On Sat, Apr 18, 2020 at 11:55:05AM -0700, Joe Perches wrote: > On Sat, 2020-04-18 at 11:53 -0700, Randy Dunlap wrote: > > On 4/18/20 11:50 AM, Matthew Wilcox wrote: > > > On Sat, Apr 18, 2020 at 11:41:09AM -0700, Randy Dunlap wrote: > > > > @@ -294,11 +295,11 @@ void dev_coredumpm(struct device *dev, s > > > > > > > > if (sysfs_create_link(&devcd->devcd_dev.kobj, &dev->kobj, > > > > "failing_device")) > > > > - /* nothing - symlink will be missing */; > > > > + do_empty(); /* nothing - symlink will be missing */ > > > > > > > > if (sysfs_create_link(&dev->kobj, &devcd->devcd_dev.kobj, > > > > "devcoredump")) > > > > - /* nothing - symlink will be missing */; > > > > + do_empty(); /* nothing - symlink will be missing */ > > > > > > > > INIT_DELAYED_WORK(&devcd->del_wk, devcd_del); > > > > schedule_delayed_work(&devcd->del_wk, DEVCD_TIMEOUT); > > > > > > Could just remove the 'if's? > > > > > > + sysfs_create_link(&devcd->devcd_dev.kobj, &dev->kobj, > > > + "failing_device"); > > > > > > > OK. > > sysfs_create_link is __must_check Oh, I missed the declaration -- I just saw the definition. This is a situation where __must_check hurts us and it should be removed. Or this code is wrong and it should be WARN(sysfs_create_link(&devcd->devcd_dev.kobj, &dev->kobj, "failing_device"); like drivers/pci/controller/vmd.c and drivers/i2c/i2c-mux.c Either way, the do_empty() construct feels like the wrong way of covering up the warning.