| Message ID | 20210615162342.1669332-4-willy@infradead.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Further set_page_dirty cleanups | expand |
On Tue, Jun 15, 2021 at 05:23:39PM +0100, Matthew Wilcox (Oracle) wrote: > The only difference between iomap_set_page_dirty() and > __set_page_dirty_nobuffers() is that the latter includes a debugging > check that a !Uptodate page has private data. Looks good, Reviewed-by: Christoph Hellwig <hch@lst.de>
On Tue, Jun 15, 2021 at 05:23:39PM +0100, Matthew Wilcox (Oracle) wrote: > The only difference between iomap_set_page_dirty() and > __set_page_dirty_nobuffers() is that the latter includes a debugging > check that a !Uptodate page has private data. > > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> > --- > fs/gfs2/aops.c | 2 +- > fs/iomap/buffered-io.c | 27 +-------------------------- > fs/xfs/xfs_aops.c | 2 +- > fs/zonefs/super.c | 2 +- > include/linux/iomap.h | 1 - > 5 files changed, 4 insertions(+), 30 deletions(-) > > diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c > index 50dd1771d00c..746b78c3a91d 100644 > --- a/fs/gfs2/aops.c > +++ b/fs/gfs2/aops.c > @@ -784,7 +784,7 @@ static const struct address_space_operations gfs2_aops = { > .writepages = gfs2_writepages, > .readpage = gfs2_readpage, > .readahead = gfs2_readahead, > - .set_page_dirty = iomap_set_page_dirty, > + .set_page_dirty = __set_page_dirty_nobuffers, Using __ functions in structures in different modules feels odd to me. Why not just have iomap_set_page_dirty be a #define to this function now if you want to do this? Or take the __ off of the function name? Anyway, logic here is fine, but feels odd. greg k-h
On Tue, Jun 15, 2021 at 07:19:59PM +0200, Greg Kroah-Hartman wrote: > On Tue, Jun 15, 2021 at 05:23:39PM +0100, Matthew Wilcox (Oracle) wrote: > Using __ functions in structures in different modules feels odd to me. > Why not just have iomap_set_page_dirty be a #define to this function now > if you want to do this? > > Or take the __ off of the function name? > > Anyway, logic here is fine, but feels odd. heh, that was how I did it the first time. Then I thought that it was better to follow Christoph's patch: static const struct address_space_operations adfs_aops = { + .set_page_dirty = __set_page_dirty_buffers, (etc)
On Tue, Jun 15, 2021 at 06:32:37PM +0100, Matthew Wilcox wrote: > On Tue, Jun 15, 2021 at 07:19:59PM +0200, Greg Kroah-Hartman wrote: > > On Tue, Jun 15, 2021 at 05:23:39PM +0100, Matthew Wilcox (Oracle) wrote: > > Using __ functions in structures in different modules feels odd to me. > > Why not just have iomap_set_page_dirty be a #define to this function now > > if you want to do this? > > > > Or take the __ off of the function name? > > > > Anyway, logic here is fine, but feels odd. > > heh, that was how I did it the first time. Then I thought that it was > better to follow Christoph's patch: > > static const struct address_space_operations adfs_aops = { > + .set_page_dirty = __set_page_dirty_buffers, > (etc) Eventually everything around set_page_dirty should be changed to operate on folios, and that will be a good time to come up with a sane naming scheme without introducing extra churn.
