Messages in this thread | | | Date | Thu, 02 Oct 1997 15:02:08 -0400 | From | Bill Hawes <> | Subject | Re: use counts for /proc modules |
| |
Richard Henderson wrote:
> Hmm. I guess what you really need is to be able to determine if the > thing is a module. Which might come in handy for other things as well. > > Have a look at what the scsi subsystem does with its > > if (rscsi_disks[target].device->host->hostt->module) > __MOD_INC_USE_COUNT(rscsi_disks[target].device->host->hostt->module); > > It gets set via init_module in scsi_module.c, btw.
I'm assuming that whoever maintains each subsystem will decide whether it can ever be used as a module, and if so will install a function pointer in the proc_dir_entry structure before registering the directory. The conditional compilation switch for the module can either leave the function pointer as NULL, or just pass a stub function.
What's missing now is that there's no way for the VFS to let the module (if any) know that its data structures are still in use. There needs to be a standardized access point so that proc_iget can add a count when it makes an inode, and delete_inode can decrement it.
Regards, Bill
| |