mm/vmalloc: export clear_vm_uninitialized_flag()

Patch series "Fix KASAN support for KHO restored vmalloc regions".

When KHO restores a vmalloc area, it maps existing physical pages into a
newly allocated virtual memory area.  However, because these areas were
not properly unpoisoned, KASAN would treat any access to the restored
region as out-of-bounds, as seen in the following trace:

BUG: KASAN: vmalloc-out-of-bounds in kho_test_restore_data.isra.0+0x17b/0x2cd
Read of size 8 at addr ffffc90000025000 by task swapper/0/1
[...]
Call Trace:
[...]
kasan_report+0xe8/0x120
kho_test_restore_data.isra.0+0x17b/0x2cd
kho_test_init+0x15a/0x1f0
do_one_initcall+0xd5/0x4b0

The fix involves deferring KASAN's default poisoning by using the
VM_UNINITIALIZED flag during allocation, manually unpoisoning the memory
once it is correctly mapped, and then clearing the uninitialized flag
using a newly exported helper.

This patch (of 2):

Make clear_vm_uninitialized_flag() available to other parts of the kernel
that need to manage vmalloc areas manually, such as KHO for restoring
vmallocs.

Link: https://lkml.kernel.org/r/20260225220223.1695350-1-pasha.tatashin@soleen.com
Link: https://lkml.kernel.org/r/20260225223857.1714801-2-pasha.tatashin@soleen.com
Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Acked-by: Pratyush Yadav (Google) <pratyush@kernel.org>
Cc: Alexander Graf <graf@amazon.com>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: "Uladzislau Rezki (Sony)" <urezki@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

diff --git a/mm/internal.h b/mm/internal.h
index 39ab37bb0e1dd604509d1bc49f51729189d7b0d2..2daa6a744172ef47501776db6bf293c352a1844c 100644 (file)
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -1469,6 +1469,8 @@ int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
 }
 #endif
 
+void clear_vm_uninitialized_flag(struct vm_struct *vm);
+
 int __must_check __vmap_pages_range_noflush(unsigned long addr,
                               unsigned long end, pgprot_t prot,
                               struct page **pages, unsigned int page_shift);

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 6dda97c3799e888c7801c32e6994707969e10d68..b2c2ed650840ceda6d517e122675143b482d5f09 100644 (file)
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -3183,7 +3183,7 @@ void __init vm_area_register_early(struct vm_struct *vm, size_t align)
        kasan_populate_early_vm_area_shadow(vm->addr, vm->size);
 }
 
-static void clear_vm_uninitialized_flag(struct vm_struct *vm)
+void clear_vm_uninitialized_flag(struct vm_struct *vm)
 {
        /*
         * Before removing VM_UNINITIALIZED,