Re: [PATCH] drm/radeon: don't add files at control minor debugfs directory

["Christian König" <christian.koenig@amd.com>]

Am 05.12.2016 um 09:39 schrieb Nicolai Stange:
> Christian König <christian.koenig@amd.com> writes:
>
>> Am 05.12.2016 um 08:27 schrieb Daniel Vetter:
>>> On Sat, Dec 03, 2016 at 03:47:00PM +0100, Nicolai Stange wrote:
>>>> Since commit 8a357d10043c ("drm: Nerf DRM_CONTROL nodes"), a
>>>> struct drm_device's ->control member is always NULL.
>>>>
>>>> In the case of CONFIG_DEBUG_FS=y, radeon_debugfs_add_files() accesses
>>>> ->control->debugfs_root though. This results in the following Oops:
>>>>
>>>>     BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
>>>>     IP: radeon_debugfs_add_files+0x90/0x100 [radeon]
>>>>     PGD 0
>>>>     Oops: 0000 [#1] SMP
>>>>     [...]
>>>>     Call Trace:
>>>>      ? work_on_cpu+0xb0/0xb0
>>>>      radeon_fence_driver_init+0x120/0x150 [radeon]
>>>>      si_init+0x122/0xd50 [radeon]
>>>>      ? _raw_spin_unlock_irq+0x2c/0x40
>>>>      ? device_pm_check_callbacks+0xb3/0xc0
>>>>      radeon_device_init+0x958/0xda0 [radeon]
>>>>      radeon_driver_load_kms+0x9a/0x210 [radeon]
>>>>      drm_dev_register+0xa9/0xd0 [drm]
>>>>      drm_get_pci_dev+0x9c/0x1e0 [drm]
>>>>      radeon_pci_probe+0xb8/0xe0 [radeon]
>>>>     [...]
>>>>
>>>> Fix this by omitting the drm_debugfs_create_files() call for the
>>>> control minor debugfs directory which is now non-existent anyway.
>>>>
>>>> Fixes: 8a357d10043c ("drm: Nerf DRM_CONTROL nodes")
>>>> Signed-off-by: Nicolai Stange <nicstange@gmail.com>
>>> Applied to drm-misc with Dave's irc ack, thanks for your patch.
>> If it's still worth it the patch is Reviewed-by: Christian König
>> <christian.koenig@amd.com>.
>>
>> On the other hand when ->control is always NULL, why do we still have
>> ->control anyway?
> Yes, I was wondering about that, too.
>
> Quoting from 8a357d10043c ("drm: Nerf DRM_CONTROL nodes"):
>
>    Since I don't like dead uabi, let's remove it. But since this would be
>    a really big change I think it's better to start out small by simply
>    not registering anything. We can garbage-collect the dead code later
>    on, once we're sure it's really not used anywhere.
>
> I'd too prefer compile time errors by purging ->control here. Daniel?

Seconded.

>
>
>> And BTW: Please double check the other drivers as well.
>    # git grep '\->control' -- drivers/gpu/
>    drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:                              adev->ddev->control->debugfs_root,
>    drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:                              adev->ddev->control);
>    drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:                                      adev->ddev->control);
>
> Oops.

Yeah, that's what I expected as well but Daniel said it would only 
affect qxl.

>
>    drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c:        ib_packet->control = (1 << 23) | (1 << 31) |
>    drivers/gpu/drm/drm_drv.c:              return &dev->control;
>
> That's drm_minor_get_slot(dev, type), but grepping for DRM_MINOR_CONTROL
> doesn't yield anything -> dead code.
>
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:        switch (sdvo->controlled_output) {
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:                psb_intel_sdvo->controlled_output |= SDVO_OUTPUT_TMDS0;
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:                psb_intel_sdvo->controlled_output |= SDVO_OUTPUT_TMDS1;
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:        psb_intel_sdvo->controlled_output |= type;
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:                psb_intel_sdvo->controlled_output |= SDVO_OUTPUT_RGB0;
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:                psb_intel_sdvo->controlled_output |= SDVO_OUTPUT_RGB1;
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:                psb_intel_sdvo->controlled_output |= SDVO_OUTPUT_LVDS0;
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:                psb_intel_sdvo->controlled_output |= SDVO_OUTPUT_LVDS1;
>    drivers/gpu/drm/gma500/psb_intel_sdvo.c:                psb_intel_sdvo->controlled_output = 0;
>    drivers/gpu/drm/i915/intel_sdvo.c:      switch (sdvo->controlled_output) {
>    drivers/gpu/drm/i915/intel_sdvo.c:              intel_sdvo->controlled_output |= SDVO_OUTPUT_TMDS0;
>    drivers/gpu/drm/i915/intel_sdvo.c:              intel_sdvo->controlled_output |= SDVO_OUTPUT_TMDS1;
>    drivers/gpu/drm/i915/intel_sdvo.c:      intel_sdvo->controlled_output |= type;
>    drivers/gpu/drm/i915/intel_sdvo.c:              intel_sdvo->controlled_output |= SDVO_OUTPUT_RGB0;
>    drivers/gpu/drm/i915/intel_sdvo.c:              intel_sdvo->controlled_output |= SDVO_OUTPUT_RGB1;
>    drivers/gpu/drm/i915/intel_sdvo.c:              intel_sdvo->controlled_output |= SDVO_OUTPUT_LVDS0;
>    drivers/gpu/drm/i915/intel_sdvo.c:              intel_sdvo->controlled_output |= SDVO_OUTPUT_LVDS1;
>    drivers/gpu/drm/i915/intel_sdvo.c:              intel_sdvo->controlled_output = 0;
>    drivers/gpu/drm/msm/msm_debugfs.c:      ret = late_init_minor(dev->control);
>
> Not an oops but dead code.
>
>    drivers/gpu/drm/qxl/qxl_debugfs.c:                               qdev->ddev->control->debugfs_root,
>    drivers/gpu/drm/qxl/qxl_debugfs.c:                               qdev->ddev->control);
>    drivers/gpu/drm/qxl/qxl_debugfs.c:                                       qdev->ddev->control);
>
> Oops.
>
>
> I'll send compile-only tested patches either tonight or tomorrow. Shall
> they cover the oopses only or the dead code as well?

Please start with the ops, cause we certainly will get complains about 
that rather fast.

Regards,
Christian.

>
>
> Thanks,
>
> Nicolai