On Tue, Jun 15, 2021 at 07:34:53PM +0200, Christoph Hellwig wrote: > On Tue, Jun 15, 2021 at 06:32:37PM +0100, Matthew Wilcox wrote: > > On Tue, Jun 15, 2021 at 07:19:59PM +0200, Greg Kroah-Hartman wrote: > > > On Tue, Jun 15, 2021 at 05:23:39PM +0100, Matthew Wilcox (Oracle) wrote: > > > Using __ functions in structures in different modules feels odd to me. > > > Why not just have iomap_set_page_dirty be a #define to this function now > > > if you want to do this? > > > > > > Or take the __ off of the function name? > > > > > > Anyway, logic here is fine, but feels odd. > > > > heh, that was how I did it the first time. Then I thought that it was > > better to follow Christoph's patch: > > > > static const struct address_space_operations adfs_aops = { > > + .set_page_dirty = __set_page_dirty_buffers, > > (etc) > > Eventually everything around set_page_dirty should be changed to operate > on folios, and that will be a good time to come up with a sane > naming scheme without introducing extra churn. Ok, that's fine, I don't normally touch these files, so it's not an issue for me :)
On Tue, Jun 15, 2021 at 07:34:53PM +0200, Christoph Hellwig wrote: > On Tue, Jun 15, 2021 at 06:32:37PM +0100, Matthew Wilcox wrote: > > On Tue, Jun 15, 2021 at 07:19:59PM +0200, Greg Kroah-Hartman wrote: > > > On Tue, Jun 15, 2021 at 05:23:39PM +0100, Matthew Wilcox (Oracle) wrote: > > > Using __ functions in structures in different modules feels odd to me. > > > Why not just have iomap_set_page_dirty be a #define to this function now > > > if you want to do this? > > > > > > Or take the __ off of the function name? > > > > > > Anyway, logic here is fine, but feels odd. > > > > heh, that was how I did it the first time. Then I thought that it was > > better to follow Christoph's patch: > > > > static const struct address_space_operations adfs_aops = { > > + .set_page_dirty = __set_page_dirty_buffers, > > (etc) > > Eventually everything around set_page_dirty should be changed to operate > on folios, and that will be a good time to come up with a sane > naming scheme without introducing extra churn. The way it currently looks in my tree ... set_page_dirty(page) is a thin wrapper that calls folio_mark_dirty(folio). folio_mark_dirty() calls a_ops->dirty_folio(mapping, folio) (which returns bool). __set_page_dirty_nobuffers() becomes filemap_dirty_folio() __set_page_dirty_buffers() becomes block_dirty_folio() __set_page_dirty_no_writeback() becomes dirty_folio_no_writeback() Now I look at it, maybe that last should be nowb_dirty_folio().
On Tue, Jun 15, 2021 at 07:13:16PM +0100, Matthew Wilcox wrote: > On Tue, Jun 15, 2021 at 07:34:53PM +0200, Christoph Hellwig wrote: > > On Tue, Jun 15, 2021 at 06:32:37PM +0100, Matthew Wilcox wrote: > > > On Tue, Jun 15, 2021 at 07:19:59PM +0200, Greg Kroah-Hartman wrote: > > > > On Tue, Jun 15, 2021 at 05:23:39PM +0100, Matthew Wilcox (Oracle) wrote: > > > > Using __ functions in structures in different modules feels odd to me. > > > > Why not just have iomap_set_page_dirty be a #define to this function now > > > > if you want to do this? > > > > > > > > Or take the __ off of the function name? > > > > > > > > Anyway, logic here is fine, but feels odd. > > > > > > heh, that was how I did it the first time. Then I thought that it was > > > better to follow Christoph's patch: > > > > > > static const struct address_space_operations adfs_aops = { > > > + .set_page_dirty = __set_page_dirty_buffers, > > > (etc) > > > > Eventually everything around set_page_dirty should be changed to operate > > on folios, and that will be a good time to come up with a sane > > naming scheme without introducing extra churn. > > The way it currently looks in my tree ... > > set_page_dirty(page) is a thin wrapper that calls folio_mark_dirty(folio). > folio_mark_dirty() calls a_ops->dirty_folio(mapping, folio) (which > returns bool). > __set_page_dirty_nobuffers() becomes filemap_dirty_folio() > __set_page_dirty_buffers() becomes block_dirty_folio() > __set_page_dirty_no_writeback() becomes dirty_folio_no_writeback() > > Now I look at it, maybe that last should be nowb_dirty_folio(). Not to be a pain, but you are mixing "folio" at the front and back of the api name? We messed up in the driver core with this for some things (get_device() being one), I would recommend just sticking with one naming scheme now as you are getting to pick what you want to use. So perhaps for the above: folio_mark_dirty() folio_dirty_no_writeback() folio_dirty_filemap() folio_dirty_block() much like "set_page" is used today? Anyway, just bikeshedding, it's your code, your choice :) thanks for doing this work overall. greg k-h
