[17:17] <jafa> seeing a possible problem with virtual memory reporting when using huge tlb pages
[17:18] <jafa> app allocates 97 x 1GB huge tlb pages (locked) using mmap64 so the app is using 97GB plus normal memory allocation
[17:19] <jafa> proc/n/status reports VmLck as 0kB and VmRSS as 3GB
[17:20] <jafa> all the common utils (top, htop, ps, etc) report 3GB of usage
[17:20] <jafa> proc/n/status: https://pastebin.com/JqiCMbs3
[17:22] <jafa> this app also mmap's 1TB of NVMe space which is why the VmSize value is >1TB
[17:23] <jafa> it is a problem for administrators who run top and see 3GB of memory usage when the app is actually using 97+3=100GB of ram
[17:33] <jafa> should the kernel be reporting the 97GB of locked hugetlb pages in VmLck and VmRSS?
[19:13] <mouai> Hi