's still some active
> > references.
> 
> So maybe I'm missing something, but devm is tied to the device's
> existence, not to the lifetime of the struct device.

Yes, devm_ resources can be released even if somebody is keeping
positive reference to struct device FWICT, after bus->remove() you
can not access them.

> Or well, it's complicated, there's actually two cleanups, once when doing
> the hotunplug/unregister, and once when the final struct device cleanup
> happens. This is why devm_drm_dev_alloc absolutely must be in the former
> cleanup group, otherwise you have the loop you've described.

There are 2 separate managed resources list: one in drm_device and one
in struct device. Those are cleaned up separately.

But if we are talking about struct device there is only one cleanup.
However it can happen from few entry points that ended up on:
device_unbind_cleanup() -> devres_release_all()

And there is a problem when mixing devm_ and drmm_, because you have
to correctly chose devm_ resources and avoid those which can still be
accessed via drm callbacks. Or block access tho those after
bus->remove / sysfs unbind. Not sure it this is related to original,
problem presented here by Maxime though. Just my thoughts from
converting intel_vpu to use drmm.

Regards
Stanislaw