[PATCH] madvise.2: Clarify addr/length and update hugetlb support

[Mike Kravetz <mike.kravetz@oracle.com>]

Clarify that madvise only works on full pages, and remove references
to 'bytes'.

Update MADV_DONTNEED and MADV_REMOVE sections to remove notes that
HugeTLB mappings are not supported.  They now are supported.

Under 'Linux notes' describe addr requirements and length handling
for ranges in HugeTLB mappings.

Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
---
 man2/madvise.2 | 36 ++++++++++++++++++++++++++----------
 1 file changed, 26 insertions(+), 10 deletions(-)

diff --git a/man2/madvise.2 b/man2/madvise.2
index f1f384c0c..c3b0615cb 100644
--- a/man2/madvise.2
+++ b/man2/madvise.2
@@ -61,9 +61,13 @@ system call is used to give advice or directions to the kernel
 about the address range beginning at address
 .I addr
 and with size
+.IR length .
+.BR madvise ()
+only operates on whole pages, therefore
+.I addr
+must be page-aligned.  The value of
 .I length
-bytes
-In most cases,
+is rounded up to a multiple of page size.  In most cases,
 the goal of such advice is to improve system or application performance.
 .PP
 Initially, the system call supported a set of "conventional"
@@ -143,7 +147,7 @@ The resident set size (RSS) of the calling process will be immediately
 reduced however.
 .IP
 .B MADV_DONTNEED
-cannot be applied to locked pages, Huge TLB pages, or
+cannot be applied to locked pages, or
 .BR VM_PFNMAP
 pages.
 (Pages marked with the kernel-internal
@@ -170,24 +174,24 @@ Note that some of these operations change the semantics of memory accesses.
 .\" commit f6b3ec238d12c8cc6cc71490c6e3127988460349
 Free up a given range of pages
 and its associated backing store.
-This is equivalent to punching a hole in the corresponding byte
+This is equivalent to punching a hole in the corresponding
 range of the backing store (see
 .BR fallocate (2)).
 Subsequent accesses in the specified address range will see
-bytes containing zero.
+pages containing zero.
 .\" Databases want to use this feature to drop a section of their
 .\" bufferpool (shared memory segments) - without writing back to
 .\" disk/swap space.  This feature is also useful for supporting
 .\" hot-plug memory on UML.
 .IP
 The specified address range must be mapped shared and writable.
-This flag cannot be applied to locked pages, Huge TLB pages, or
+This flag cannot be applied to locked pages, or
 .BR VM_PFNMAP
 pages.
 .IP
 In the initial implementation, only
 .BR tmpfs (5)
-was supported
+supported
 .BR MADV_REMOVE ;
 but since Linux 3.5,
 .\" commit 3f31d07571eeea18a7d34db9af21d2285b807a17
@@ -196,9 +200,9 @@ any filesystem which supports the
 .BR FALLOC_FL_PUNCH_HOLE
 mode also supports
 .BR MADV_REMOVE .
-Hugetlbfs fails with the error
-.BR EINVAL
-and other filesystems fail with the error
+Filesystems which do not support
+.BR MADV_REMOVE
+fail with the error
 .BR EOPNOTSUPP .
 .TP
 .BR MADV_DONTFORK " (since Linux 2.6.16)"
@@ -596,6 +600,18 @@ that are not mapped, the Linux version of
 ignores them and applies the call to the rest (but returns
 .B ENOMEM
 from the system call, as it should).
+.PP
+If the specified address
+.I addr
+is within a mapping backed by Huge TLB pages, then
+.I addr
+must be aligned to the underlying Huge TLB page size.  If the range
+specified by
+.I addr
+and
+.I length
+ends in a mapping backed by Huge TLB pages, then the end of the range
+will be rounded up to a multiple of the underlying Huge TLB page size.
 .\" .SH HISTORY
 .\" The
 .\" .BR madvise ()
-- 
2.35.3