[PATCH v6 00/21] Userspace P2PDMA with O_DIRECT NVMe devices

[PATCH v6 00/21] Userspace P2PDMA with O_DIRECT NVMe devices

[Logan Gunthorpe]

Hi,

This patchset continues my work to add userspace P2PDMA access using
O_DIRECT NVMe devices. This posting contains some minor fixes and a
rebase onto v5.18-rc1 which contains cleanup from Christoph around
free_zone_device_page() that helps to enable this patchset. The
previous posting was here[1].

The patchset enables userspace P2PDMA by allowing userspace to mmap()
allocated chunks of the CMB. The resulting VMA can be passed only
to O_DIRECT IO on NVMe backed files or block devices. A flag is added
to GUP() in Patch <>, then Patches <> through <> wire this flag up based
on whether the block queue indicates P2PDMA support. Patches <>
through <> enable the CMB to be mapped into userspace by mmaping
the nvme char device.

This is relatively straightforward, however the one significant
problem is that, presently, pci_p2pdma_map_sg() requires a homogeneous
SGL with all P2PDMA pages or all regular pages. Enhancing GUP to
support enforcing this rule would require a huge hack that I don't
expect would be all that pallatable. So the first 13 patches add
support for P2PDMA pages to dma_map_sg[table]() to the dma-direct
and dma-iommu implementations. Thus systems without an IOMMU plus
Intel and AMD IOMMUs are supported. (Other IOMMU implementations would
then be unsupported, notably ARM and PowerPC but support would be added
when they convert to dma-iommu).

dma_map_sgtable() is preferred when dealing with P2PDMA memory as it
will return -EREMOTEIO when the DMA device cannot map specific P2PDMA
pages based on the existing rules in calc_map_type_and_dist().

The other issue is dma_unmap_sg() needs a flag to determine whether a
given dma_addr_t was mapped regularly or as a PCI bus address. To allow
this, a third flag is added to the page_link field in struct
scatterlist. This effectively means support for P2PDMA will now depend
on CONFIG_64BIT.

Feedback welcome.

This series is based on v5.18-rc1. A git branch is available here:

  https://github.com/sbates130272/linux-p2pmem/  p2pdma_user_cmb_v6

Thanks,

Logan

[1] lkml.kernel.org/r/20220128002614.6136-1-logang@deltatee.com

--

Changes since v5:
  - Rebased onto v5.18-rc1 which includes Christophs cleanup to
    free_zone_device_page() (similar to Ralph's patch).
  - Fix bug with concurrent first calls to pci_p2pdma_vma_fault()
    that caused a double allocation and lost p2p memory. Noticed
    by Andrew Maier.
  - Collected a Reviewed-by tag from Chaitanya.
  - Numerous minor fixes to commit messages

Changes since v4:
  - Rebase onto v5.17-rc1.
  - Included Ralph Cambell's patches which removes the ZONE_DEVICE page
    reference count offset. This is just to demonstrate that this
    series is compatible with that direction.
  - Added a comment in pci_p2pdma_map_sg_attrs(), per Chaitanya and
    included his Reviewed-by tags.
  - Patch 1 in the last series which cleaned up scatterlist.h
    has been upstreamed.
  - Dropped NEED_SG_DMA_BUS_ADDR_FLAG seeing depends on doesn't
    work with selected symbols, per Christoph.
  - Switched iov_iter_get_pages_[alloc_]flags to be exported with
    EXPORT_SYMBOL_GPL, per Christoph.
  - Renamed zone_device_pages_are_mergeable() to
    zone_device_pages_have_same_pgmap(), per Christoph.
  - Renamed .mmap_file_open operation in nvme_ctrl_ops to
    cdev_file_open(), per Christoph.

Changes since v3:
  - Add some comment and commit message cleanups I had missed for v3,
    also moved the prototypes for some of the p2pdma helpers to
    dma-map-ops.h (which I missed in v3 and was suggested in v2).
  - Add separate cleanup patch for scatterlist.h and change the macros
    to functions. (Suggested by Chaitanya and Jason, respectively)
  - Rename sg_dma_mark_pci_p2pdma() and sg_is_dma_pci_p2pdma() to
    sg_dma_mark_bus_address() and sg_is_dma_bus_address() which
    is a more generic name (As requested by Jason)
  - Fixes to some comments and commit messages as suggested by Bjorn
    and Jason.
  - Ensure swiotlb is not used with P2PDMA pages. (Per Jason)
  - The sgtable coversion in RDMA was split out and sent upstream
    separately, the new patch is only the removal. (Per Jason)
  - Moved the FOLL_PCI_P2PDMA check outside of get_dev_pagemap() as
    Jason suggested this will be removed in the near term.
  - Add two patches to ensure that zone device pages with different
    pgmaps are never merged in the block layer or
    sg_alloc_append_table_from_pages() (Per Jason)
  - Ensure synchronize_rcu() or call_rcu() is used before returning
    pages to the genalloc. (Jason pointed out that pages are not
    gauranteed to be unused in all architectures until at least
    after an RCU grace period, and that synchronize_rcu() was likely
    too slow to use in the vma close operation.
  - Collected Acks and Reviews by Bjorn, Jason and Max.

