riscv: mm: Do not call pmd dtor on vmemmap page table teardown

In the Linux kernel, the following vulnerability has been resolved:

riscv: mm: Do not call pmd dtor on vmemmap page table teardown

The vmemmap's, which is used for RV64 with SPARSEMEM_VMEMMAP, page

tables are populated using pmd (page middle directory) hugetables.

However, the pmd allocation is not using the generic mechanism used by

the VMA code (e.g. pmd_alloc()), or the RISC-V specific

create_pgd_mapping()/alloc_pmd_late(). Instead, the vmemmap page table

code allocates a page, and calls vmemmap_set_pmd(). This results in

that the pmd ctor is *not* called, nor would it make sense to do so.

Now, when tearing down a vmemmap page table pmd, the cleanup code

would unconditionally, and incorrectly call the pmd dtor, which

results in a crash (best case).

This issue was found when running the HMM selftests:

| tools/testing/selftests/mm# ./test_hmm.sh smoke

| ... # when unloading the test_hmm.ko module

| page: refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x10915b

| flags: 0x1000000000000000(node=0|zone=1)

| raw: 1000000000000000 0000000000000000 dead000000000122 0000000000000000

| raw: 0000000000000000 0000000000000000 00000001ffffffff 0000000000000000

| page dumped because: VM_BUG_ON_PAGE(ptdesc->pmd_huge_pte)

| ------------[ cut here ]------------

| kernel BUG at include/linux/mm.h:3080!

| Kernel BUG [#1]

| Modules linked in: test_hmm(-) sch_fq_codel fuse drm drm_panel_orientation_quirks backlight dm_mod

| CPU: 1 UID: 0 PID: 514 Comm: modprobe Tainted: G W 6.12.0-00982-gf2a4f1682d07 #2

| Tainted: [W]=WARN

| Hardware name: riscv-virtio qemu/qemu, BIOS 2024.10 10/01/2024

| epc : remove_pgd_mapping+0xbec/0x1070

| ra : remove_pgd_mapping+0xbec/0x1070

| epc : ffffffff80010a68 ra : ffffffff80010a68 sp : ff20000000a73940

| gp : ffffffff827b2d88 tp : ff6000008785da40 t0 : ffffffff80fbce04

| t1 : 0720072007200720 t2 : 706d756420656761 s0 : ff20000000a73a50

| s1 : ff6000008915cff8 a0 : 0000000000000039 a1 : 0000000000000008

| a2 : ff600003fff0de20 a3 : 0000000000000000 a4 : 0000000000000000

| a5 : 0000000000000000 a6 : c0000000ffffefff a7 : ffffffff824469b8

| s2 : ff1c0000022456c0 s3 : ff1ffffffdbfffff s4 : ff6000008915c000

| s5 : ff6000008915c000 s6 : ff6000008915c000 s7 : ff1ffffffdc00000

| s8 : 0000000000000001 s9 : ff1ffffffdc00000 s10: ffffffff819a31f0

| s11: ffffffffffffffff t3 : ffffffff8000c950 t4 : ff60000080244f00

| t5 : ff60000080244000 t6 : ff20000000a73708

| status: 0000000200000120 badaddr: ffffffff80010a68 cause: 0000000000000003

| [<ffffffff80010a68>] remove_pgd_mapping+0xbec/0x1070

| [<ffffffff80fd238e>] vmemmap_free+0x14/0x1e

| [<ffffffff8032e698>] section_deactivate+0x220/0x452

| [<ffffffff8032ef7e>] sparse_remove_section+0x4a/0x58

| [<ffffffff802f8700>] __remove_pages+0x7e/0xba

| [<ffffffff803760d8>] memunmap_pages+0x2bc/0x3fe

| [<ffffffff02a3ca28>] dmirror_device_remove_chunks+0x2ea/0x518 [test_hmm]

| [<ffffffff02a3e026>] hmm_dmirror_exit+0x3e/0x1018 [test_hmm]

| [<ffffffff80102c14>] __riscv_sys_delete_module+0x15a/0x2a6

| [<ffffffff80fd020c>] do_trap_ecall_u+0x1f2/0x266

| [<ffffffff80fde0a2>] _new_vmalloc_restore_context_a0+0xc6/0xd2

| Code: bf51 7597 0184 8593 76a5 854a 4097 0029 80e7 2c00 (9002) 7597

| ---[ end trace 0000000000000000 ]---

| Kernel panic - not syncing: Fatal exception in interrupt

Add a check to avoid calling the pmd dtor, if the calling context is

vmemmap_free().