[PATCH 0/2] core: Ensure drivers are given a chance to re-probe() late-on

[PATCH 0/2] core: Ensure drivers are given a chance to re-probe() late-on

[Lee Jones]

If a driver's resources are not available at probe() time, it is
common practice to defer probe (return -EPROBE_DEFER) until a time
where the required dependencies have been satisfied.  However, if a
driver's resources are not provided by other drivers probe()ing
e.g. in a case where a driver is waiting for initramfs or rootfs to
be mounted in order to retrieve firmware located within, then
eventually we will run out of binding drivers (the trigger to re-run
deferred driver's probe()), and another re-probe() opportunity will
not be given.

This patch ensures that all drivers still on the deferred list are
given at least one more shot at probe()ing very late on in boot.  This
ensures drivers who find themselves in the position explained in the
aforementioned example are able to locate their resources, resulting
in a successful probe().

Lee Jones (2):
  driver-core: Add externally callable function to trigger deferred
    probes
  init/main: Trigger re-execution of deferred deferred probe()s

 drivers/base/dd.c      | 8 ++++++++
 include/linux/device.h | 1 +
 init/main.c            | 1 +
 3 files changed, 10 insertions(+)

-- 
2.9.0

[PATCH 1/2] driver-core: Add externally callable function to trigger deferred probes

[Lee Jones]

Deferred probes are currently only re-executed if new drivers are successfully
registered and probed.  Drivers which defer their probes based on things like
missing firmware located on rootfs will sometimes only be re-probed if kernel
modules are found.  Unfortunately, this is fragile and can not be guaranteed
or relied upon.  This call allows us to trigger probe re-execution at any
point during boot.  For instance after rootfs is mounted in order to collect
firmwares in rootfs.

Tested-by: Peter Griffin <peter.griffin@linaro.org>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
---
 drivers/base/dd.c      | 8 ++++++++
 include/linux/device.h | 1 +
 2 files changed, 9 insertions(+)

diff --git a/drivers/base/dd.c b/drivers/base/dd.c
index 16688f5..ab20851 100644
--- a/drivers/base/dd.c
+++ b/drivers/base/dd.c
@@ -203,6 +203,14 @@ void device_unblock_probing(void)
 }
 
 /**
+ * deferred_probe_tigger() - Trigger a deferred probe
+ */
+void deferred_probe_trigger(void)
+{
+	driver_deferred_probe_trigger();
+}
+
+/**
  * deferred_probe_initcall() - Enable probing of deferred devices
  *
  * We don't want to get in the way when the bulk of drivers are getting probed.
diff --git a/include/linux/device.h b/include/linux/device.h
index 38f0281..d441471 100644
--- a/include/linux/device.h
+++ b/include/linux/device.h
@@ -293,6 +293,7 @@ extern struct device_driver *driver_find(const char *name,
 					 struct bus_type *bus);
 extern int driver_probe_done(void);
 extern void wait_for_device_probe(void);
+extern void deferred_probe_trigger(void);
 
 
 /* sysfs interface for exporting driver attributes */
-- 
2.9.0

[PATCH 2/2] init/main: Trigger re-execution of deferred deferred probe()s

[Lee Jones]

Some devices require files from rootfs.

Ensure they are given a chance to re-probe() now that the rootfs is mounted.

Tested-by: Peter Griffin <peter.griffin@linaro.org>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
---
 init/main.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/init/main.c b/init/main.c
index 4c17fda..0b905a1 100644
--- a/init/main.c
+++ b/init/main.c
@@ -1035,6 +1035,7 @@ static noinline void __init kernel_init_freeable(void)
 	 * and default modules
 	 */
 
+	deferred_probe_trigger();
 	integrity_load_keys();
 	load_default_modules();
 }
-- 
2.9.0