hugetlb: No coalescing of adjacent mappings in /proc/pid/maps

[dada1]

Hello William

I noticed a small problem if a process allocates hugetlb memory using

sys_alloc_hugepages(0, 0, 4*1024*1024, PROT_READ|PROT_WRITE, 0) ;

If this process uses a lot of hugetlb pages, its /proc/pid/maps contains one
entry for each mapping.

Could it be possible to coalesce all consecutive mappings onto only one ? As
standard mmap() calls do ?