Skip to contentSkip to conversation

pci: remove pci_mmap_page_range wrapper

PATCH In review Arnd Bergmann 2 participants 3 messages Jul 15, 2022 – Jul 29, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 pci: remove pci_mmap_page_range wrapper

From: Arnd Bergmann <[email protected]>

The ARCH_GENERIC_PCI_MMAP_RESOURCE symbol came up in a recent discussion, and I noticed that this was left behind by an unfinished cleanup from 2017.

The only architecture that still relies on providing its
own pci_mmap_page_range() helper instead of using the generic pci_mmap_resource_range() is sparc. Presumably the reasons for this have not changed, but at least this can be simplified by converting sparc to use the same interface as the others.

The only difference between the two is the device specific offset that gets added to or subtracted from vma->vm_pgoff.

Change the only caller of pci_mmap_page_range() in common code to subtract this offset and call the modern interface, while adding it back in the sparc implementation to preserve the existing behavior.

This removes the complexities of the dual interfaces from the common code, and keeps it all specific to the sparc architecture code. According to David Miller, the sparc code lets user space poke into the VGA I/O port registers by mmapping the I/O space of the parent bridge device, which is something that the generic pci_mmap_resource_range() code apparently does not.

---
 Documentation/PCI/sysfs-pci.rst |  2 +-
 arch/sparc/kernel/pci.c         | 13 +++++++---
 drivers/pci/mmap.c              | 44 ---------------------------------
 drivers/pci/proc.c              |  7 +++++-
 include/linux/pci.h             | 12 +--------
 5 files changed, 17 insertions(+), 61 deletions(-)
Documentation/PCI/sysfs-pci.rst +1 -1
index 742fbd21dc1f..f495185aa88a 100644
@@ -125,7 +125,7 @@ implementation of that functionality. To support the historical interface of
 mmap() through files in /proc/bus/pci, platforms may also set HAVE_PCI_MMAP.
 
 Alternatively, platforms which set HAVE_PCI_MMAP may provide their own
-implementation of pci_mmap_page_range() instead of defining
+implementation of pci_mmap_resource_range() instead of defining
 ARCH_GENERIC_PCI_MMAP_RESOURCE.
 
 Platforms which support write-combining maps of PCI resources must define
arch/sparc/kernel/pci.c +9 -4
index 31b0c1983286..f580db840bf7 100644
@@ -876,17 +876,22 @@ static void __pci_mmap_set_pgprot(struct pci_dev *dev, struct vm_area_struct *vm
 
 /* Perform the actual remap of the pages for a PCI device mapping, as appropriate
  * for this architecture.  The region in the process to map is described by vm_start
- * and vm_end members of VMA, the base physical address is found in vm_pgoff.
+ * and vm_end members of VMA, the BAR relative address is found in vm_pgoff.
  * The pci device structure is provided so that architectures may make mapping
  * decisions on a per-device or per-bus basis.
  *
  * Returns a negative error code on failure, zero on success.
  */
-int pci_mmap_page_range(struct pci_dev *dev, int bar,
-			struct vm_area_struct *vma,
-			enum pci_mmap_state mmap_state, int write_combine)
+int pci_mmap_resource_range(struct pci_dev *dev, int bar,
+			    struct vm_area_struct *vma,
+			    enum pci_mmap_state mmap_state, int write_combine)
 {
 	int ret;
+	resource_size_t start, end;
+
+	/* convert per-BAR address to PCI bus address */
+	pci_resource_to_user(dev, bar, &dev->resource[bar], &start, &end);
+	vma->vm_pgoff += start >> PAGE_SHIFT;
 
 	ret = __pci_mmap_make_offset(dev, vma, mmap_state);
 	if (ret < 0)
drivers/pci/mmap.c +0 -44

54 lines not on this page · raw message

drivers/pci/proc.c +6 -1
index 31b26d8ea6cc..f967709082d6 100644
@@ -244,6 +244,7 @@ static int proc_bus_pci_mmap(struct file *file, struct vm_area_struct *vma)
 {
 	struct pci_dev *dev = pde_data(file_inode(file));
 	struct pci_filp_private *fpriv = file->private_data;
+	resource_size_t start, end;
 	int i, ret, write_combine = 0, res_bit = IORESOURCE_MEM;
 
 	if (!capable(CAP_SYS_RAWIO) ||
@@ -278,7 +279,11 @@ static int proc_bus_pci_mmap(struct file *file, struct vm_area_struct *vma)
 	    iomem_is_exclusive(dev->resource[i].start))
 		return -EINVAL;
 