On Wed, Jun 16, 2021 at 08:50:40AM +0200, Greg Kroah-Hartman wrote: > On Tue, Jun 15, 2021 at 07:13:16PM +0100, Matthew Wilcox wrote: > > On Tue, Jun 15, 2021 at 07:34:53PM +0200, Christoph Hellwig wrote: > > > Eventually everything around set_page_dirty should be changed to operate > > > on folios, and that will be a good time to come up with a sane > > > naming scheme without introducing extra churn. > > > > The way it currently looks in my tree ... > > > > set_page_dirty(page) is a thin wrapper that calls folio_mark_dirty(folio). > > folio_mark_dirty() calls a_ops->dirty_folio(mapping, folio) (which > > returns bool). > > __set_page_dirty_nobuffers() becomes filemap_dirty_folio() > > __set_page_dirty_buffers() becomes block_dirty_folio() > > __set_page_dirty_no_writeback() becomes dirty_folio_no_writeback() > > > > Now I look at it, maybe that last should be nowb_dirty_folio(). > > Not to be a pain, but you are mixing "folio" at the front and back of > the api name? We messed up in the driver core with this for some things > (get_device() being one), I would recommend just sticking with one > naming scheme now as you are getting to pick what you want to use. That is mostly what I'm doing. eg, get_page -> folio_get lock_page -> folio_lock PageUptodate -> folio_uptodate set_page_dirty -> folio_mark_dirty What I haven't dealt with yet is the naming of the address_space_operations. My thinking with those is that they should be verb_folio, since they _aren't_ the functions that get called. ie it looks like this: folio_mark_dirty() aops->dirty_folio() ext4_dirty_folio() buffer_dirty_folio() I actually see the inconsistency here as a good thing -- these are implementations of the aop, so foo_verb_folio() means you're doing something weird and internal instead of going through the vfs/mm. That implies doing things like renaming ->readpage to ->read_folio, but if we're changing the API from passing a struct page to a struct folio, that can all be done at the same time with no additional disruption.
On Wed, Jun 16, 2021 at 05:28:14PM +0100, Matthew Wilcox wrote: > On Wed, Jun 16, 2021 at 08:50:40AM +0200, Greg Kroah-Hartman wrote: > > On Tue, Jun 15, 2021 at 07:13:16PM +0100, Matthew Wilcox wrote: > > > On Tue, Jun 15, 2021 at 07:34:53PM +0200, Christoph Hellwig wrote: > > > > Eventually everything around set_page_dirty should be changed to operate > > > > on folios, and that will be a good time to come up with a sane > > > > naming scheme without introducing extra churn. > > > > > > The way it currently looks in my tree ... > > > > > > set_page_dirty(page) is a thin wrapper that calls folio_mark_dirty(folio). > > > folio_mark_dirty() calls a_ops->dirty_folio(mapping, folio) (which > > > returns bool). > > > __set_page_dirty_nobuffers() becomes filemap_dirty_folio() > > > __set_page_dirty_buffers() becomes block_dirty_folio() > > > __set_page_dirty_no_writeback() becomes dirty_folio_no_writeback() > > > > > > Now I look at it, maybe that last should be nowb_dirty_folio(). > > > > Not to be a pain, but you are mixing "folio" at the front and back of > > the api name? We messed up in the driver core with this for some things > > (get_device() being one), I would recommend just sticking with one > > naming scheme now as you are getting to pick what you want to use. > > That is mostly what I'm doing. eg, > > get_page -> folio_get > lock_page -> folio_lock > PageUptodate -> folio_uptodate > set_page_dirty -> folio_mark_dirty Nice. > What I haven't dealt with yet is the naming of the > address_space_operations. My thinking with those is that they should > be verb_folio, since they _aren't_ the functions that get called. > ie it looks like this: > > folio_mark_dirty() > aops->dirty_folio() > ext4_dirty_folio() > buffer_dirty_folio() > > I actually see the inconsistency here as a good thing -- these are > implementations of the aop, so foo_verb_folio() means you're doing > something weird and internal instead of going through the vfs/mm. > > That implies doing things like renaming ->readpage to ->read_folio, but > if we're changing the API from passing a struct page to a struct folio, > that can all be done at the same time with no additional disruption. Ok, as long as there's a reason for the naming scheme, I'm happy as hopefully it will make sense to others as well. thanks, greg k-h
diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c index 50dd1771d00c..746b78c3a91d 100644 --- a/fs/gfs2/aops.c +++ b/fs/gfs2/aops.c @@ -784,7 +784,7 @@ static const struct address_space_operations gfs2_aops = { .writepages = gfs2_writepages, .readpage = gfs2_readpage, .readahead = gfs2_readahead, - .set_page_dirty = iomap_set_page_dirty, + .set_page_dirty = __set_page_dirty_nobuffers, .releasepage = iomap_releasepage, .invalidatepage = iomap_invalidatepage, .bmap = gfs2_bmap, diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c index 2bf4778f2098..41da4f14c00b 100644 --- a/fs/iomap/buffered-io.c +++ b/fs/iomap/buffered-io.c @@ -640,31 +640,6 @@ iomap_write_begin(struct inode *inode, loff_t pos, unsigned len, unsigned flags, return status; } -int -iomap_set_page_dirty(struct page *page) -{ - struct address_space *mapping = page_mapping(page); - int newly_dirty; - - if (unlikely(!mapping)) - return !TestSetPageDirty(page); - - /* - * Lock out page's memcg migration to keep PageDirty - * synchronized with per-memcg dirty page counters. - */ - lock_page_memcg(page); - newly_dirty = !TestSetPageDirty(page); - if (newly_dirty) - __set_page_dirty(page, mapping, 0); - unlock_page_memcg(page); - - if (newly_dirty) - __mark_inode_dirty(mapping->host, I_DIRTY_PAGES); - return newly_dirty; -} -EXPORT_SYMBOL_GPL(iomap_set_page_dirty); - static size_t __iomap_write_end(struct inode *inode, loff_t pos, size_t len, size_t copied, struct page *page) { @@ -684,7 +659,7 @@ static size_t __iomap_write_end(struct inode *inode, loff_t pos, size_t len, if (unlikely(copied < len && !PageUptodate(page))) return 0; iomap_set_range_uptodate(page, offset_in_page(pos), len); - iomap_set_page_dirty(page); + __set_page_dirty_nobuffers(page); return copied; } diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index 826caa6b4a5a..a335d79dcff8 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -561,7 +561,7 @@ const struct address_space_operations xfs_address_space_operations = { .readahead = xfs_vm_readahead, .writepage = xfs_vm_writepage, .writepages = xfs_vm_writepages, - .set_page_dirty = iomap_set_page_dirty, + .set_page_dirty = __set_page_dirty_nobuffers, .releasepage = iomap_releasepage, .invalidatepage = iomap_invalidatepage, .bmap = xfs_vm_bmap, diff --git a/fs/zonefs/super.c b/fs/zonefs/super.c index cd145d318b17..3aacf016c7c2 100644 --- a/fs/zonefs/super.c +++ b/fs/zonefs/super.c @@ -185,7 +185,7 @@ static const struct address_space_operations zonefs_file_aops = { .readahead = zonefs_readahead, .writepage = zonefs_writepage, .writepages = zonefs_writepages, - .set_page_dirty = iomap_set_page_dirty, + .set_page_dirty = __set_page_dirty_nobuffers, .releasepage = iomap_releasepage, .invalidatepage = iomap_invalidatepage, .migratepage = iomap_migrate_page, diff --git a/include/linux/iomap.h b/include/linux/iomap.h index c87d0cb0de6d..479c1da3e221 100644 --- a/include/linux/iomap.h +++ b/include/linux/iomap.h @@ -159,7 +159,6 @@ ssize_t iomap_file_buffered_write(struct kiocb *iocb, struct iov_iter *from, const struct iomap_ops *ops); int iomap_readpage(struct page *page, const struct iomap_ops *ops); void iomap_readahead(struct readahead_control *, const struct iomap_ops *ops); -int iomap_set_page_dirty(struct page *page); int iomap_is_partially_uptodate(struct page *page, unsigned long from, unsigned long count); int iomap_releasepage(struct page *page, gfp_t gfp_mask);
The only difference between iomap_set_page_dirty() and __set_page_dirty_nobuffers() is that the latter includes a debugging check that a !Uptodate page has private data. Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> --- fs/gfs2/aops.c | 2 +- fs/iomap/buffered-io.c | 27 +-------------------------- fs/xfs/xfs_aops.c | 2 +- fs/zonefs/super.c | 2 +- include/linux/iomap.h | 1 - 5 files changed, 4 insertions(+), 30 deletions(-)