docs: x86: Remove obsolete information about x86_64 vmalloc() faulting
authorPeilin Ye <peilin.ye@bytedance.com>
Wed, 18 Aug 2021 22:01:23 +0000 (15:01 -0700)
committerJonathan Corbet <corbet@lwn.net>
Fri, 20 Aug 2021 17:14:11 +0000 (11:14 -0600)
commit251a7b3edc197a3947b8cb56fffe61d811aba0a5
treed20bd8c38816a94ba09eb887fdeda95c239b8934
parentd44f571ff5ce51298df520cc46c3a9f5b983fc0a
docs: x86: Remove obsolete information about x86_64 vmalloc() faulting

x86_64 vmalloc() mappings are no longer "synchronized" among page tables
via faulting since commit 6eb82f994026 ("x86/mm: Pre-allocate P4D/PUD
pages for vmalloc area"), since the corresponding P4D or PUD pages are
now preallocated at boot, by preallocate_vmalloc_pages().  Drop the
"lazily synchronized" description for less confusion.

While this file is x86_64-specific, it is worth noting that things are
different for x86_32, where vmalloc()-related changes to `init_mm.pgd` are
synchronized to all page tables in the system during runtime, via
arch_sync_kernel_mappings().  Unfortunately, this synchronization is
subject to race condition, which is further handled via faulting, see
vmalloc_fault().  See commit 4819e15f740e ("x86/mm/32: Bring back vmalloc
faulting on x86_32") for more details.

Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Peilin Ye <peilin.ye@bytedance.com>
Reviewed-by: Joerg Roedel <jroedel@suse.de>
Link: https://lore.kernel.org/r/20210818220123.2623-1-yepeilin.cs@gmail.com
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Documentation/x86/x86_64/mm.rst