Re: [Qemu-devel] [PATCH 05/15] postcopy: enhance ram_discard_range for hugepages

[Juan Quintela <quintela@redhat.com>]

"Dr. David Alan Gilbert" <dgilbert@redhat.com> wrote:
> * Juan Quintela (quintela@redhat.com) wrote:
>> "Dr. David Alan Gilbert (git)" <dgilbert@redhat.com> wrote:
>> > From: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
>> >
>> > Unfortunately madvise DONTNEED doesn't work on hugepagetlb
>> > so use fallocate(FALLOC_FL_PUNCH_HOLE)
>> > qemu_fd_getpagesize only sets the page based off a file
>> > if the file is from hugetlbfs.
>> >
>> > Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>

>> > +#if defined(CONFIG_FALLOCATE_PUNCH_HOLE)
>> > +#include <linux/falloc.h>
>> > +#endif

I hate this in generic code :-()





I think that the function will have to be called:

qemu_ram_punch_hole(RAMBblock *, length);

Put it all together at the beggining of the file?
What I don't want is that if someone arrives with a way to do this in
other OS, we need to put yet more ifdefs.  I preffer very much that it
just have to define a function with that semantics.  Agreed that this is
a mess, but I can't think of an easier way of doing it either :-()



>> > @@ -1874,15 +1878,27 @@ int ram_discard_range(MigrationIncomingState *mis,
>> >  
>> >      if ((start + length) <= rb->used_length) {
>> >          uint8_t *host_endaddr = host_startaddr + length;
>> > -        if ((uintptr_t)host_endaddr & (qemu_host_page_size - 1)) {
>> > +        if ((uintptr_t)host_endaddr & (rb->page_size - 1)) {
>> >              error_report("ram_discard_range: Unaligned end address: %p",
>> >                           host_endaddr);
>> >              goto err;
>> >          }
>> > -        errno = ENOTSUP;
>> > +        errno = ENOTSUP; /* If we are missing MADVISE etc */
>> > +
>> > +        if (rb->page_size == qemu_host_page_size) {
>> >  #if defined(CONFIG_MADVISE)
>> > -        ret = qemu_madvise(host_startaddr, length, QEMU_MADV_DONTNEED);
>> > +            ret = qemu_madvise(host_startaddr, length, QEMU_MADV_DONTNEED);
>> >  #endif
>> > +        } else {
>> > +            /* Huge page case  - unfortunately it can't do DONTNEED, but
>> > +             * it can do the equivalent by FALLOC_FL_PUNCH_HOLE in the
>> > +             * huge page file.
>> > +             */
>> > +#ifdef CONFIG_FALLOCATE_PUNCH_HOLE
>> > +            ret = fallocate(rb->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
>> > +                            start, length);
>> > +#endif
>> > +        }
>> >          if (ret) {
>> >              error_report("ram_discard_range: Failed to discard  range "
>> >                           "%s:%" PRIx64 " +%zx (%d)",
>> 
>> Can we move this to qemu-posix or similar?
>> qemu_punch_hole() or similar and just put all the magic there?
>
> I'm trying but it's tricky.
> The problem is that:
>    a) To be able to tell which you need you need the pagesize of the
>      area
>    b) Then you need the fd if it is a hugepage
>      (You can get (a) from (b) by a syscall we already do once)
>    c) If it's normal RAM you need the memory address but...
>    d) If it's a hugepage you need the offset in the file
>
>   which is a mess; so you either have to pass all those parameters,
> or end up passing a RAMBlock* which doesn't feel like it should
> make its way into any of the os-* files.
>
>   I could move it to exec.c that already has some ifdef on OSs;
> what do you think?
>
> Dave
>
>> For the rest, I am ok with it.
> --
> Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK