[PATCH v2 4/8] linux: Add MAP_DROPPABLE from Linux 6.11

[H.J. Lu]

On Wed, Oct 9, 2024 at 9:37 PM Adhemerval Zanella Netto
<adhemerval.zanella@linaro.org> wrote:
>
>
>
> On 09/10/24 08:48, H.J. Lu wrote:
> >
> > On Wed, Oct 9, 2024, 7:45 PM Adhemerval Zanella Netto <adhemerval.zanella@linaro.org <mailto:adhemerval.zanella@linaro.org>> wrote:
> >
> >
> >
> >     On 08/10/24 17:28, H.J. Lu wrote:
> >     > On Wed, Oct 9, 2024 at 2:47 AM Adhemerval Zanella
> >     > <adhemerval.zanella@linaro.org <mailto:adhemerval.zanella@linaro.org>> wrote:
> >     >>
> >     >> This request the page to be never written out to swap, it will be zeroed
> >     >> under memory pressure (so kernel can just drop the page), it is inherited
> >     >> by fork, it is not counted against @code{mlock} budget, and if there is
> >     >> no enough memory to service a page faults there is no fatal error (so not
> >     >> signal is sent).
> >     >>
> >     >> Tested with build-many-glibcs.py.
> >     >> ---
> >     >>  manual/llio.texi                           | 9 +++++++++
> >     >>  sysdeps/unix/sysv/linux/bits/mman-linux.h  | 1 +
> >     >>  sysdeps/unix/sysv/linux/tst-mman-consts.py | 2 +-
> >     >>  3 files changed, 11 insertions(+), 1 deletion(-)
> >     >>
> >     >> diff --git a/manual/llio.texi b/manual/llio.texi
> >     >> index a035c3e20f..05ab44c6e7 100644
> >     >> --- a/manual/llio.texi
> >     >> +++ b/manual/llio.texi
> >     >> @@ -1762,6 +1762,15 @@ write dirty metadata out whenever dirty data is written out.  Unlike
> >     >>  most other flags, this one will fail unless @code{MAP_SHARED_VALIDATE}
> >     >>  is also given.
> >     >>
> >     >> +@item MAP_DROPPABLE
> >     >> +Request the page to be never written out to swap, it will be zeroed
> >     >> +under memory pressure (so kernel can just drop the page), it is inherited
> >     >> +by fork, it is not counted against @code{mlock} budget, and if there is
> >     >> +not enough memory to service a page fault there is no fatal error (so no
> >     >> +signal is sent).
> >     >> +
> >     >> +The @code{MAP_DROPPABLE} flag is specific to Linux.
> >     >> +
> >     >>  @end vtable
> >     >>
> >     >>  @code{mmap} returns the address of the new mapping, or
> >     >> diff --git a/sysdeps/unix/sysv/linux/bits/mman-linux.h b/sysdeps/unix/sysv/linux/bits/mman-linux.h
> >     >> index 522333c50a..161a885096 100644
> >     >> --- a/sysdeps/unix/sysv/linux/bits/mman-linux.h
> >     >> +++ b/sysdeps/unix/sysv/linux/bits/mman-linux.h
> >     >> @@ -43,6 +43,7 @@
> >     >>  #define MAP_PRIVATE    0x02            /* Changes are private.  */
> >     >>  #define MAP_SHARED_VALIDATE    0x03    /* Share changes and validate
> >     >>                                            extension flags.  */
> >     >> +#define MAP_DROPPABLE  0x08            /* Zero memory under memory pressure.  */
> >     >>  #define MAP_TYPE       0x0f            /* Mask for type of mapping.  */
> >     >>
> >     >>  /* Other flags.  */
> >     >> diff --git a/sysdeps/unix/sysv/linux/tst-mman-consts.py b/sysdeps/unix/sysv/linux/tst-mman-consts.py
> >     >> index a1137eb1d5..2191bea36f 100644
> >     >> --- a/sysdeps/unix/sysv/linux/tst-mman-consts.py
> >     >> +++ b/sysdeps/unix/sysv/linux/tst-mman-consts.py
> >     >> @@ -33,7 +33,7 @@ def main():
> >     >>                          help='C compiler (including options) to use')
> >     >>      args = parser.parse_args()
> >     >>      linux_version_headers = glibcsyscalls.linux_kernel_version(args.cc)
> >     >> -    linux_version_glibc = (6, 10)
> >     >> +    linux_version_glibc = (6, 11)
> >     >>      sys.exit(glibcextract.compare_macro_consts(
> >     >>          '#define _GNU_SOURCE 1\n'
> >     >>          '#include <sys/mman.h>\n',
> >     >> --
> >     >> 2.43.0
> >     >>
> >     >
> >     > We should add a test for it.
> >     >
> >
> >     I think it would make sense when glibc itself would use this new flag, this patch
> >     is just the usual sync one that just update glibc when a kernel release.
> >
> >
> > mmap is a special case.   A test is desirable.
>
> I am not sure if we should do functional kernel testing in this case, and the VM_DROPPABLE
> test in also tricky because to fully tests requires to apply some system memory pressure
> (so the page is dropped).
>

We have run into the MAP_XXX problem before.  We should add a test
for each new MAP_XXX.  Ther kernel test, tools/testing/selftests/mm/droppable.c,
looks reasonable for glibc.

-- 
H.J.