Logan Gunthorpe (21):
  lib/scatterlist: add flag for indicating P2PDMA segments in an SGL
  PCI/P2PDMA: Attempt to set map_type if it has not been set
  PCI/P2PDMA: Expose pci_p2pdma_map_type()
  PCI/P2PDMA: Introduce helpers for dma_map_sg implementations
  dma-mapping: allow EREMOTEIO return code for P2PDMA transfers
  dma-direct: support PCI P2PDMA pages in dma-direct map_sg
  dma-mapping: add flags to dma_map_ops to indicate PCI P2PDMA support
  iommu/dma: support PCI P2PDMA pages in dma-iommu map_sg
  nvme-pci: check DMA ops when indicating support for PCI P2PDMA
  nvme-pci: convert to using dma_map_sgtable()
  RDMA/core: introduce ib_dma_pci_p2p_dma_supported()
  RDMA/rw: drop pci_p2pdma_[un]map_sg()
  PCI/P2PDMA: Remove pci_p2pdma_[un]map_sg()
  mm: introduce FOLL_PCI_P2PDMA to gate getting PCI P2PDMA pages
  iov_iter: introduce iov_iter_get_pages_[alloc_]flags()
  block: add check when merging zone device pages
  lib/scatterlist: add check when merging zone device pages
  block: set FOLL_PCI_P2PDMA in __bio_iov_iter_get_pages()
  block: set FOLL_PCI_P2PDMA in bio_map_user_iov()
  PCI/P2PDMA: Introduce pci_mmap_p2pmem()
  nvme-pci: allow mmaping the CMB in userspace

 block/bio.c                  |  10 +-
 block/blk-map.c              |   7 +-
 drivers/infiniband/core/rw.c |  45 +---
 drivers/iommu/dma-iommu.c    |  68 ++++-
 drivers/nvme/host/core.c     |  18 +-
 drivers/nvme/host/nvme.h     |   4 +-
 drivers/nvme/host/pci.c      |  97 ++++----
 drivers/nvme/target/rdma.c   |   2 +-
 drivers/pci/Kconfig          |   5 +
 drivers/pci/p2pdma.c         | 467 ++++++++++++++++++++++++++++++-----
 include/linux/dma-map-ops.h  |  76 ++++++
 include/linux/dma-mapping.h  |   5 +
 include/linux/mm.h           |  24 ++
 include/linux/pci-p2pdma.h   |  38 +--
 include/linux/scatterlist.h  |  44 +++-
 include/linux/uio.h          |   6 +
 include/rdma/ib_verbs.h      |  11 +
 include/uapi/linux/magic.h   |   1 +
 kernel/dma/direct.c          |  43 +++-
 kernel/dma/direct.h          |   8 +-
 kernel/dma/mapping.c         |  22 +-
 lib/iov_iter.c               |  25 +-
 lib/scatterlist.c            |  25 +-
 mm/gup.c                     |  22 +-
 24 files changed, 867 insertions(+), 206 deletions(-)


base-commit: 3123109284176b1532874591f7c81f3837bbdc17
-- 
2.30.2

[PATCH v6 01/21] lib/scatterlist: add flag for indicating P2PDMA segments in an SGL

[Logan Gunthorpe]

Make use of the third free LSB in scatterlist's page_link on 64bit systems.

The extra bit will be used by dma_[un]map_sg_p2pdma() to determine when a
given SGL segments dma_address points to a PCI bus address.
dma_unmap_sg_p2pdma() will need to perform different cleanup when a
segment is marked as a bus address.

The new bit will only be used when CONFIG_PCI_P2PDMA is set; this means
PCI P2PDMA will require CONFIG_64BIT. This should be acceptable as the
majority of P2PDMA use cases are restricted to newer root complexes and
roughly require the extra address space for memory BARs used in the
transactions.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
---
 drivers/pci/Kconfig         |  5 +++++
 include/linux/scatterlist.h | 44 ++++++++++++++++++++++++++++++++++++-
 2 files changed, 48 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
index 133c73207782..5cc7cba1941f 100644
--- a/drivers/pci/Kconfig
+++ b/drivers/pci/Kconfig
@@ -164,6 +164,11 @@ config PCI_PASID
 config PCI_P2PDMA
 	bool "PCI peer-to-peer transfer support"
 	depends on ZONE_DEVICE
+	#
+	# The need for the scatterlist DMA bus address flag means PCI P2PDMA
+	# requires 64bit
+	#
+	depends on 64BIT
 	select GENERIC_ALLOCATOR
 	help
 	  Enableѕ drivers to do PCI peer-to-peer transactions to and from
diff --git a/include/linux/scatterlist.h b/include/linux/scatterlist.h
index 7ff9d6386c12..6561ca8aead8 100644
--- a/include/linux/scatterlist.h
+++ b/include/linux/scatterlist.h
@@ -64,12 +64,24 @@ struct sg_append_table {
 #define SG_CHAIN	0x01UL
 #define SG_END		0x02UL
 
