[PATCH 0/4] put_user_page: new put_user_page_dirty*() helpers

[PATCH 0/4] put_user_page: new put_user_page_dirty*() helpers

[john.hubbard]

From: John Hubbard <jhubbard@nvidia.com>

Hi,

Here is the first small batch of call site conversions for put_page()
to put_user_page().

This batch includes some, but not all of the places that benefit from the
two new put_user_page_dirty*() helper functions. (The ordering of call site
conversion patch submission makes it better to wait until later, to convert
the rest.)

There are about 50+ patches in my tree [1], and I'll be sending out the
remaining ones in a few more groups:

    * The block/bio related changes (Jerome mostly wrote those, but I've
      had to move stuff around extensively, and add a little code)

    * mm/ changes

    * other subsystem patches

    * an RFC that shows the current state of the tracking patch set. That
      can only be applied after all call sites are converted, but it's
      good to get an early look at it.

This is part a tree-wide conversion, as described in commit fc1d8e7cca2d
("mm: introduce put_user_page*(), placeholder versions").

[1] https://github.com/johnhubbard/linux/tree/gup_dma_core

John Hubbard (4):
  drivers/gpu/drm/via: convert put_page() to put_user_page*()
  net/xdp: convert put_page() to put_user_page*()
  net/rds: convert put_page() to put_user_page*()
  gup: new put_user_page_dirty*() helpers

 drivers/gpu/drm/via/via_dmablit.c        |  5 +++--
 drivers/infiniband/core/umem.c           |  2 +-
 drivers/infiniband/hw/usnic/usnic_uiom.c |  2 +-
 include/linux/mm.h                       | 10 ++++++++++
 net/rds/info.c                           |  5 ++---
 net/rds/message.c                        |  2 +-
 net/rds/rdma.c                           | 15 +++++++--------
 net/xdp/xdp_umem.c                       |  3 +--
 8 files changed, 26 insertions(+), 18 deletions(-)

-- 
2.22.0

[PATCH 1/3] drivers/gpu/drm/via: convert put_page() to put_user_page*()

[john.hubbard]

From: John Hubbard <jhubbard@nvidia.com>

For pages that were retained via get_user_pages*(), release those pages
via the new put_user_page*() routines, instead of via put_page() or
release_pages().

This is part a tree-wide conversion, as described in commit fc1d8e7cca2d
("mm: introduce put_user_page*(), placeholder versions").

Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
---
 drivers/gpu/drm/via/via_dmablit.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/via/via_dmablit.c b/drivers/gpu/drm/via/via_dmablit.c
index 062067438f1d..219827ae114f 100644
--- a/drivers/gpu/drm/via/via_dmablit.c
+++ b/drivers/gpu/drm/via/via_dmablit.c
@@ -189,8 +189,9 @@ via_free_sg_info(struct pci_dev *pdev, drm_via_sg_info_t *vsg)
 		for (i = 0; i < vsg->num_pages; ++i) {
 			if (NULL != (page = vsg->pages[i])) {
 				if (!PageReserved(page) && (DMA_FROM_DEVICE == vsg->direction))
-					SetPageDirty(page);
-				put_page(page);
+					put_user_pages_dirty(&page, 1);
+				else
+					put_user_page(page);
 			}
 		}
 		/* fall through */
-- 
2.22.0

[PATCH 2/3] net/xdp: convert put_page() to put_user_page*()

[john.hubbard]

From: John Hubbard <jhubbard@nvidia.com>

For pages that were retained via get_user_pages*(), release those pages
via the new put_user_page*() routines, instead of via put_page() or
release_pages().

This is part a tree-wide conversion, as described in commit fc1d8e7cca2d
("mm: introduce put_user_page*(), placeholder versions").

Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Magnus Karlsson <magnus.karlsson@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
---
 net/xdp/xdp_umem.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
index 83de74ca729a..9cbbb96c2a32 100644
--- a/net/xdp/xdp_umem.c
+++ b/net/xdp/xdp_umem.c
@@ -171,8 +171,7 @@ static void xdp_umem_unpin_pages(struct xdp_umem *umem)
 	for (i = 0; i < umem->npgs; i++) {
 		struct page *page = umem->pgs[i];
 
-		set_page_dirty_lock(page);
-		put_page(page);
+		put_user_pages_dirty_lock(&page, 1);
 	}
 
 	kfree(umem->pgs);
-- 
2.22.0

[PATCH 3/3] gup: new put_user_page_dirty*() helpers

[john.hubbard]

From: John Hubbard <jhubbard@nvidia.com>

While converting call sites to use put_user_page*() [1], quite a few
places ended up needing a single-page routine to put and dirty a
page.

Provide put_user_page_dirty() and put_user_page_dirty_lock(),
and use them in a few places: net/xdp, drm/via/, drivers/infiniband.

Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
---
 drivers/gpu/drm/via/via_dmablit.c        |  2 +-
 drivers/infiniband/core/umem.c           |  2 +-
 drivers/infiniband/hw/usnic/usnic_uiom.c |  2 +-
 include/linux/mm.h                       | 10 ++++++++++
 net/xdp/xdp_umem.c                       |  2 +-
 5 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/via/via_dmablit.c b/drivers/gpu/drm/via/via_dmablit.c
index 219827ae114f..d30b2d75599f 100644
--- a/drivers/gpu/drm/via/via_dmablit.c
+++ b/drivers/gpu/drm/via/via_dmablit.c
@@ -189,7 +189,7 @@ via_free_sg_info(struct pci_dev *pdev, drm_via_sg_info_t *vsg)
 		for (i = 0; i < vsg->num_pages; ++i) {
 			if (NULL != (page = vsg->pages[i])) {
 				if (!PageReserved(page) && (DMA_FROM_DEVICE == vsg->direction))
-					put_user_pages_dirty(&page, 1);
+					put_user_page_dirty(page);
 				else
 					put_user_page(page);
 			}
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index 08da840ed7ee..a7337cc3ca20 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -55,7 +55,7 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
 	for_each_sg_page(umem->sg_head.sgl, &sg_iter, umem->sg_nents, 0) {
 		page = sg_page_iter_page(&sg_iter);
 		if (umem->writable && dirty)
-			put_user_pages_dirty_lock(&page, 1);
+			put_user_page_dirty_lock(page);
 		else
 			put_user_page(page);
 	}
diff --git a/drivers/infiniband/hw/usnic/usnic_uiom.c b/drivers/infiniband/hw/usnic/usnic_uiom.c
index 0b0237d41613..d2ded624fb2a 100644
--- a/drivers/infiniband/hw/usnic/usnic_uiom.c
+++ b/drivers/infiniband/hw/usnic/usnic_uiom.c
@@ -76,7 +76,7 @@ static void usnic_uiom_put_pages(struct list_head *chunk_list, int dirty)
 			page = sg_page(sg);
 			pa = sg_phys(sg);
 			if (dirty)
-				put_user_pages_dirty_lock(&page, 1);
+				put_user_page_dirty_lock(page);
 			else
 				put_user_page(page);
 			usnic_dbg("pa: %pa\n", &pa);
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 0334ca97c584..c0584c6d9d78 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1061,6 +1061,16 @@ void put_user_pages_dirty(struct page **pages, unsigned long npages);
 void put_user_pages_dirty_lock(struct page **pages, unsigned long npages);
 void put_user_pages(struct page **pages, unsigned long npages);
 
