On Tue, Jul 7, 2020 at 6:48 PM Abhishek Pandit-Subedi
[off-list ref] wrote:

Hi Rafael,

(resent in plain text)

On Tue, Jul 7, 2020 at 9:28 AM Rafael J. Wysocki [off-list ref] wrote:

quoted

On Tue, Jul 7, 2020 at 6:24 PM Abhishek Pandit-Subedi
[off-list ref] wrote:

quoted

Udev rules that depend on the power/wakeup attribute don't get triggered
correctly if device_set_wakeup_capable is called after the device is
created. This can happen for several reasons (driver sets wakeup after
device is created, wakeup is changed on parent device, etc) and it seems
reasonable to emit a changed event when adding or removing attributes on
the device.

Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org>
---

Changes in v4:
- Fix warning where returning from void and tested on device

Changes in v3:
- Simplified error handling

Changes in v2:
- Add newline at end of bt_dev_err

 drivers/base/power/sysfs.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/base/power/sysfs.c b/drivers/base/power/sysfs.c
index 24d25cf8ab1487..aeb58d40aac8de 100644
--- a/drivers/base/power/sysfs.c
+++ b/drivers/base/power/sysfs.c

@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0
 /* sysfs entries for device PM */
 #include <linux/device.h>
+#include <linux/kobject.h>
 #include <linux/string.h>
 #include <linux/export.h>
 #include <linux/pm_qos.h>

@@ -739,12 +740,18 @@ int dpm_sysfs_change_owner(struct device *dev, kuid_t kuid, kgid_t kgid)

 int wakeup_sysfs_add(struct device *dev)
 {
-       return sysfs_merge_group(&dev->kobj, &pm_wakeup_attr_group);
+       int ret = sysfs_merge_group(&dev->kobj, &pm_wakeup_attr_group);
+
+       if (ret)
+               return ret;
+
+       return kobject_uevent(&dev->kobj, KOBJ_CHANGE);

So let me repeat the previous comment:

If you return an error here, it may confuse the caller to think that
the operation has failed completely, whereas the merging of the
attribute group has been successful already.

I don't think that an error can be returned at this point.

The caller looks at the return code and just logs that an error
occurred (no other action). It's also unlikely for kobject_uevent to
fail (I saw mostly -ENOMEM and an -ENOENT when the kobj wasn't in the
correct set).

Call site:
    int ret = wakeup_sysfs_add(dev);

    if (ret)
        dev_info(dev, "Wakeup sysfs attributes not added\n");

Yes, which is confusing, because the wakeup attributes may in fact
have been added.  Which is my point.

So I'm ok with either keeping this as-is (caller isn't getting
confused, just logging) or swallowing the return of kobject_uevent.

I would just ignore the return value of kobject_uevent() along the
lines of wakeup_sysfs_remove() below.

Thanks!

quoted

quoted

 }

 void wakeup_sysfs_remove(struct device *dev)
 {
        sysfs_unmerge_group(&dev->kobj, &pm_wakeup_attr_group);
+       kobject_uevent(&dev->kobj, KOBJ_CHANGE);
 }

 int pm_qos_sysfs_add_resume_latency(struct device *dev)
--
2.27.0.212.ge8ba1cc988-goog