+/*
+ * bit 2 is the third free bit in the page_link on 64bit systems which
+ * is used by dma_unmap_sg() to determine if the dma_address is a
+ * bus address when doing P2PDMA.
+ */
+#ifdef CONFIG_PCI_P2PDMA
+#define SG_DMA_BUS_ADDRESS	0x04UL
+static_assert(__alignof__(struct page) >= 8);
+#else
+#define SG_DMA_BUS_ADDRESS	0x00UL
+#endif
+
 /*
  * We overload the LSB of the page pointer to indicate whether it's
  * a valid sg entry, or whether it points to the start of a new scatterlist.
  * Those low bits are there for everyone! (thanks mason :-)
  */
-#define SG_PAGE_LINK_MASK (SG_CHAIN | SG_END)
+#define SG_PAGE_LINK_MASK (SG_CHAIN | SG_END | SG_DMA_BUS_ADDRESS)
 
 static inline unsigned int __sg_flags(struct scatterlist *sg)
 {
@@ -91,6 +103,11 @@ static inline bool sg_is_last(struct scatterlist *sg)
 	return __sg_flags(sg) & SG_END;
 }
 
+static inline bool sg_is_dma_bus_address(struct scatterlist *sg)
+{
+	return __sg_flags(sg) & SG_DMA_BUS_ADDRESS;
+}
+
 /**
  * sg_assign_page - Assign a given page to an SG entry
  * @sg:		    SG entry
@@ -245,6 +262,31 @@ static inline void sg_unmark_end(struct scatterlist *sg)
 	sg->page_link &= ~SG_END;
 }
 
+/**
+ * sg_dma_mark_bus address - Mark the scatterlist entry as a bus address
+ * @sg:		 SG entryScatterlist
+ *
+ * Description:
+ *   Marks the passed in sg entry to indicate that the dma_address is
+ *   a bus address and doesn't need to be unmapped.
+ **/
+static inline void sg_dma_mark_bus_address(struct scatterlist *sg)
+{
+	sg->page_link |= SG_DMA_BUS_ADDRESS;
+}
+
+/**
+ * sg_unmark_pci_p2pdma - Unmark the scatterlist entry as a bus address
+ * @sg:		 SG entryScatterlist
+ *
+ * Description:
+ *   Clears the bus address mark.
+ **/
+static inline void sg_dma_unmark_bus_address(struct scatterlist *sg)
+{
+	sg->page_link &= ~SG_DMA_BUS_ADDRESS;
+}
+
 /**
  * sg_phys - Return physical address of an sg entry
  * @sg:	     SG entry
-- 
2.30.2

[PATCH v6 02/21] PCI/P2PDMA: Attempt to set map_type if it has not been set

[Logan Gunthorpe]

Attempt to find the mapping type for P2PDMA pages on the first
DMA map attempt if it has not been done ahead of time.

Previously, the mapping type was expected to be calculated ahead of
time, but if pages are to come from userspace then there's no
way to ensure the path was checked ahead of time.

This change will calculate the mapping type if it hasn't pre-calculated
so it is no longer invalid to call pci_p2pdma_map_sg() before the mapping
type is calculated, so drop the WARN_ON when that is the case.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
---
 drivers/pci/p2pdma.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
index 30b1df3c9d2f..c3a68e82cf36 100644
--- a/drivers/pci/p2pdma.c
+++ b/drivers/pci/p2pdma.c
@@ -849,6 +849,7 @@ static enum pci_p2pdma_map_type pci_p2pdma_map_type(struct dev_pagemap *pgmap,
 	struct pci_dev *provider = to_p2p_pgmap(pgmap)->provider;
 	struct pci_dev *client;
 	struct pci_p2pdma *p2pdma;
+	int dist;
 
 	if (!provider->p2pdma)
 		return PCI_P2PDMA_MAP_NOT_SUPPORTED;
@@ -865,6 +866,10 @@ static enum pci_p2pdma_map_type pci_p2pdma_map_type(struct dev_pagemap *pgmap,
 		type = xa_to_value(xa_load(&p2pdma->map_types,
 					   map_types_idx(client)));
 	rcu_read_unlock();
+
+	if (type == PCI_P2PDMA_MAP_UNKNOWN)
+		return calc_map_type_and_dist(provider, client, &dist, true);
+
 	return type;
 }
 