+static inline void put_user_page_dirty(struct page *page)
+{
+	put_user_pages_dirty(&page, 1);
+}
+
+static inline void put_user_page_dirty_lock(struct page *page)
+{
+	put_user_pages_dirty_lock(&page, 1);
+}
+
 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 #define SECTION_IN_PAGE_FLAGS
 #endif
diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
index 9cbbb96c2a32..1d122e52c6de 100644
--- a/net/xdp/xdp_umem.c
+++ b/net/xdp/xdp_umem.c
@@ -171,7 +171,7 @@ static void xdp_umem_unpin_pages(struct xdp_umem *umem)
 	for (i = 0; i < umem->npgs; i++) {
 		struct page *page = umem->pgs[i];
 
-		put_user_pages_dirty_lock(&page, 1);
+		put_user_page_dirty_lock(page);
 	}
 
 	kfree(umem->pgs);
-- 
2.22.0

Re: [PATCH 1/3] drivers/gpu/drm/via: convert put_page() to put_user_page*()

[Christoph Hellwig]

On Sun, Jul 21, 2019 at 09:30:10PM -0700, john.hubbard@gmail.com wrote:
>  		for (i = 0; i < vsg->num_pages; ++i) {
>  			if (NULL != (page = vsg->pages[i])) {
>  				if (!PageReserved(page) && (DMA_FROM_DEVICE == vsg->direction))
> -					SetPageDirty(page);
> -				put_page(page);
> +					put_user_pages_dirty(&page, 1);
> +				else
> +					put_user_page(page);
>  			}

Can't just pass a dirty argument to put_user_pages?  Also do we really
need a separate put_user_page for the single page case?
put_user_pages_dirty?

Also the PageReserved check looks bogus, as I can't see how a reserved
page can end up here.  So IMHO the above snippled should really look
something like this:

	put_user_pages(vsg->pages[i], vsg->num_pages,
			vsg->direction == DMA_FROM_DEVICE);

in the end.

Re: [PATCH 2/3] net/xdp: convert put_page() to put_user_page*()

[Christoph Hellwig]

> diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
> index 83de74ca729a..9cbbb96c2a32 100644
> --- a/net/xdp/xdp_umem.c
> +++ b/net/xdp/xdp_umem.c
> @@ -171,8 +171,7 @@ static void xdp_umem_unpin_pages(struct xdp_umem *umem)
>  	for (i = 0; i < umem->npgs; i++) {
>  		struct page *page = umem->pgs[i];
>  
> -		set_page_dirty_lock(page);
> -		put_page(page);
> +		put_user_pages_dirty_lock(&page, 1);

Same here, we really should avoid the need for the loop here and
do the looping inside the helper.

Re: [PATCH 1/3] drivers/gpu/drm/via: convert put_page() to put_user_page*()

[John Hubbard]

On 7/22/19 2:33 AM, Christoph Hellwig wrote:
> On Sun, Jul 21, 2019 at 09:30:10PM -0700, john.hubbard@gmail.com wrote:
>>  		for (i = 0; i < vsg->num_pages; ++i) {
>>  			if (NULL != (page = vsg->pages[i])) {
>>  				if (!PageReserved(page) && (DMA_FROM_DEVICE == vsg->direction))
>> -					SetPageDirty(page);
>> -				put_page(page);
>> +					put_user_pages_dirty(&page, 1);
>> +				else
>> +					put_user_page(page);
>>  			}
> 
> Can't just pass a dirty argument to put_user_pages?  Also do we really

Yes, and in fact that would help a lot more than the single page case,
which is really just cosmetic after all.

> need a separate put_user_page for the single page case?
> put_user_pages_dirty?

Not really. I'm still zeroing in on the ideal API for all these call sites,
and I agree that the approach below is cleaner.

> 
> Also the PageReserved check looks bogus, as I can't see how a reserved
> page can end up here.  So IMHO the above snippled should really look
> something like this:
> 
> 	put_user_pages(vsg->pages[i], vsg->num_pages,
> 			vsg->direction == DMA_FROM_DEVICE);
> 
> in the end.
> 

Agreed.

thanks,
-- 
John Hubbard
NVIDIA

Re: [PATCH 3/3] gup: new put_user_page_dirty*() helpers

[John Hubbard]

On 7/21/19 9:30 PM, john.hubbard@gmail.com wrote:
> From: John Hubbard <jhubbard@nvidia.com>
> 
> While converting call sites to use put_user_page*() [1], quite a few
> places ended up needing a single-page routine to put and dirty a
> page.
> 
> Provide put_user_page_dirty() and put_user_page_dirty_lock(),
> and use them in a few places: net/xdp, drm/via/, drivers/infiniband.
> 

Please disregard this one, I'm going to drop it, as per the
discussion in patch 1.

thanks,
-- 
John Hubbard
NVIDIA

> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Cc: Jan Kara <jack@suse.cz>
> Signed-off-by: John Hubbard <jhubbard@nvidia.com>
> ---
>  drivers/gpu/drm/via/via_dmablit.c        |  2 +-
>  drivers/infiniband/core/umem.c           |  2 +-
>  drivers/infiniband/hw/usnic/usnic_uiom.c |  2 +-
>  include/linux/mm.h                       | 10 ++++++++++
>  net/xdp/xdp_umem.c                       |  2 +-
>  5 files changed, 14 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/via/via_dmablit.c b/drivers/gpu/drm/via/via_dmablit.c
> index 219827ae114f..d30b2d75599f 100644
> --- a/drivers/gpu/drm/via/via_dmablit.c
> +++ b/drivers/gpu/drm/via/via_dmablit.c
> @@ -189,7 +189,7 @@ via_free_sg_info(struct pci_dev *pdev, drm_via_sg_info_t *vsg)
>  		for (i = 0; i < vsg->num_pages; ++i) {
>  			if (NULL != (page = vsg->pages[i])) {
>  				if (!PageReserved(page) && (DMA_FROM_DEVICE == vsg->direction))
> -					put_user_pages_dirty(&page, 1);
> +					put_user_page_dirty(page);
>  				else
>  					put_user_page(page);
>  			}
> diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
> index 08da840ed7ee..a7337cc3ca20 100644
> --- a/drivers/infiniband/core/umem.c
> +++ b/drivers/infiniband/core/umem.c
> @@ -55,7 +55,7 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
>  	for_each_sg_page(umem->sg_head.sgl, &sg_iter, umem->sg_nents, 0) {
>  		page = sg_page_iter_page(&sg_iter);
>  		if (umem->writable && dirty)
> -			put_user_pages_dirty_lock(&page, 1);
> +			put_user_page_dirty_lock(page);
>  		else
>  			put_user_page(page);
>  	}
> diff --git a/drivers/infiniband/hw/usnic/usnic_uiom.c b/drivers/infiniband/hw/usnic/usnic_uiom.c
> index 0b0237d41613..d2ded624fb2a 100644
> --- a/drivers/infiniband/hw/usnic/usnic_uiom.c
> +++ b/drivers/infiniband/hw/usnic/usnic_uiom.c
> @@ -76,7 +76,7 @@ static void usnic_uiom_put_pages(struct list_head *chunk_list, int dirty)
>  			page = sg_page(sg);
>  			pa = sg_phys(sg);
>  			if (dirty)
> -				put_user_pages_dirty_lock(&page, 1);
> +				put_user_page_dirty_lock(page);
>  			else
>  				put_user_page(page);
>  			usnic_dbg("pa: %pa\n", &pa);
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 0334ca97c584..c0584c6d9d78 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -1061,6 +1061,16 @@ void put_user_pages_dirty(struct page **pages, unsigned long npages);
>  void put_user_pages_dirty_lock(struct page **pages, unsigned long npages);
>  void put_user_pages(struct page **pages, unsigned long npages);
>  
> +static inline void put_user_page_dirty(struct page *page)
> +{
> +	put_user_pages_dirty(&page, 1);
> +}
> +
> +static inline void put_user_page_dirty_lock(struct page *page)
> +{
> +	put_user_pages_dirty_lock(&page, 1);
> +}
> +
>  #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
>  #define SECTION_IN_PAGE_FLAGS
>  #endif
> diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
> index 9cbbb96c2a32..1d122e52c6de 100644
> --- a/net/xdp/xdp_umem.c
> +++ b/net/xdp/xdp_umem.c
> @@ -171,7 +171,7 @@ static void xdp_umem_unpin_pages(struct xdp_umem *umem)
>  	for (i = 0; i < umem->npgs; i++) {
>  		struct page *page = umem->pgs[i];
>  
> -		put_user_pages_dirty_lock(&page, 1);
> +		put_user_page_dirty_lock(page);
>  	}
>  
>  	kfree(umem->pgs);
> 

