Insufficient HugePages allocation results in excessive expansion of PageTables, leading to severe memory pressure

Issue

• The system is highly loaded, under high memory pressure.
• Significant growth of PageTables is observed.

proc/meminfo 
---
MemTotal:       198064408 kB
MemFree:        17773544 kB
MemAvailable:   25111912 kB
Buffers:           22880 kB
Cached:         56739016 kB
SwapCached:      2250624 kB
Active:         45941504 kB
Inactive:       17868432 kB
Active(anon):   44585508 kB
Inactive(anon): 10605308 kB
Active(file):    1355996 kB
Inactive(file):  7263124 kB
    ...
SwapTotal:      121630712 kB
SwapFree:       84829628 kB
    ...
Slab:            2442184 kB
SReclaimable:    1758232 kB
SUnreclaim:       683952 kB
    ...
PageTables:     109033236 kB <<-----
    ...
CommitLimit:    218565764 kB
Committed_AS:   106736164 kB
    ...
HugePages_Total:    2048
HugePages_Free:      117
HugePages_Rsvd:        5
HugePages_Surp:        0
Hugepagesize:       2048 kB
    ...

• Many tasks are entering D-state showing call traces just like this:

[<0>] __lock_page+0x12d/0x230
[<0>] shmem_swapin_page+0x4ea/0x630
[<0>] shmem_getpage_gfp+0x1fc/0x8a0
[<0>] shmem_fault+0x78/0x220
[<0>] __do_fault+0x38/0xc0
[<0>] handle_pte_fault+0x55d/0x880
[<0>] __handle_mm_fault+0x453/0x6c0
[<0>] handle_mm_fault+0xc1/0x1e0
[<0>] do_user_addr_fault+0x1b9/0x450
[<0>] do_page_fault+0x37/0x130
[<0>] page_fault+0x1e/0x30

• The call trace just right above is indicative of major page fault on tmpfs/shmem that is bringing a swapped‑out page back into RAM.
• Since many tasks show the same stack, it appears likely a swap‑in storm.