@@ -907,7 +912,7 @@ int pci_p2pdma_map_sg_attrs(struct device *dev, struct scatterlist *sg,
 	case PCI_P2PDMA_MAP_BUS_ADDR:
 		return __pci_p2pdma_map_sg(p2p_pgmap, dev, sg, nents);
 	default:
-		WARN_ON_ONCE(1);
+		/* Mapping is not Supported */
 		return 0;
 	}
 }
-- 
2.30.2

[PATCH v6 03/21] PCI/P2PDMA: Expose pci_p2pdma_map_type()

[Logan Gunthorpe]

pci_p2pdma_map_type() will be needed by the dma-iommu map_sg
implementation because it will need to determine the mapping type
ahead of actually doing the mapping to create the actual IOMMU mapping.

Prototypes for this helper are added to dma-map-ops.h as they are only
useful to dma map implementations and don't need to pollute the public
pci-p2pdma header

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
---
 drivers/pci/p2pdma.c        | 25 +++++++++++++--------
 include/linux/dma-map-ops.h | 45 +++++++++++++++++++++++++++++++++++++
 2 files changed, 61 insertions(+), 9 deletions(-)

diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
index c3a68e82cf36..8573bf9d651a 100644
--- a/drivers/pci/p2pdma.c
+++ b/drivers/pci/p2pdma.c
@@ -10,6 +10,7 @@
 
 #define pr_fmt(fmt) "pci-p2pdma: " fmt
 #include <linux/ctype.h>
+#include <linux/dma-map-ops.h>
 #include <linux/pci-p2pdma.h>
 #include <linux/module.h>
 #include <linux/slab.h>
@@ -20,13 +21,6 @@
 #include <linux/seq_buf.h>
 #include <linux/xarray.h>
 
-enum pci_p2pdma_map_type {
-	PCI_P2PDMA_MAP_UNKNOWN = 0,
-	PCI_P2PDMA_MAP_NOT_SUPPORTED,
-	PCI_P2PDMA_MAP_BUS_ADDR,
-	PCI_P2PDMA_MAP_THRU_HOST_BRIDGE,
-};
-
 struct pci_p2pdma {
 	struct gen_pool *pool;
 	bool p2pmem_published;
@@ -842,8 +836,21 @@ void pci_p2pmem_publish(struct pci_dev *pdev, bool publish)
 }
 EXPORT_SYMBOL_GPL(pci_p2pmem_publish);
 
-static enum pci_p2pdma_map_type pci_p2pdma_map_type(struct dev_pagemap *pgmap,
-						    struct device *dev)
+/**
+ * pci_p2pdma_map_type - return the type of mapping that should be used for
+ *	a given device and pgmap
+ * @pgmap: the pagemap of a page to determine the mapping type for
+ * @dev: device that is mapping the page
+ *
+ * Returns one of:
+ *	PCI_P2PDMA_MAP_NOT_SUPPORTED - The mapping should not be done
+ *	PCI_P2PDMA_MAP_BUS_ADDR - The mapping should use the PCI bus address
+ *	PCI_P2PDMA_MAP_THRU_HOST_BRIDGE - The mapping should be done normally
+ *		using the CPU physical address (in dma-direct) or an IOVA
+ *		mapping for the IOMMU.
+ */
+enum pci_p2pdma_map_type pci_p2pdma_map_type(struct dev_pagemap *pgmap,
+					     struct device *dev)
 {
 	enum pci_p2pdma_map_type type = PCI_P2PDMA_MAP_NOT_SUPPORTED;
 	struct pci_dev *provider = to_p2p_pgmap(pgmap)->provider;
diff --git a/include/linux/dma-map-ops.h b/include/linux/dma-map-ops.h
index 0d5b06b3a4a6..d693a0e33bac 100644
--- a/include/linux/dma-map-ops.h
+++ b/include/linux/dma-map-ops.h
@@ -379,4 +379,49 @@ static inline void debug_dma_dump_mappings(struct device *dev)
 
 extern const struct dma_map_ops dma_dummy_ops;
 
