I'm looking for code reviews for a fix to 6878048, if anyone's willing to look at it. It's not a large change, less than 300 lines changed. The webrev is here: http://cr.opensolaris.org/~cmynhier/6878048/.
The problem with fuser(1M) is that the kernel function dofusers() does a walk of the address space (the AVL tree) for every process trying to match the target vnode. What I've done is add a hash table to struct as to keep track of what vnodes are mapped into the AS. In the best case, almost none of the AS segments will be mmap'd files, so this will save a significant amount of time for fuser(1M). In this case, the additional space requirement is the essentially the size of an empty hash table per process. The following is an example of the performance benefit to be had (where the bulk of the AS segments are shared memory segments and the time is in milliseconds): AS segments current with fix 0 10.93 5.72 200000 40.08 5.73 400000 69.22 5.74 600000 98.42 5.78 800000 128.05 5.81 1000000 156.22 5.75 1200000 184.31 5.69 1400000 211.76 5.74 In the worst case, all of the AS segments will be mmap'd files, each segment from a different file. In this case, there is no performance benefit, and the additional space requirement is essentially a second copy of the AVL tree representing the AS. Chad _______________________________________________ opensolaris-code mailing list opensolaris-code@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
