Is there any argument against using the DNOTIFY/poll upcall approach
(against pseudo files e.g. in /proc or /sysfs) that e.g. nfs uses now to
get from kernel space to get data back from user space helper routines
(e.g. for idmapping and gssapi)? Since there could be security and
potential denial of service implications in getting it wrong, I wanted
to check if there are recent changes that I have missed that would be
better approaches to this.
I was hoping that the kernel event mechanism would standardize a way to
do this with less code someday.