+enum pci_p2pdma_map_type {
+	/*
+	 * PCI_P2PDMA_MAP_UNKNOWN: Used internally for indicating the mapping
+	 * type hasn't been calculated yet. Functions that return this enum
+	 * never return this value.
+	 */
+	PCI_P2PDMA_MAP_UNKNOWN = 0,
+
+	/*
+	 * PCI_P2PDMA_MAP_NOT_SUPPORTED: Indicates the transaction will
+	 * traverse the host bridge and the host bridge is not in the
+	 * allowlist. DMA Mapping routines should return an error when
+	 * this is returned.
+	 */
+	PCI_P2PDMA_MAP_NOT_SUPPORTED,
+
+	/*
+	 * PCI_P2PDMA_BUS_ADDR: Indicates that two devices can talk to
+	 * each other directly through a PCI switch and the transaction will
+	 * not traverse the host bridge. Such a mapping should program
+	 * the DMA engine with PCI bus addresses.
+	 */
+	PCI_P2PDMA_MAP_BUS_ADDR,
+
+	/*
+	 * PCI_P2PDMA_MAP_THRU_HOST_BRIDGE: Indicates two devices can talk
+	 * to each other, but the transaction traverses a host bridge on the
+	 * allowlist. In this case, a normal mapping either with CPU physical
+	 * addresses (in the case of dma-direct) or IOVA addresses (in the
+	 * case of IOMMUs) should be used to program the DMA engine.
+	 */
+	PCI_P2PDMA_MAP_THRU_HOST_BRIDGE,
+};
+
+#ifdef CONFIG_PCI_P2PDMA
+enum pci_p2pdma_map_type pci_p2pdma_map_type(struct dev_pagemap *pgmap,
+					     struct device *dev);
+#else /* CONFIG_PCI_P2PDMA */
+static inline enum pci_p2pdma_map_type
+pci_p2pdma_map_type(struct dev_pagemap *pgmap, struct device *dev)
+{
+	return PCI_P2PDMA_MAP_NOT_SUPPORTED;
+}
+#endif /* CONFIG_PCI_P2PDMA */
+
 #endif /* _LINUX_DMA_MAP_OPS_H */
-- 
2.30.2

[PATCH v6 04/21] PCI/P2PDMA: Introduce helpers for dma_map_sg implementations

[Logan Gunthorpe]

Add pci_p2pdma_map_segment() as a helper for simple dma_map_sg()
implementations. It takes an scatterlist segment that must point to a
pci_p2pdma struct page and will map it if the mapping requires a bus
address.

The return value indicates whether the mapping required a bus address
or whether the caller still needs to map the segment normally. If the
segment should not be mapped, -EREMOTEIO is returned.

This helper uses a state structure to track the changes to the
pgmap across calls and avoid needing to lookup into the xarray for
every page.

Also add pci_p2pdma_map_bus_segment() which is useful for IOMMU
dma_map_sg() implementations where the sg segment containing the page
differs from the sg segment containing the DMA address.

Prototypes for these helpers are added to dma-map-ops.h as they are only
useful to dma map implementations and don't need to pollute the public
pci-p2pdma header.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
---
 drivers/pci/p2pdma.c        | 59 +++++++++++++++++++++++++++++++++++++
 include/linux/dma-map-ops.h | 21 +++++++++++++
 2 files changed, 80 insertions(+)

diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
index 8573bf9d651a..9032c2ed2cdf 100644
--- a/drivers/pci/p2pdma.c
+++ b/drivers/pci/p2pdma.c
@@ -946,6 +946,65 @@ void pci_p2pdma_unmap_sg_attrs(struct device *dev, struct scatterlist *sg,
 }
 EXPORT_SYMBOL_GPL(pci_p2pdma_unmap_sg_attrs);
 
+/**
+ * pci_p2pdma_map_segment - map an sg segment determining the mapping type
+ * @state: State structure that should be declared outside of the for_each_sg()
+ *	loop and initialized to zero.
+ * @dev: DMA device that's doing the mapping operation
+ * @sg: scatterlist segment to map
+ *
+ * This is a helper to be used by non-IOMMU dma_map_sg() implementations where
+ * the sg segment is the same for the page_link and the dma_address.
+ *
+ * Attempt to map a single segment in an SGL with the PCI bus address.
+ * The segment must point to a PCI P2PDMA page and thus must be
+ * wrapped in a is_pci_p2pdma_page(sg_page(sg)) check.
+ *
+ * Returns the type of mapping used and maps the page if the type is
+ * PCI_P2PDMA_MAP_BUS_ADDR.
+ */
+enum pci_p2pdma_map_type
+pci_p2pdma_map_segment(struct pci_p2pdma_map_state *state, struct device *dev,
+		       struct scatterlist *sg)
+{
+	if (state->pgmap != sg_page(sg)->pgmap) {
+		state->pgmap = sg_page(sg)->pgmap;
+		state->map = pci_p2pdma_map_type(state->pgmap, dev);
+		state->bus_off = to_p2p_pgmap(state->pgmap)->bus_offset;
+	}
+
+	if (state->map == PCI_P2PDMA_MAP_BUS_ADDR) {
+		sg->dma_address = sg_phys(sg) + state->bus_off;
+		sg_dma_len(sg) = sg->length;
+		sg_dma_mark_bus_address(sg);
+	}
+
+	return state->map;
+}
+
+/**
+ * pci_p2pdma_map_bus_segment - map an sg segment pre determined to
+ *	be mapped with PCI_P2PDMA_MAP_BUS_ADDR
+ * @pg_sg: scatterlist segment with the page to map
+ * @dma_sg: scatterlist segment to assign a DMA address to
+ *
+ * This is a helper for iommu dma_map_sg() implementations when the
+ * segment for the DMA address differs from the segment containing the
+ * source page.
+ *
+ * pci_p2pdma_map_type() must have already been called on the pg_sg and
+ * returned PCI_P2PDMA_MAP_BUS_ADDR.
+ */
+void pci_p2pdma_map_bus_segment(struct scatterlist *pg_sg,
+				struct scatterlist *dma_sg)
+{
+	struct pci_p2pdma_pagemap *pgmap = to_p2p_pgmap(sg_page(pg_sg)->pgmap);
+
+	dma_sg->dma_address = sg_phys(pg_sg) + pgmap->bus_offset;
+	sg_dma_len(dma_sg) = pg_sg->length;
+	sg_dma_mark_bus_address(dma_sg);
+}
+
 /**
  * pci_p2pdma_enable_store - parse a configfs/sysfs attribute store
  *		to enable p2pdma
diff --git a/include/linux/dma-map-ops.h b/include/linux/dma-map-ops.h
index d693a0e33bac..752f91e5eb5d 100644
--- a/include/linux/dma-map-ops.h
+++ b/include/linux/dma-map-ops.h
@@ -413,15 +413,36 @@ enum pci_p2pdma_map_type {
 	PCI_P2PDMA_MAP_THRU_HOST_BRIDGE,
 };
 
+struct pci_p2pdma_map_state {
+	struct dev_pagemap *pgmap;
+	int map;
+	u64 bus_off;
+};
+
 #ifdef CONFIG_PCI_P2PDMA
 enum pci_p2pdma_map_type pci_p2pdma_map_type(struct dev_pagemap *pgmap,
 					     struct device *dev);
+enum pci_p2pdma_map_type
+pci_p2pdma_map_segment(struct pci_p2pdma_map_state *state, struct device *dev,
+		       struct scatterlist *sg);
+void pci_p2pdma_map_bus_segment(struct scatterlist *pg_sg,
+				struct scatterlist *dma_sg);
 #else /* CONFIG_PCI_P2PDMA */
 static inline enum pci_p2pdma_map_type
 pci_p2pdma_map_type(struct dev_pagemap *pgmap, struct device *dev)
 {
 	return PCI_P2PDMA_MAP_NOT_SUPPORTED;
 }
+static inline enum pci_p2pdma_map_type
+pci_p2pdma_map_segment(struct pci_p2pdma_map_state *state, struct device *dev,
+		       struct scatterlist *sg)
+{
+	return PCI_P2PDMA_MAP_NOT_SUPPORTED;
+}
+static inline void pci_p2pdma_map_bus_segment(struct scatterlist *pg_sg,
+					      struct scatterlist *dma_sg)
+{
+}
 #endif /* CONFIG_PCI_P2PDMA */
 
 #endif /* _LINUX_DMA_MAP_OPS_H */
-- 
2.30.2

[PATCH v6 05/21] dma-mapping: allow EREMOTEIO return code for P2PDMA transfers

[Logan Gunthorpe]

Add EREMOTEIO error return to dma_map_sgtable() which will be used
by .map_sg() implementations that detect P2PDMA pages that the
underlying DMA device cannot access.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
---
 kernel/dma/mapping.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
index db7244291b74..9f65d1041638 100644
--- a/kernel/dma/mapping.c
+++ b/kernel/dma/mapping.c
@@ -197,7 +197,7 @@ static int __dma_map_sg_attrs(struct device *dev, struct scatterlist *sg,
 	if (ents > 0)
 		debug_dma_map_sg(dev, sg, nents, ents, dir, attrs);
 	else if (WARN_ON_ONCE(ents != -EINVAL && ents != -ENOMEM &&
-			      ents != -EIO))
+			      ents != -EIO && ents != -EREMOTEIO))
 		return -EIO;
 
 	return ents;
@@ -255,6 +255,8 @@ EXPORT_SYMBOL(dma_map_sg_attrs);
  *		complete the mapping. Should succeed if retried later.
  *   -EIO	Legacy error code with an unknown meaning. eg. this is
  *		returned if a lower level call returned DMA_MAPPING_ERROR.
+ *   -EREMOTEIO	The DMA device cannot access P2PDMA memory specified in
+ *		the sg_table. This will not succeed if retried.
  */
 int dma_map_sgtable(struct device *dev, struct sg_table *sgt,
 		    enum dma_data_direction dir, unsigned long attrs)
-- 
2.30.2

[PATCH v6 06/21] dma-direct: support PCI P2PDMA pages in dma-direct map_sg

[Logan Gunthorpe]

Add PCI P2PDMA support for dma_direct_map_sg() so that it can map
PCI P2PDMA pages directly without a hack in the callers. This allows
for heterogeneous SGLs that contain both P2PDMA and regular pages.

A P2PDMA page may have three possible outcomes when being mapped:
  1) If the data path between the two devices doesn't go through the
     root port, then it should be mapped with a PCI bus address
  2) If the data path goes through the host bridge, it should be mapped
     normally, as though it were a CPU physical address
  3) It is not possible for the two devices to communicate and thus
     the mapping operation should fail (and it will return -EREMOTEIO).

SGL segments that contain PCI bus addresses are marked with
sg_dma_mark_pci_p2pdma() and are ignored when unmapped.

P2PDMA mappings are also failed if swiotlb needs to be used on the
mapping.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 kernel/dma/direct.c | 43 +++++++++++++++++++++++++++++++++++++------
 kernel/dma/direct.h |  8 +++++++-
 2 files changed, 44 insertions(+), 7 deletions(-)

diff --git a/kernel/dma/direct.c b/kernel/dma/direct.c
index 9743c6ccce1a..33b838a3ccb2 100644
--- a/kernel/dma/direct.c
+++ b/kernel/dma/direct.c
@@ -461,29 +461,60 @@ void dma_direct_sync_sg_for_cpu(struct device *dev,
 		arch_sync_dma_for_cpu_all();
 }
 
+/*
+ * Unmaps segments, except for ones marked as pci_p2pdma which do not
+ * require any further action as they contain a bus address.
+ */
 void dma_direct_unmap_sg(struct device *dev, struct scatterlist *sgl,
 		int nents, enum dma_data_direction dir, unsigned long attrs)
 {
 	struct scatterlist *sg;
 	int i;
 
-	for_each_sg(sgl, sg, nents, i)
-		dma_direct_unmap_page(dev, sg->dma_address, sg_dma_len(sg), dir,
-			     attrs);
+	for_each_sg(sgl,  sg, nents, i) {
+		if (sg_is_dma_bus_address(sg))
+			sg_dma_unmark_bus_address(sg);
+		else
+			dma_direct_unmap_page(dev, sg->dma_address,
+					      sg_dma_len(sg), dir, attrs);
+	}
 }
 #endif
 
 int dma_direct_map_sg(struct device *dev, struct scatterlist *sgl, int nents,
 		enum dma_data_direction dir, unsigned long attrs)
 {
-	int i;
+	struct pci_p2pdma_map_state p2pdma_state = {};
+	enum pci_p2pdma_map_type map;
 	struct scatterlist *sg;
+	int i, ret;
 
 	for_each_sg(sgl, sg, nents, i) {
+		if (is_pci_p2pdma_page(sg_page(sg))) {
+			map = pci_p2pdma_map_segment(&p2pdma_state, dev, sg);
+			switch (map) {
+			case PCI_P2PDMA_MAP_BUS_ADDR:
+				continue;
+			case PCI_P2PDMA_MAP_THRU_HOST_BRIDGE:
+				/*
+				 * Any P2P mapping that traverses the PCI
+				 * host bridge must be mapped with CPU physical
+				 * address and not PCI bus addresses. This is
+				 * done with dma_direct_map_page() below.
+				 */
+				break;
+			default:
+				ret = -EREMOTEIO;
+				goto out_unmap;
+			}
+		}
+
 		sg->dma_address = dma_direct_map_page(dev, sg_page(sg),
 				sg->offset, sg->length, dir, attrs);
-		if (sg->dma_address == DMA_MAPPING_ERROR)
+		if (sg->dma_address == DMA_MAPPING_ERROR) {
+			ret = -EIO;
 			goto out_unmap;
+		}
 		sg_dma_len(sg) = sg->length;
 	}
 
@@ -491,7 +522,7 @@ int dma_direct_map_sg(struct device *dev, struct scatterlist *sgl, int nents,
 
 out_unmap:
 	dma_direct_unmap_sg(dev, sgl, i, dir, attrs | DMA_ATTR_SKIP_CPU_SYNC);
-	return -EIO;
+	return ret;
 }
 
 dma_addr_t dma_direct_map_resource(struct device *dev, phys_addr_t paddr,
diff --git a/kernel/dma/direct.h b/kernel/dma/direct.h
index 4632b0f4f72e..81b213409ce8 100644
--- a/kernel/dma/direct.h
+++ b/kernel/dma/direct.h
@@ -8,6 +8,7 @@
 #define _KERNEL_DMA_DIRECT_H
 
 #include <linux/dma-direct.h>
+#include <linux/memremap.h>
 
 int dma_direct_get_sgtable(struct device *dev, struct sg_table *sgt,
 		void *cpu_addr, dma_addr_t dma_addr, size_t size,
@@ -87,10 +88,15 @@ static inline dma_addr_t dma_direct_map_page(struct device *dev,
 	phys_addr_t phys = page_to_phys(page) + offset;
 	dma_addr_t dma_addr = phys_to_dma(dev, phys);
 