-	ret = pci_mmap_page_range(dev, i, vma,
+	pci_resource_to_user(dev, i, &dev->resource[i], &start, &end);
+
+	/* Adjust vm_pgoff to be the offset within the resource */
+	vma->vm_pgoff -= start >> PAGE_SHIFT;
+	ret = pci_mmap_resource_range(dev, i, vma,
 				  fpriv->mmap_state, write_combine);
 	if (ret < 0)
 		return ret;
include/linux/pci.h +1 -11
index 81a57b498f22..060af91bafcd 100644
@@ -1909,24 +1909,14 @@ pci_alloc_irq_vectors(struct pci_dev *dev, unsigned int min_vecs,
 
 #include <asm/pci.h>
 
-/* These two functions provide almost identical functionality. Depending
- * on the architecture, one will be implemented as a wrapper around the
- * other (in drivers/pci/mmap.c).
- *
+/*
  * pci_mmap_resource_range() maps a specific BAR, and vm->vm_pgoff
  * is expected to be an offset within that region.
  *
- * pci_mmap_page_range() is the legacy architecture-specific interface,
- * which accepts a "user visible" resource address converted by
- * pci_resource_to_user(), as used in the legacy mmap() interface in
- * /proc/bus/pci/.
  */
 int pci_mmap_resource_range(struct pci_dev *dev, int bar,
 			    struct vm_area_struct *vma,
 			    enum pci_mmap_state mmap_state, int write_combine);
-int pci_mmap_page_range(struct pci_dev *pdev, int bar,
-			struct vm_area_struct *vma,
-			enum pci_mmap_state mmap_state, int write_combine);
 
 #ifndef arch_can_pci_mmap_wc
 #define arch_can_pci_mmap_wc()		0
Signature · 1 line
2.29.2
#1 lore raw

RFC PATCH 2/2 sparc: Use generic pci_mmap_resource_range()

From: Arnd Bergmann <[email protected]>

The main feature of the sparc specific implementation of pci_mmap_resource_range() is that it allows mapping the entire PCI I/O space for a PCI host bridge using the /proc/bus/pci interface on a bridge device.

The generic implementation cannot do this, but it also appears that this got broken for sparc by commit 9eff02e2042f ("PCI: check mmap range of /proc/bus/pci files too"), which enforces that each address is part of a BAR for kernels after 2.6.28.

Remove it all, assuming that the corresponding user space code has already been changed to access /dev/ioport instead a long time ago. The pci_iobar_pfn() function needs to be added to make it possible to map I/O resources. This is adapted from the powerpc version.

---
It's quite possible that I missed something important, but it appears that David Miller was thinking of the pre-2.6.28 behavior that was already broken when David Woodhouse sent his series.

---
 arch/sparc/include/asm/pci.h |   1 +
 arch/sparc/kernel/pci.c      | 154 +----------------------------------
 2 files changed, 5 insertions(+), 150 deletions(-)
arch/sparc/include/asm/pci.h +1 -0
index 4deddf430e5d..dff90dce6cb7 100644
@@ -37,6 +37,7 @@ static inline int pci_proc_domain(struct pci_bus *bus)
 #define HAVE_PCI_MMAP
 #define arch_can_pci_mmap_io()	1
 #define HAVE_ARCH_PCI_GET_UNMAPPED_AREA
+#define ARCH_GENERIC_PCI_MMAP_RESOURCE
 #define get_pci_unmapped_area get_fb_unmapped_area
 #endif /* CONFIG_SPARC64 */
 
arch/sparc/kernel/pci.c +4 -150

165 lines not on this page · raw message

Signature · 1 line
2.29.2
#2 lore raw
In reply to Arnd Bergmann
Quoted 26 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

The ARCH_GENERIC_PCI_MMAP_RESOURCE symbol came up in a recent discussion, and I noticed that this was left behind by an unfinished cleanup from 2017.

