Re: [PATCH 12/13] writeback: add a writeback iterator

[Brian Foster]

On Mon, Feb 05, 2024 at 10:33:52AM -0500, Brian Foster wrote:
> On Sat, Feb 03, 2024 at 08:11:46AM +0100, Christoph Hellwig wrote:
> > Refactor the code left in write_cache_pages into an iterator that the
> > file system can call to get the next folio for a writeback operation:
> > 
> > 	struct folio *folio = NULL;
> > 
> > 	while ((folio = writeback_iter(mapping, wbc, folio, &error))) {
> > 		error = <do per-foli writeback>;
> > 	}
> > 
> > The twist here is that the error value is passed by reference, so that
> > the iterator can restore it when breaking out of the loop.
> > 
> > Handling of the magic AOP_WRITEPAGE_ACTIVATE value stays outside the
> > iterator and needs is just kept in the write_cache_pages legacy wrapper.
> > in preparation for eventually killing it off.
> > 
> > Heavily based on a for_each* based iterator from Matthew Wilcox.
> > 
> > Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> > ---
> >  include/linux/writeback.h |   4 +
> >  mm/page-writeback.c       | 192 ++++++++++++++++++++++----------------
> >  2 files changed, 118 insertions(+), 78 deletions(-)
> > 
> ...
> > diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> > index 3abb053e70580e..5fe4cdb7dbd61a 100644
> > --- a/mm/page-writeback.c
> > +++ b/mm/page-writeback.c
> ...
> > @@ -2434,69 +2434,68 @@ static struct folio *writeback_get_folio(struct address_space *mapping,
> >  }
> >  
> >  /**
> ...
> >   */
> > -int write_cache_pages(struct address_space *mapping,
> > -		      struct writeback_control *wbc, writepage_t writepage,
> > -		      void *data)
> > +struct folio *writeback_iter(struct address_space *mapping,
> > +		struct writeback_control *wbc, struct folio *folio, int *error)
> >  {
> ...
> > +	} else {
> >  		wbc->nr_to_write -= folio_nr_pages(folio);
> >  
> > -		if (error == AOP_WRITEPAGE_ACTIVATE) {
> > -			folio_unlock(folio);
> > -			error = 0;
> > -		}
> > +		WARN_ON_ONCE(*error > 0);
> 
> Why the warning on writeback error here? It looks like new behavior, but
> maybe I missed something. Otherwise the factoring LGTM.

Err, sorry.. I glossed over the > 0 check and read it as < 0.
Disregard, this seems reasonable to me as long as we no longer expect
those AOP returns (which I'm not really clear on either, but anyways..):

Reviewed-by: Brian Foster <bfoster@xxxxxxxxxx>

> 
> Brian
> 
> >  
> >  		/*
> >  		 * For integrity writeback we have to keep going until we have
> > @@ -2510,33 +2509,70 @@ int write_cache_pages(struct address_space *mapping,
> >  		 * wbc->nr_to_write or encounter the first error.
> >  		 */
> >  		if (wbc->sync_mode == WB_SYNC_ALL) {
> > -			if (error && !ret)
> > -				ret = error;
> > +			if (*error && !wbc->saved_err)
> > +				wbc->saved_err = *error;
> >  		} else {
> > -			if (error || wbc->nr_to_write <= 0)
> > +			if (*error || wbc->nr_to_write <= 0)
> >  				goto done;
> >  		}
> >  	}
> >  
> > -	/*
> > -	 * For range cyclic writeback we need to remember where we stopped so
> > -	 * that we can continue there next time we are called.  If  we hit the
> > -	 * last page and there is more work to be done, wrap back to the start
> > -	 * of the file.
> > -	 *
> > -	 * For non-cyclic writeback we always start looking up at the beginning
> > -	 * of the file if we are called again, which can only happen due to
> > -	 * -ENOMEM from the file system.
> > -	 */
> > -	folio_batch_release(&wbc->fbatch);
> > -	if (wbc->range_cyclic)
> > -		mapping->writeback_index = 0;
> > -	return ret;
> > +	folio = writeback_get_folio(mapping, wbc);
> > +	if (!folio) {
> > +		/*
> > +		 * To avoid deadlocks between range_cyclic writeback and callers
> > +		 * that hold pages in PageWriteback to aggregate I/O until
> > +		 * the writeback iteration finishes, we do not loop back to the
> > +		 * start of the file.  Doing so causes a page lock/page
> > +		 * writeback access order inversion - we should only ever lock
> > +		 * multiple pages in ascending page->index order, and looping
> > +		 * back to the start of the file violates that rule and causes
> > +		 * deadlocks.
> > +		 */
> > +		if (wbc->range_cyclic)
> > +			mapping->writeback_index = 0;
> > +
> > +		/*
> > +		 * Return the first error we encountered (if there was any) to
> > +		 * the caller.
> > +		 */
> > +		*error = wbc->saved_err;
> > +	}
> > +	return folio;
> >  
> >  done:
> >  	folio_batch_release(&wbc->fbatch);
> >  	if (wbc->range_cyclic)
> >  		mapping->writeback_index = folio->index + folio_nr_pages(folio);
> > +	return NULL;
> > +}
> > +
> > +/**
> > + * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
> > + * @mapping: address space structure to write
> > + * @wbc: subtract the number of written pages from *@wbc->nr_to_write
> > + * @writepage: function called for each page
> > + * @data: data passed to writepage function
> > + *
> > + * Return: %0 on success, negative error code otherwise
> > + *
> > + * Note: please use writeback_iter() instead.
> > + */
> > +int write_cache_pages(struct address_space *mapping,
> > +		      struct writeback_control *wbc, writepage_t writepage,
> > +		      void *data)
> > +{
> > +	struct folio *folio = NULL;
> > +	int error;
> > +
> > +	while ((folio = writeback_iter(mapping, wbc, folio, &error))) {
> > +		error = writepage(folio, wbc, data);
> > +		if (error == AOP_WRITEPAGE_ACTIVATE) {
> > +			folio_unlock(folio);
> > +			error = 0;
> > +		}
> > +	}
> > +
> >  	return error;
> >  }
> >  EXPORT_SYMBOL(write_cache_pages);
> > -- 
> > 2.39.2
> > 
> > 
> 
>