Hi linux-mm,
Do we expect the reporter process-wide RSS figure to differ from that of the sum of the RSS fields of the individual VMAs as reported via smaps? They're tracked very differently: the former is the sum of MM_FILEPAGES, MM_ANONPAGES, and MM_SHMEMPAGES, while the latter comes from counting pages in smaps_pte_entry (huge pages don't appear in this context). The sum of the smaps rss fields is sometimes larger than the counter-based values from status. Same with the anonymous sizes and the new anonymous RSS in status.
Weirdly, I can't reproduce the discrepancy in a minimal UML boot with init=/bin/sh (either with 4.4.88 or with latest master), but I can see this discrepancy appear on both Android systems and normal Ubuntu 14.04 systems.
Before I spend more time debugging: is there something obvious that I missed? Where should I be looking?
Thanks!