Re: [PATCH 1/3] drivers/gpu/drm/via: convert put_page() to put_user_page*()

[Matthew Wilcox]

On Mon, Jul 22, 2019 at 11:53:54AM -0700, John Hubbard wrote:
> On 7/22/19 2:33 AM, Christoph Hellwig wrote:
> > On Sun, Jul 21, 2019 at 09:30:10PM -0700, john.hubbard@gmail.com wrote:
> >>  		for (i = 0; i < vsg->num_pages; ++i) {
> >>  			if (NULL != (page = vsg->pages[i])) {
> >>  				if (!PageReserved(page) && (DMA_FROM_DEVICE == vsg->direction))
> >> -					SetPageDirty(page);
> >> -				put_page(page);
> >> +					put_user_pages_dirty(&page, 1);
> >> +				else
> >> +					put_user_page(page);
> >>  			}
> > 
> > Can't just pass a dirty argument to put_user_pages?  Also do we really
> 
> Yes, and in fact that would help a lot more than the single page case,
> which is really just cosmetic after all.
> 
> > need a separate put_user_page for the single page case?
> > put_user_pages_dirty?
> 
> Not really. I'm still zeroing in on the ideal API for all these call sites,
> and I agree that the approach below is cleaner.

so enum { CLEAN = 0, DIRTY = 1, LOCK = 2, DIRTY_LOCK = 3 };
?

Re: [PATCH 1/3] drivers/gpu/drm/via: convert put_page() to put_user_page*()

[John Hubbard]

On 7/22/19 12:07 PM, Matthew Wilcox wrote:
> On Mon, Jul 22, 2019 at 11:53:54AM -0700, John Hubbard wrote:
>> On 7/22/19 2:33 AM, Christoph Hellwig wrote:
>>> On Sun, Jul 21, 2019 at 09:30:10PM -0700, john.hubbard@gmail.com wrote:
>>>>  		for (i = 0; i < vsg->num_pages; ++i) {
>>>>  			if (NULL != (page = vsg->pages[i])) {
>>>>  				if (!PageReserved(page) && (DMA_FROM_DEVICE == vsg->direction))
>>>> -					SetPageDirty(page);
>>>> -				put_page(page);
>>>> +					put_user_pages_dirty(&page, 1);
>>>> +				else
>>>> +					put_user_page(page);
>>>>  			}
>>>
>>> Can't just pass a dirty argument to put_user_pages?  Also do we really
>>
>> Yes, and in fact that would help a lot more than the single page case,
>> which is really just cosmetic after all.
>>
>>> need a separate put_user_page for the single page case?
>>> put_user_pages_dirty?
>>
>> Not really. I'm still zeroing in on the ideal API for all these call sites,
>> and I agree that the approach below is cleaner.
> 
> so enum { CLEAN = 0, DIRTY = 1, LOCK = 2, DIRTY_LOCK = 3 };
> ?
> 

Sure. In fact, I just applied something similar to bio_release_pages()
locally, in order to reconcile Christoph's and Jerome's approaches 
(they each needed to add a bool arg), so I'm all about the enums in the
arg lists. :)

I'm going to post that one shortly, let's see how it goes over. heh.

thanks,
-- 
John Hubbard
NVIDIA