[PATCH RFC] madvise.2: Document MADV_POPULATE

[David Hildenbrand]

Preview of MADV_POPULATE documentation, which is still under discussion:
https://lkml.kernel.org/r/20210217154844.12392-1-david@xxxxxxxxxx

Once/if merged, there will be an official patch to man-page folks.

Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: Arnd Bergmann <arnd@xxxxxxxx>
Cc: Michal Hocko <mhocko@xxxxxxxx>
Cc: Oscar Salvador <osalvador@xxxxxxx>
Cc: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Cc: Minchan Kim <minchan@xxxxxxxxxx>
Cc: Jann Horn <jannh@xxxxxxxxxx>
Cc: Jason Gunthorpe <jgg@xxxxxxxx>
Cc: Dave Hansen <dave.hansen@xxxxxxxxx>
Cc: Hugh Dickins <hughd@xxxxxxxxxx>
Cc: Rik van Riel <riel@xxxxxxxxxxx>
Cc: Michael S. Tsirkin <mst@xxxxxxxxxx>
Cc: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
Cc: Vlastimil Babka <vbabka@xxxxxxx>
Cc: Richard Henderson <rth@xxxxxxxxxxx>
Cc: Ivan Kokshaysky <ink@xxxxxxxxxxxxxxxxxxxx>
Cc: Matt Turner <mattst88@xxxxxxxxx>
Cc: Thomas Bogendoerfer <tsbogend@xxxxxxxxxxxxxxxx>
Cc: "James E.J. Bottomley" <James.Bottomley@xxxxxxxxxxxxxxxxxxxxx>
Cc: Helge Deller <deller@xxxxxx>
Cc: Chris Zankel <chris@xxxxxxxxxx>
Cc: Max Filippov <jcmvbkbc@xxxxxxxxx>
Cc: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
Cc: Rolf Eike Beer <eike-kernel@xxxxxxxxx>
Cc: linux-alpha@xxxxxxxxxxxxxxx
Cc: linux-mips@xxxxxxxxxxxxxxx
Cc: linux-parisc@xxxxxxxxxxxxxxx
Cc: linux-xtensa@xxxxxxxxxxxxxxxx
Cc: linux-arch@xxxxxxxxxxxxxxx
Cc: Linux API <linux-api@xxxxxxxxxxxxxxx>
Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
---
 man2/madvise.2 | 59 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 59 insertions(+)

diff --git a/man2/madvise.2 b/man2/madvise.2
index 2af407212..ff08768a3 100644
--- a/man2/madvise.2
+++ b/man2/madvise.2
@@ -469,6 +469,48 @@ If a page is file-backed and dirty, it will be written back to the backing
 storage.
 The advice might be ignored for some pages in the range when it is not
 applicable.
+.TP
+.BR MADV_POPULATE " (since Linux 5.13)
+Populate (prefault) page tables for the whole range.
+Depending on the underlying mapping, preallocate memory or read the
+underlying file.
+Do not generate
+.B SIGBUS
+when populating fails, return an error instead.
+The populate semantics match
+.BR MAP_POPULATE
+(see
+.BR mmap (2))
+with the exception that
+.B MADV_POPULATE
+fails if there is a proplem populating page tables.
+.B MADV_POPULATE
+simulates user space access to all pages in the range without actually
+reading/writing the pages.
+For private, writable mappings, simulate a write access; for all other
+mappings, simulate a read access.
+.IP
+If
+.B MADV_POPULATE
+succeeds, all page tables have been populated (prefaulted) once.
+If
+.B MADV_POPULATE
+fails, some page tables might have been populated.
+.IP
+.B MADV_POPULATE
+cannot be applied to
+.B PROT_NONE
+and special mappings marked with the kernel-internal
+.B VM_PFNMAP
+and
+.BR VM_IO .
+.IP
+Note that
+.B MADV_POPULATE
+will ignore any poisoned pages in the range.
+Similar to
+.BR MAP_POPULATE ,
+it cannot protect from the OOM (Out Of Memory) handler killing the process.
 .SH RETURN VALUE
 On success,
 .BR madvise ()
@@ -533,6 +575,17 @@ or
 .BR VM_PFNMAP
 ranges.
 .TP
+.B EINVAL
+.I advice
+is
+.BR MADV_POPULATE ,
+but the specified address range includes
+.BR PROT_NONE ,
+.B VM_IO
+or
+.B VM_PFNMAP
+ranges.
+.TP
 .B EIO
 (for
 .BR MADV_WILLNEED )
@@ -548,6 +601,12 @@ Not enough memory: paging in failed.
 Addresses in the specified range are not currently
 mapped, or are outside the address space of the process.
 .TP
+.B ENOMEM
+.I advice
+is
+.BR MADV_POPULATE ,
+but populating (prefaulting) page tables failed.
+.TP
 .B EPERM
 .I advice
 is
-- 
2.29.2