On Wed, 2019-11-06 at 13:03 +0100, Jan Kara wrote:
> On Tue 05-11-19 13:22:11, Johannes Weiner wrote:
> > What I don't quite understand yet is why the fault path doesn't
> > make
> > progress eventually. We must drop the mmap_sem without changing the
> > state in any way. How can we keep looping on the same page?
>
> That may be a slight suboptimality with Josef's patches. If the page
> is marked as PageReadahead, we always drop mmap_sem if we can and
> start
> readahead without checking whether that makes sense or not in
> do_async_mmap_readahead(). OTOH page_cache_async_readahead() then
> clears
> PageReadahead so the only way how I can see we could loop like this
> is when
> file->ra->ra_pages is 0. Not sure if that's what's happening through.
> We'd
> need to find which of the paths in filemap_fault() calls
> maybe_unlock_mmap_for_io() to tell more.

Yes, ra_pages==0
Attached the dmesg + smaps outputs


[   91.803976] _mm_populate 0 7ffffffff000 7ffffffff000 1 LOOP
5637e3a2b000 1 0
[   91.803978] _mm_populate 0 7ffffffff000 7ffffffff000 1 vma-
>vm_start=7fa32afe5000 vma->vm_end=7fa32b00a000 vma->vm_flags=8002071
[   91.803980] _mm_populate 0 7ffffffff000 7ffffffff000 1
nend=7fa32b00a000 7ffffffff000 7fa32b00a000
[   91.803981] __get_user_pages start=7fa32afe5000 nr_pages=37
gup_flags=1052 ctx.page_mask=0
[   91.803982] __get_user_pages @1 vma==NULL
[   91.803984] __get_user_pages @2 start=7fa32afe5000 vma-
>vm_start=7fa32afe5000 vma->vm_end=7fa32b00a000 vma->vm_flags=8002071
[   91.803986] __get_user_pages @2 vma->vm_file->f_mode=80a801d
ra.async_size=0 ra.ra_pages=0 ra.mmap_miss=0 ra.size=0 ra.start=0
ra.prev_pos=948
[   91.803988] __get_user_pages @7 follow_page_mask -->
1  ctx.page_mask=0
[   91.803989] __get_user_pages @9 page_increm=1 ctx.page_mask=0
[   91.803990] __get_user_pages @10 i=1 start=7fa32afe6000 nr_pages=36
[   91.803991] __get_user_pages @7 follow_page_mask -->
0  ctx.page_mask=0
[   91.803996] faultin_page handle_mm_fault --> ret = 1024
[   91.803997] faultin_page-->EBUSY VM_FAULT_RETRY non-blocking?1
FAULT_FLAG_RETRY_NOWAIT?0
[   91.803998] __get_user_pages @8 faultin_page --> -16
[   91.803999] __get_user_pages LEAVE i=1 ret=0
[   91.804001] _mm_populate 0 7ffffffff000 7ffffffff000 1 LOOP-3 ret=1
nend=7fa32afe6000
[   91.804002] _mm_populate 0 7ffffffff000 7ffffffff000 1 LOOP
7fa32afe6000 0 0
[   91.804004] _mm_populate 0 7ffffffff000 7ffffffff000 1 vma-
>vm_start=7fa32afe5000 vma->vm_end=7fa32b00a000 vma->vm_flags=8002071
[   91.804006] _mm_populate 0 7ffffffff000 7ffffffff000 1
nend=7fa32b00a000 7ffffffff000 7fa32b00a000
[   91.804007] __get_user_pages start=7fa32afe6000 nr_pages=36
gup_flags=1052 ctx.page_mask=0
[   91.804008] __get_user_pages @1 vma==NULL
[   91.804010] __get_user_pages @2 start=7fa32afe6000 vma-
>vm_start=7fa32afe5000 vma->vm_end=7fa32b00a000 vma->vm_flags=8002071
[   91.804012] __get_user_pages @2 vma->vm_file->f_mode=80a801d
ra.async_size=0 ra.ra_pages=0 ra.mmap_miss=0 ra.size=0 ra.start=0
ra.prev_pos=948
[   91.804013] __get_user_pages @7 follow_page_mask -->
0  ctx.page_mask=0
[   91.804016] faultin_page handle_mm_fault --> ret = 1024
[   91.804017] faultin_page-->EBUSY VM_FAULT_RETRY non-blocking?1
FAULT_FLAG_RETRY_NOWAIT?0
[   91.804018] __get_user_pages @8 faultin_page --> -16
[   91.804019] __get_user_pages LEAVE i=0 ret=0
[   91.804020] _mm_populate 0 7ffffffff000 7ffffffff000 1 LOOP-3 ret=0
nend=7fa32afe6000