drm/radeon: Return -EPROBE_DEFER when amdkfd not loaded

radeon must load only after amdkfd's loading has been completed. If that
is not enforced, then radeon's call into amdkfd's functions will cause a
kernel BUG.

When radeon and amdkfd are built as kernel modules, that rule is
enforced by the kernel's modules loading mechanism. When radeon and
amdkfd are built inside the kernel image, that rule is enforced by
ordering in the drm Makefile (amdkfd before radeon).

Instead of using drm Makefile ordering, we can now use deferred
loading as amdkfd now returns -EPROBE_DEFER in kgd2kfd_init() when it is
not yet loaded.

This patch defers radeon loading by propagating -EPROBE_DEFER to the
kernel's drivers loading infrastructure. That will put radeon into the
pending drivers list (see description in dd.c). Once amdkfd is loaded,
a call to kgd2kfd_init() will return successfully and radeon will be
able to load.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/radeon/radeon_drv.c

@@ -321,6 +321,14 @@ static int radeon_pci_probe(struct pci_dev *pdev,
 {
 	int ret;
 
+	/*
+	 * Initialize amdkfd before starting radeon. If it was not loaded yet,
+	 * defer radeon probing
+	 */
+	ret = radeon_kfd_init();
+	if (ret == -EPROBE_DEFER)
+		return ret;
+
 	/*
 	 * apple-gmux is needed on dual GPU MacBook Pro
 	 * to probe the panel if we're the inactive GPU.
@@ -581,8 +589,6 @@ static int __init radeon_init(void)
 		return -EINVAL;
 	}
 
-	radeon_kfd_init();
-
 	/* let modprobe override vga console setting */
 	return drm_pci_init(driver, pdriver);
 }

drivers/gpu/drm/radeon/radeon_kfd.c

@@ -132,35 +132,34 @@ static const struct kfd2kgd_calls kfd2kgd = {
 
 static const struct kgd2kfd_calls *kgd2kfd;
 
-bool radeon_kfd_init(void)
+int radeon_kfd_init(void)
 {
+	int ret;
+
 #if defined(CONFIG_HSA_AMD_MODULE)
-	bool (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**);
+	int (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**);
 
 	kgd2kfd_init_p = symbol_request(kgd2kfd_init);
 
 	if (kgd2kfd_init_p == NULL)
-		return false;
+		return -ENOENT;
 
-	if (kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kgd2kfd)) {
+	ret = kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kgd2kfd);
+	if (ret) {
 		symbol_put(kgd2kfd_init);
 		kgd2kfd = NULL;
-
-		return false;
 	}
 
-	return true;
 #elif defined(CONFIG_HSA_AMD)
-	if (kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd)) {
+	ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd);
+	if (ret)
 		kgd2kfd = NULL;
 
-		return false;
-	}
-
-	return true;
 #else
-	return false;
+	ret = -ENOENT;
 #endif
+
+	return ret;
 }
 
 void radeon_kfd_fini(void)

drivers/gpu/drm/radeon/radeon_kfd.h

@@ -33,7 +33,7 @@
 
 struct radeon_device;
 
-bool radeon_kfd_init(void);
+int radeon_kfd_init(void);
 void radeon_kfd_fini(void);
 
 void radeon_kfd_suspend(struct radeon_device *rdev);