The only architecture that still relies on providing its
own pci_mmap_page_range() helper instead of using the generic pci_mmap_resource_range() is sparc. Presumably the reasons for this have not changed, but at least this can be simplified by converting sparc to use the same interface as the others.

The only difference between the two is the device specific offset that gets added to or subtracted from vma->vm_pgoff.

Change the only caller of pci_mmap_page_range() in common code to subtract this offset and call the modern interface, while adding it back in the sparc implementation to preserve the existing behavior.

This removes the complexities of the dual interfaces from the common code, and keeps it all specific to the sparc architecture code. According to David Miller, the sparc code lets user space poke into the VGA I/O port registers by mmapping the I/O space of the parent bridge device, which is something that the generic pci_mmap_resource_range() code apparently does not.

Quoted 6 lines

I applied both these patches on a pci/resource branch for v5.20, thanks!

Sparc folks, let me know if you object or prefer to take these via your tree.

Quoted 170 lines
---
 Documentation/PCI/sysfs-pci.rst |  2 +-
 arch/sparc/kernel/pci.c         | 13 +++++++---
 drivers/pci/mmap.c              | 44 ---------------------------------
 drivers/pci/proc.c              |  7 +++++-
 include/linux/pci.h             | 12 +--------
 5 files changed, 17 insertions(+), 61 deletions(-)
Documentation/PCI/sysfs-pci.rst +1 -1
index 742fbd21dc1f..f495185aa88a 100644
@@ -125,7 +125,7 @@ implementation of that functionality. To support the historical interface of
 mmap() through files in /proc/bus/pci, platforms may also set HAVE_PCI_MMAP.
 
 Alternatively, platforms which set HAVE_PCI_MMAP may provide their own
-implementation of pci_mmap_page_range() instead of defining
+implementation of pci_mmap_resource_range() instead of defining
 ARCH_GENERIC_PCI_MMAP_RESOURCE.
 
 Platforms which support write-combining maps of PCI resources must define
arch/sparc/kernel/pci.c +9 -4
index 31b0c1983286..f580db840bf7 100644
@@ -876,17 +876,22 @@ static void __pci_mmap_set_pgprot(struct pci_dev *dev, struct vm_area_struct *vm
 
 /* Perform the actual remap of the pages for a PCI device mapping, as appropriate
  * for this architecture.  The region in the process to map is described by vm_start
- * and vm_end members of VMA, the base physical address is found in vm_pgoff.
+ * and vm_end members of VMA, the BAR relative address is found in vm_pgoff.
  * The pci device structure is provided so that architectures may make mapping
  * decisions on a per-device or per-bus basis.
  *
  * Returns a negative error code on failure, zero on success.
  */
-int pci_mmap_page_range(struct pci_dev *dev, int bar,
-			struct vm_area_struct *vma,
-			enum pci_mmap_state mmap_state, int write_combine)
+int pci_mmap_resource_range(struct pci_dev *dev, int bar,
+			    struct vm_area_struct *vma,
+			    enum pci_mmap_state mmap_state, int write_combine)
 {
 	int ret;
+	resource_size_t start, end;
+
+	/* convert per-BAR address to PCI bus address */
+	pci_resource_to_user(dev, bar, &dev->resource[bar], &start, &end);
+	vma->vm_pgoff += start >> PAGE_SHIFT;
 
 	ret = __pci_mmap_make_offset(dev, vma, mmap_state);
 	if (ret < 0)
drivers/pci/mmap.c +0 -44
index b8c9011987f4..4504039056d1 100644
@@ -13,27 +13,6 @@
 
 #ifdef ARCH_GENERIC_PCI_MMAP_RESOURCE
 
-/*
- * Modern setup: generic pci_mmap_resource_range(), and implement the legacy
- * pci_mmap_page_range() (if needed) as a wrapper round it.
- */
-
-#ifdef HAVE_PCI_MMAP
-int pci_mmap_page_range(struct pci_dev *pdev, int bar,
-			struct vm_area_struct *vma,
-			enum pci_mmap_state mmap_state, int write_combine)
-{
-	resource_size_t start, end;
-
-	pci_resource_to_user(pdev, bar, &pdev->resource[bar], &start, &end);
-
-	/* Adjust vm_pgoff to be the offset within the resource */
-	vma->vm_pgoff -= start >> PAGE_SHIFT;
-	return pci_mmap_resource_range(pdev, bar, vma, mmap_state,
-				       write_combine);
-}
-#endif
-
 static const struct vm_operations_struct pci_phys_vm_ops = {
 #ifdef CONFIG_HAVE_IOREMAP_PROT
 	.access = generic_access_phys,
@@ -70,27 +49,4 @@ int pci_mmap_resource_range(struct pci_dev *pdev, int bar,
 				  vma->vm_page_prot);
 }
 
