Re: kobj->state_initialized is never set to 0
On Fri, Jul 20, 2018 at 3:39 PM, Greg KH wrote: > On Fri, Jul 20, 2018 at 03:33:22PM -0300, Helen Koike wrote: >> Hi, >> >> I am simulating hotplugging a platform device (to the >> media/platform/vimc driver), so I register it using >> platform_device_register, then I unregister it with >> platform_device_unregister, then I register it again, but the code >> doesn't allow me to register something that was registered before due >> to this line: >> >> https://github.com/torvalds/linux/blob/master/lib/kobject.c#L334 >> >> The variable kobj->state_initialized is set to 1 at >> >> https://github.com/torvalds/linux/blob/master/lib/kobject.c#L194 >> >> but it is never set to 0 again. > > Yes, do not do that, that is why we wrote that code. Never "recycle" a > struct device, or a kobject, as you never really know if it is still in > use or not. That is why it needs to be dynamically created, not > statically created. > >> Is this how it is supposed to work? Or kobj->state_initialized is set >> to 0 somewhere else I couldn't find? > > Nope, don't do this please. Again, never recycle these objects, bad > things will happen as the code is not designed to do that for good > reasons (think object lifetimes, you do not control them fully, other > things do which you have no control over.) > > hope this helps, > > greg k-h Yes, this helps a lot, thank you! Helen ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: kobj->state_initialized is never set to 0
On Fri, Jul 20, 2018 at 03:33:22PM -0300, Helen Koike wrote: > Hi, > > I am simulating hotplugging a platform device (to the > media/platform/vimc driver), so I register it using > platform_device_register, then I unregister it with > platform_device_unregister, then I register it again, but the code > doesn't allow me to register something that was registered before due > to this line: > > https://github.com/torvalds/linux/blob/master/lib/kobject.c#L334 > > The variable kobj->state_initialized is set to 1 at > > https://github.com/torvalds/linux/blob/master/lib/kobject.c#L194 > > but it is never set to 0 again. Yes, do not do that, that is why we wrote that code. Never "recycle" a struct device, or a kobject, as you never really know if it is still in use or not. That is why it needs to be dynamically created, not statically created. > Is this how it is supposed to work? Or kobj->state_initialized is set > to 0 somewhere else I couldn't find? Nope, don't do this please. Again, never recycle these objects, bad things will happen as the code is not designed to do that for good reasons (think object lifetimes, you do not control them fully, other things do which you have no control over.) hope this helps, greg k-h ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
kobj->state_initialized is never set to 0
Hi, I am simulating hotplugging a platform device (to the media/platform/vimc driver), so I register it using platform_device_register, then I unregister it with platform_device_unregister, then I register it again, but the code doesn't allow me to register something that was registered before due to this line: https://github.com/torvalds/linux/blob/master/lib/kobject.c#L334 The variable kobj->state_initialized is set to 1 at https://github.com/torvalds/linux/blob/master/lib/kobject.c#L194 but it is never set to 0 again. Is this how it is supposed to work? Or kobj->state_initialized is set to 0 somewhere else I couldn't find? Thanks Helen ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies