Skip to contentSkip to conversation

fpga: dfl: afu: convert get_user_pages() --> pin_user_pages()

PATCH Superseded John Hubbard 3 participants 6 messages May 19, 2020 – May 25, 2020

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 fpga: dfl: afu: convert get_user_pages() --> pin_user_pages()

This code was using get_user_pages_fast(), in a "Case 2" scenario (DMA/RDMA), using the categorization from [1]. That means that it's time to convert the get_user_pages_fast() + put_page() calls to pin_user_pages_fast() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small part of fixing a long-standing disconnect between pinning pages, and file systems' use of those pages.

[1] Documentation/core-api/pin_user_pages.rst

[2] "Explicit pinning of user-space pages":

    https://lwn.net/Articles/807108/

---

Hi,

Changes since v1:

Changed the label from "put_pages", to "unpin_pages".

thanks,
John Hubbard
NVIDIA

 drivers/fpga/dfl-afu-dma-region.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)
drivers/fpga/dfl-afu-dma-region.c +5 -14
index 62f924489db5..a31dd3a7e581 100644
@@ -16,15 +16,6 @@
 
 #include "dfl-afu.h"
 
-static void put_all_pages(struct page **pages, int npages)
-{
-	int i;
-
-	for (i = 0; i < npages; i++)
-		if (pages[i])
-			put_page(pages[i]);
-}
-
 void afu_dma_region_init(struct dfl_feature_platform_data *pdata)
 {
 	struct dfl_afu *afu = dfl_fpga_pdata_get_private(pdata);
@@ -57,11 +48,11 @@ static int afu_dma_pin_pages(struct dfl_feature_platform_data *pdata,
 		goto unlock_vm;
 	}
 
-	pinned = get_user_pages_fast(region->user_addr, npages, FOLL_WRITE,
+	pinned = pin_user_pages_fast(region->user_addr, npages, FOLL_WRITE,
 				     region->pages);
 	if (pinned < 0) {
 		ret = pinned;
-		goto put_pages;
+		goto unpin_pages;
 	} else if (pinned != npages) {
 		ret = -EFAULT;
 		goto free_pages;
@@ -71,8 +62,8 @@ static int afu_dma_pin_pages(struct dfl_feature_platform_data *pdata,
 
 	return 0;
 
-put_pages:
-	put_all_pages(region->pages, pinned);
+unpin_pages:
+	unpin_user_pages(region->pages, pinned);
 free_pages:
 	kfree(region->pages);
 unlock_vm:
@@ -94,7 +85,7 @@ static void afu_dma_unpin_pages(struct dfl_feature_platform_data *pdata,
 	long npages = region->length >> PAGE_SHIFT;
 	struct device *dev = &pdata->dev->dev;
 
-	put_all_pages(region->pages, npages);
+	unpin_user_pages(region->pages, npages);
 	kfree(region->pages);
 	account_locked_vm(current->mm, npages, false);
 
Signature · 1 line
2.26.2
#1 lore raw
Quoted 11 lines · John Hubbard

This code was using get_user_pages_fast(), in a "Case 2" scenario (DMA/RDMA), using the categorization from [1]. That means that it's time to convert the get_user_pages_fast() + put_page() calls to pin_user_pages_fast() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small part of fixing a long-standing disconnect between pinning pages, and file systems' use of those pages.

[1] Documentation/core-api/pin_user_pages.rst

Quoted 8 lines

[2] "Explicit pinning of user-space pages":

     https://lwn.net/Articles/807108/

Hi Moritz and FPGA developers,

Is this OK? And if so, is it going into your git tree? Or should I send it up through a different tree? (I'm new to the FPGA development model).

thanks,

Signature · 2 lines
John Hubbard
NVIDIA
Quoted 70 lines

---

Hi,

Changes since v1:

Changed the label from "put_pages", to "unpin_pages".

thanks,
John Hubbard
NVIDIA

  drivers/fpga/dfl-afu-dma-region.c | 19 +++++--------------
  1 file changed, 5 insertions(+), 14 deletions(-)
drivers/fpga/dfl-afu-dma-region.c +5 -14
index 62f924489db5..a31dd3a7e581 100644
@@ -16,15 +16,6 @@
 
  #include "dfl-afu.h"
 
-static void put_all_pages(struct page **pages, int npages)
-{
-	int i;
-
-	for (i = 0; i < npages; i++)
-		if (pages[i])
-			put_page(pages[i]);
-}
-
  void afu_dma_region_init(struct dfl_feature_platform_data *pdata)
  {
  	struct dfl_afu *afu = dfl_fpga_pdata_get_private(pdata);
@@ -57,11 +48,11 @@ static int afu_dma_pin_pages(struct dfl_feature_platform_data *pdata,
  		goto unlock_vm;
  	}
 
-	pinned = get_user_pages_fast(region->user_addr, npages, FOLL_WRITE,
+	pinned = pin_user_pages_fast(region->user_addr, npages, FOLL_WRITE,
  				     region->pages);
  	if (pinned < 0) {
  		ret = pinned;
-		goto put_pages;
+		goto unpin_pages;
  	} else if (pinned != npages) {
  		ret = -EFAULT;
  		goto free_pages;
@@ -71,8 +62,8 @@ static int afu_dma_pin_pages(struct dfl_feature_platform_data *pdata,
 
  	return 0;
 
-put_pages:
-	put_all_pages(region->pages, pinned);
+unpin_pages:
+	unpin_user_pages(region->pages, pinned);
  free_pages:
  	kfree(region->pages);
  unlock_vm:
@@ -94,7 +85,7 @@ static void afu_dma_unpin_pages(struct dfl_feature_platform_data *pdata,
  	long npages = region->length >> PAGE_SHIFT;
  	struct device *dev = &pdata->dev->dev;
 
-	put_all_pages(region->pages, npages);
+	unpin_user_pages(region->pages, npages);
  	kfree(region->pages);
  	account_locked_vm(current->mm, npages, false);
 
maintainer
#2 lore raw
Quoted 22 lines · John Hubbard
Quoted 11 lines · John Hubbard

This code was using get_user_pages_fast(), in a "Case 2" scenario (DMA/RDMA), using the categorization from [1]. That means that it's time to convert the get_user_pages_fast() + put_page() calls to pin_user_pages_fast() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small part of fixing a long-standing disconnect between pinning pages, and file systems' use of those pages.

[1] Documentation/core-api/pin_user_pages.rst

Quoted 8 lines

[2] "Explicit pinning of user-space pages":

     https://lwn.net/Articles/807108/
Quoted 5 lines

Hi Moritz and FPGA developers,

Is this OK? And if so, is it going into your git tree? Or should I send it up through a different tree? (I'm new to the FPGA development model).

I can take it, sorry for sluggish response.

Cheers,
Moritz

Quoted 72 lines

thanks,

Signature · 2 lines
John Hubbard
NVIDIA
Quoted 62 lines

---

Hi,

Changes since v1:

Changed the label from "put_pages", to "unpin_pages".

thanks,
John Hubbard
NVIDIA

  drivers/fpga/dfl-afu-dma-region.c | 19 +++++--------------
  1 file changed, 5 insertions(+), 14 deletions(-)
drivers/fpga/dfl-afu-dma-region.c +5 -14
index 62f924489db5..a31dd3a7e581 100644
@@ -16,15 +16,6 @@
  #include "dfl-afu.h"
-static void put_all_pages(struct page **pages, int npages)
-{
-	int i;
-
-	for (i = 0; i < npages; i++)
-		if (pages[i])
-			put_page(pages[i]);
-}
-
  void afu_dma_region_init(struct dfl_feature_platform_data *pdata)
  {
  	struct dfl_afu *afu = dfl_fpga_pdata_get_private(pdata);
@@ -57,11 +48,11 @@ static int afu_dma_pin_pages(struct dfl_feature_platform_data *pdata,
  		goto unlock_vm;
  	}
-	pinned = get_user_pages_fast(region->user_addr, npages, FOLL_WRITE,
+	pinned = pin_user_pages_fast(region->user_addr, npages, FOLL_WRITE,
  				     region->pages);
  	if (pinned < 0) {
  		ret = pinned;
-		goto put_pages;
+		goto unpin_pages;
  	} else if (pinned != npages) {
  		ret = -EFAULT;
  		goto free_pages;
@@ -71,8 +62,8 @@ static int afu_dma_pin_pages(struct dfl_feature_platform_data *pdata,
  	return 0;
-put_pages:
-	put_all_pages(region->pages, pinned);
+unpin_pages:
+	unpin_user_pages(region->pages, pinned);
  free_pages:
  	kfree(region->pages);
  unlock_vm:
@@ -94,7 +85,7 @@ static void afu_dma_unpin_pages(struct dfl_feature_platform_data *pdata,
  	long npages = region->length >> PAGE_SHIFT;
  	struct device *dev = &pdata->dev->dev;
-	put_all_pages(region->pages, npages);
+	unpin_user_pages(region->pages, npages);
  	kfree(region->pages);
  	account_locked_vm(current->mm, npages, false);
 
#3 lore raw
Quoted 23 lines · Moritz Fischer
Quoted 22 lines · John Hubbard
Quoted 19 lines · John Hubbard

This code was using get_user_pages_fast(), in a "Case 2" scenario (DMA/RDMA), using the categorization from [1]. That means that it's time to convert the get_user_pages_fast() + put_page() calls to pin_user_pages_fast() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small part of fixing a long-standing disconnect between pinning pages, and file systems' use of those pages.

[1] Documentation/core-api/pin_user_pages.rst

[2] "Explicit pinning of user-space pages":

      https://lwn.net/Articles/807108/
Quoted 8 lines
Quoted 5 lines

Hi Moritz and FPGA developers,

Is this OK? And if so, is it going into your git tree? Or should I send it up through a different tree? (I'm new to the FPGA development model).

I can take it, sorry for sluggish response.

That's great news, thanks Moritz! Sorry to be pushy, just didn't want it to get lost. :)

thanks,

Signature · 2 lines
John Hubbard
NVIDIA
#4 lore raw
Quoted 11 lines
Quoted 8 lines
Quoted 5 lines

Hi Moritz and FPGA developers,

Is this OK? And if so, is it going into your git tree? Or should I send it up through a different tree? (I'm new to the FPGA development model).

I can take it, sorry for sluggish response.

That's great news, thanks Moritz! Sorry to be pushy, just didn't want it to get lost. :)

Thanks John for this patch, and thanks Moritz for taking care of this. Sorry for late response, one thing here we may need to be careful is a recent bug fixing patch, that fixing patch has been merged by Greg in char-misc-next tree, and may have some conflict with this one.

https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc.git/commit/?h=char-misc-next&id=c9d7e3da1f3c4cf5dddfc5d7ce4d76d013aba1cc
fpga: dfl: afu: Corrected error handling levels

I guess we need to rebase this patch on top of it? Moritz, do you have any suggestion?

Thanks
Hao

Quoted 5 lines

thanks,
--
John Hubbard
NVIDIA

#5 lore raw
Quoted 17 lines · Hao
Quoted 11 lines
Quoted 8 lines
Quoted 5 lines

Hi Moritz and FPGA developers,

Is this OK? And if so, is it going into your git tree? Or should I send it up through a different tree? (I'm new to the FPGA development model).

I can take it, sorry for sluggish response.

That's great news, thanks Moritz! Sorry to be pushy, just didn't want it to get lost. :)

Thanks John for this patch, and thanks Moritz for taking care of this. Sorry for late response, one thing here we may need to be careful is a recent bug fixing patch, that fixing patch has been merged by Greg in char-misc-next tree, and may have some conflict with this one.

Quoted 6 lines

https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc.git/commit/?h=char-misc-next&id=c9d7e3da1f3c4cf5dddfc5d7ce4d76d013aba1cc
fpga: dfl: afu: Corrected error handling levels

I guess we need to rebase this patch on top of it? Moritz, do you have any suggestion?

I'll send you a v2, rebased on top of the latest char-misc.git, no problem.

thanks,

Signature · 2 lines
John Hubbard
NVIDIA