-#elif defined(HAVE_PCI_MMAP) /* && !ARCH_GENERIC_PCI_MMAP_RESOURCE */
-
-/*
- * Legacy setup: Implement pci_mmap_resource_range() as a wrapper around
- * the architecture's pci_mmap_page_range(), converting to "user visible"
- * addresses as necessary.
- */
-
-int pci_mmap_resource_range(struct pci_dev *pdev, int bar,
-			    struct vm_area_struct *vma,
-			    enum pci_mmap_state mmap_state, int write_combine)
-{
-	resource_size_t start, end;
-
-	/*
-	 * pci_mmap_page_range() expects the same kind of entry as coming
-	 * from /proc/bus/pci/ which is a "user visible" value. If this is
-	 * different from the resource itself, arch will do necessary fixup.
-	 */
-	pci_resource_to_user(pdev, bar, &pdev->resource[bar], &start, &end);
-	vma->vm_pgoff += start >> PAGE_SHIFT;
-	return pci_mmap_page_range(pdev, bar, vma, mmap_state, write_combine);
-}
 #endif
drivers/pci/proc.c +6 -1
index 31b26d8ea6cc..f967709082d6 100644
@@ -244,6 +244,7 @@ static int proc_bus_pci_mmap(struct file *file, struct vm_area_struct *vma)
 {
 	struct pci_dev *dev = pde_data(file_inode(file));
 	struct pci_filp_private *fpriv = file->private_data;
+	resource_size_t start, end;
 	int i, ret, write_combine = 0, res_bit = IORESOURCE_MEM;
 
 	if (!capable(CAP_SYS_RAWIO) ||
@@ -278,7 +279,11 @@ static int proc_bus_pci_mmap(struct file *file, struct vm_area_struct *vma)
 	    iomem_is_exclusive(dev->resource[i].start))
 		return -EINVAL;
 
-	ret = pci_mmap_page_range(dev, i, vma,
+	pci_resource_to_user(dev, i, &dev->resource[i], &start, &end);
+
+	/* Adjust vm_pgoff to be the offset within the resource */
+	vma->vm_pgoff -= start >> PAGE_SHIFT;
+	ret = pci_mmap_resource_range(dev, i, vma,
 				  fpriv->mmap_state, write_combine);
 	if (ret < 0)
 		return ret;
include/linux/pci.h +1 -11
index 81a57b498f22..060af91bafcd 100644
@@ -1909,24 +1909,14 @@ pci_alloc_irq_vectors(struct pci_dev *dev, unsigned int min_vecs,
 
 #include <asm/pci.h>
 
-/* These two functions provide almost identical functionality. Depending
- * on the architecture, one will be implemented as a wrapper around the
- * other (in drivers/pci/mmap.c).
- *
+/*
  * pci_mmap_resource_range() maps a specific BAR, and vm->vm_pgoff
  * is expected to be an offset within that region.
  *
- * pci_mmap_page_range() is the legacy architecture-specific interface,
- * which accepts a "user visible" resource address converted by
- * pci_resource_to_user(), as used in the legacy mmap() interface in
- * /proc/bus/pci/.
  */
 int pci_mmap_resource_range(struct pci_dev *dev, int bar,
 			    struct vm_area_struct *vma,
 			    enum pci_mmap_state mmap_state, int write_combine);
-int pci_mmap_page_range(struct pci_dev *pdev, int bar,
-			struct vm_area_struct *vma,
-			enum pci_mmap_state mmap_state, int write_combine);
 
 #ifndef arch_can_pci_mmap_wc
 #define arch_can_pci_mmap_wc()		0
Signature · 1 line
2.29.2