Re: [PATCH] fs: annotate refault stalls from bdev_read_page

[Minchan Kim <minchan@kernel.org>]

On Wed, Oct 09, 2019 at 03:25:10PM -0700, Shakeel Butt wrote:
> On Wed, Oct 9, 2019 at 2:19 PM Minchan Kim <minchan@kernel.org> wrote:
> >
> > From: Minchan Kim <minchan@google.com>
> >
> > If block device supports rw_page operation, it doesn't submit bio
> > so annotation in submit_bio for refault stall doesn't work.
> > It happens with zram in android, especially swap read path which
> > could consume CPU cycle for decompress.
> 
> What about zswap? Do we need the same in zswap_frontswap_load()?

Yub, it needs it. Maybe, a annotation in swap_readpage will cover both
all at once in real pratice unless we need to take care of nvdimms
which supports rw_page operation.

Thanks.
> 
> >
> > Annotate bdev_read_page() to account the synchronous IO overhead
> > to prevent underreport memory pressure.
> >
> > Cc: Johannes Weiner <hannes@cmpxchg.org>
> > Signed-off-by: Minchan Kim <minchan@google.com>
> > ---
> >  fs/block_dev.c | 13 +++++++++++++
> >  mm/memory.c    |  1 +
> >  2 files changed, 14 insertions(+)
> >
> > diff --git a/fs/block_dev.c b/fs/block_dev.c
> > index 9c073dbdc1b0..82ca28eb9a57 100644
> > --- a/fs/block_dev.c
> > +++ b/fs/block_dev.c
> > @@ -26,6 +26,7 @@
> >  #include <linux/writeback.h>
> >  #include <linux/mpage.h>
> >  #include <linux/mount.h>
> > +#include <linux/psi.h>
> >  #include <linux/pseudo_fs.h>
> >  #include <linux/uio.h>
> >  #include <linux/namei.h>
> > @@ -701,6 +702,8 @@ int bdev_read_page(struct block_device *bdev, sector_t sector,
> >  {
> >         const struct block_device_operations *ops = bdev->bd_disk->fops;
> >         int result = -EOPNOTSUPP;
> > +       unsigned long pflags;
> > +       bool workingset_read;
> >
> >         if (!ops->rw_page || bdev_get_integrity(bdev))
> >                 return result;
> > @@ -708,9 +711,19 @@ int bdev_read_page(struct block_device *bdev, sector_t sector,
> >         result = blk_queue_enter(bdev->bd_queue, 0);
> >         if (result)
> >                 return result;
> > +
> > +       workingset_read = PageWorkingset(page);
> > +       if (workingset_read)
> > +               psi_memstall_enter(&pflags);
> > +
> >         result = ops->rw_page(bdev, sector + get_start_sect(bdev), page,
> >                               REQ_OP_READ);
> > +
> > +       if (workingset_read)
> > +               psi_memstall_leave(&pflags);
> > +
> >         blk_queue_exit(bdev->bd_queue);
> > +
> >         return result;
> >  }
> >  EXPORT_SYMBOL_GPL(bdev_read_page);
> > diff --git a/mm/memory.c b/mm/memory.c
> > index 06935826d71e..6357d5a0a2a5 100644
> > --- a/mm/memory.c
> > +++ b/mm/memory.c
> > @@ -2801,6 +2801,7 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> >                         if (page) {
> >                                 __SetPageLocked(page);
> >                                 __SetPageSwapBacked(page);
> > +                               SetPageWorkingset(page);
> >                                 set_page_private(page, entry.val);
> >                                 lru_cache_add_anon(page);
> >                                 swap_readpage(page, true);
> > --
> > 2.23.0.581.g78d2f28ef7-goog
> >