On 07/23/2015 01:49 AM, David Rientjes wrote:
> On Wed, 22 Jul 2015, Mike Kravetz wrote:
>
>> On 07/21/2015 05:41 PM, David Rientjes wrote:
>>> munmap(2) will fail with an errno of EINVAL for hugetlb memory if the
>>> length is not a multiple of the underlying page size.
>>>
>>> Documentation/vm/hugetlbpage.txt was updated to specify this behavior
>>> since Linux 4.1 in commit 80d6b94bd69a ("mm, doc: cleanup and clarify
>>> munmap behavior for hugetlb memory").
>>>
>>> Signed-off-by: David Rientjes <rientjes@xxxxxxxxxx>
>>> ---
>>> man2/mmap.2 | 4 ++++
>>> 1 file changed, 4 insertions(+)
>>>
>>> diff --git a/man2/mmap.2 b/man2/mmap.2
>>> --- a/man2/mmap.2
>>> +++ b/man2/mmap.2
>>> @@ -383,6 +383,10 @@ All pages containing a part
>>> of the indicated range are unmapped, and subsequent references
>>> to these pages will generate
>>> .BR SIGSEGV .
>>> +An exception is when the underlying memory is not of the native page
>>> +size, such as hugetlb page sizes, whereas
>>> +.I length
>>> +must be a multiple of the underlying page size.
>>> It is not an error if the
>>> indicated range does not contain any mapped pages.
>>> .SS Timestamps changes for file-backed mappings
>>>
>>> --
>>
>> Should we also add a similar comment for the mmap offset? Currently
>> the man page says:
>>
>> "offset must be a multiple of the page size as returned by
>> sysconf(_SC_PAGE_SIZE)."
>>
>> For hugetlbfs, I beieve the offset must be a multiple of the
>> hugetlb page size. A similar comment/exception about using
>> the "underlying page size" would apply here as well.
>>
>
> Yes, that makes sense, thanks. We should also explicitly say that mmap(2)
> automatically aligns length to be hugepage aligned if backed by hugetlbfs.
And, surely, it also does something similar for mmap()'s 'addr'
argument?
I suggest we add a subsection to describe the HugeTLB differences. How
about something like:
Huge page (Huge TLB) mappings
For mappings that employ huge pages, the requirements for the
arguments of mmap() and munmap() differ somewhat from the
requirements for mappings that use the native system page size.
For mmap(), offset must be a multiple of the underlying huge page
size. The system automatically aligns length to be a multiple of
the underlying huge page size.
For munmap(), addr and length must both be a multiple of the
underlying huge page size.
?
Thanks,
Michael
--
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
