[PATCH xf86-video-amdgpu] amdgpu_probe: Do not close server managed drm fds

[Hans de Goede]

Hi,

On 19-10-16 04:42, Michel Dänzer wrote:
>
> [ Moving to the amd-gfx mailing list, where xf86-video-amdgpu (and -ati)
> patches are reviewed ]
>
> On 18/10/16 11:48 PM, Hans de Goede wrote:
>> This fixes the xserver only seeing AMD/ATI devices supported by the amdgpu
>> driver, as by the time xf86-video-ati gets a chance to probe them, the
>> fd has been closed.
>
> That basically makes sense, but I have one question: Why does amdgpu get
> probed in the first place in that case? I was under the impression that
> this should only happen for GPUs bound to the amdgpu kernel driver, due
> to Section "OutputClass" in /usr/share/X11/xorg.conf.d/10-amdgpu.conf .

Good question, I honestly don't know and I do not have time to investigate.
You should be able to reproduce this yourself, if not let me know.

>> diff --git a/src/amdgpu_probe.c b/src/amdgpu_probe.c
>> index 213d245..5d17fe5 100644
>> --- a/src/amdgpu_probe.c
>> +++ b/src/amdgpu_probe.c
>> @@ -142,6 +142,15 @@ static int amdgpu_kernel_open_fd(ScrnInfoPtr pScrn, char *busid,
>>  	return fd;
>>  }
>>
>> +static void amdgpu_kernel_close_fd(AMDGPUEntPtr pAMDGPUEnt)
>> +{
>> +#ifdef XF86_PDEV_SERVER_FD
>> +	if (!(pAMDGPUEnt->platform_dev &&
>> +	      pAMDGPUEnt->platform_dev->flags & XF86_PDEV_SERVER_FD))
>> +#endif
>> +		drmClose(pAMDGPUEnt->fd);
>> +}
>
> AMDGPUFreeRec could also be refactored to call this function.

Ack.

>> @@ -164,7 +173,7 @@ static Bool amdgpu_open_drm_master(ScrnInfoPtr pScrn, AMDGPUEntPtr pAMDGPUEnt,
>>  	if (err != 0) {
>>  		xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
>>  			   "[drm] failed to set drm interface version.\n");
>> -		drmClose(pAMDGPUEnt->fd);
>> +		amdgpu_kernel_close_fd(pAMDGPUEnt);
>>  		return FALSE;
>>  	}
>> @@ -252,7 +261,7 @@ static Bool amdgpu_get_scrninfo(int entity_num, struct pci_device *pci_dev)
>>  	return TRUE;
>>
>>  error_amdgpu:
>> -	drmClose(pAMDGPUEnt->fd);
>> +	amdgpu_kernel_close_fd(pAMDGPUEnt);
>>  error_fd:
>>  	free(pPriv->ptr);
>>  error:
>
> These two hunks aren't really necessary, right? These only get called
> from amdgpu_pci_probe, in which case pAMDGPUEnt->platform_dev remains NULL.

The names of the functions do not imply amdgpu_pci_probe() use only, so even
though that is true today it might not be true in the future. IOW better
safe then sorry.

>> @@ -347,6 +356,7 @@ amdgpu_platform_probe(DriverPtr pDriver,
>>
>>  		pPriv->ptr = xnfcalloc(sizeof(AMDGPUEntRec), 1);
>>  		pAMDGPUEnt = pPriv->ptr;
>> +		pAMDGPUEnt->platform_dev = dev;
>>  		pAMDGPUEnt->fd = amdgpu_kernel_open_fd(pScrn, busid, dev);
>>  		if (pAMDGPUEnt->fd < 0)
>>  			goto error_fd;
>> @@ -365,7 +375,6 @@ amdgpu_platform_probe(DriverPtr pDriver,
>>  		pAMDGPUEnt = pPriv->ptr;
>>  		pAMDGPUEnt->fd_ref++;
>>  	}
>> -	pAMDGPUEnt->platform_dev = dev;
>>
>>  	xf86SetEntityInstanceForScreen(pScrn, pEnt->index,
>>  				       xf86GetNumEntityInstances(pEnt->
>
> These two hunks aren't really necessary either, are they?

Actually they are, quoting some more of the (new after the patch) code:

                 pAMDGPUEnt->platform_dev = dev;
                 pAMDGPUEnt->fd = amdgpu_kernel_open_fd(pScrn, busid, dev);
                 if (pAMDGPUEnt->fd < 0)
                         goto error_fd;

                 pAMDGPUEnt->fd_ref = 1;

                 if (amdgpu_device_initialize(pAMDGPUEnt->fd,
                                              &major_version,
                                              &minor_version,
                                              &pAMDGPUEnt->pDev)) {
                         xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
                                    "amdgpu_device_initialize failed\n");
                         goto error_amdgpu;
		}

...

error_amdgpu:
         amdgpu_kernel_close_fd(pAMDGPUEnt);
error_fd:
         free(pPriv->ptr);
error:
         free(busid);
         return FALSE;
}

And amdgpu_kernel_close_fd accesses pAMDGPUEnt->platform_dev, so it needs
to be set earlier then before.

Shall I send a v2 with AMDGPUFreeRec refactored to call amdgpu_kernel_close_fd
and the rest kept as is ?

Regards,

Hans