-	if (is_swiotlb_force_bounce(dev))
+	if (is_swiotlb_force_bounce(dev)) {
+		if (is_pci_p2pdma_page(page))
+			return DMA_MAPPING_ERROR;
 		return swiotlb_map(dev, phys, size, dir, attrs);
+	}
 
 	if (unlikely(!dma_capable(dev, dma_addr, size, true))) {
+		if (is_pci_p2pdma_page(page))
+			return DMA_MAPPING_ERROR;
 		if (swiotlb_force != SWIOTLB_NO_FORCE)
 			return swiotlb_map(dev, phys, size, dir, attrs);
 
-- 
2.30.2

[PATCH v6 07/21] dma-mapping: add flags to dma_map_ops to indicate PCI P2PDMA support

[Logan Gunthorpe]

Add a flags member to the dma_map_ops structure with one flag to
indicate support for PCI P2PDMA.

Also, add a helper to check if a device supports PCI P2PDMA.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
---
 include/linux/dma-map-ops.h | 10 ++++++++++
 include/linux/dma-mapping.h |  5 +++++
 kernel/dma/mapping.c        | 18 ++++++++++++++++++
 3 files changed, 33 insertions(+)

diff --git a/include/linux/dma-map-ops.h b/include/linux/dma-map-ops.h
index 752f91e5eb5d..4d4161d58ce0 100644
--- a/include/linux/dma-map-ops.h
+++ b/include/linux/dma-map-ops.h
@@ -11,7 +11,17 @@
 
 struct cma;
 
+/*
+ * Values for struct dma_map_ops.flags:
+ *
+ * DMA_F_PCI_P2PDMA_SUPPORTED: Indicates the dma_map_ops implementation can
+ * handle PCI P2PDMA pages in the map_sg/unmap_sg operation.
+ */
+#define DMA_F_PCI_P2PDMA_SUPPORTED     (1 << 0)
+
 struct dma_map_ops {
+	unsigned int flags;
+
 	void *(*alloc)(struct device *dev, size_t size,
 			dma_addr_t *dma_handle, gfp_t gfp,
 			unsigned long attrs);
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index dca2b1355bb1..f7c61b2b4b5e 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -140,6 +140,7 @@ int dma_mmap_attrs(struct device *dev, struct vm_area_struct *vma,
 		unsigned long attrs);
 bool dma_can_mmap(struct device *dev);
 int dma_supported(struct device *dev, u64 mask);
+bool dma_pci_p2pdma_supported(struct device *dev);
 int dma_set_mask(struct device *dev, u64 mask);
 int dma_set_coherent_mask(struct device *dev, u64 mask);
 u64 dma_get_required_mask(struct device *dev);
@@ -250,6 +251,10 @@ static inline int dma_supported(struct device *dev, u64 mask)
 {
 	return 0;
 }
+static inline bool dma_pci_p2pdma_supported(struct device *dev)
+{
+	return false;
+}
 static inline int dma_set_mask(struct device *dev, u64 mask)
 {
 	return -EIO;
diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
index 9f65d1041638..21793506fdb6 100644
--- a/kernel/dma/mapping.c
+++ b/kernel/dma/mapping.c
@@ -722,6 +722,24 @@ int dma_supported(struct device *dev, u64 mask)
 }
 EXPORT_SYMBOL(dma_supported);
 
+bool dma_pci_p2pdma_supported(struct device *dev)
+{
+	const struct dma_map_ops *ops = get_dma_ops(dev);
+
+	/* if ops is not set, dma direct will be used which supports P2PDMA */
+	if (!ops)
+		return true;
+
+	/*
+	 * Note: dma_ops_bypass is not checked here because P2PDMA should
+	 * not be used with dma mapping ops that do not have support even
+	 * if the specific device is bypassing them.
+	 */
+
+	return ops->flags & DMA_F_PCI_P2PDMA_SUPPORTED;
+}
+EXPORT_SYMBOL_GPL(dma_pci_p2pdma_supported);
+
 #ifdef CONFIG_ARCH_HAS_DMA_SET_MASK
 void arch_dma_set_mask(struct device *dev, u64 mask);
 #else
-- 
2.30.2

[PATCH v6 08/21] iommu/dma: support PCI P2PDMA pages in dma-iommu map_sg

[Logan Gunthorpe]

When a PCI P2PDMA page is seen, set the IOVA length of the segment
to zero so that it is not mapped into the IOVA. Then, in finalise_sg(),
apply the appropriate bus address to the segment. The IOVA is not
created if the scatterlist only consists of P2PDMA pages.

A P2PDMA page may have three possible outcomes when being mapped:
  1) If the data path between the two devices doesn't go through
     the root port, then it should be mapped with a PCI bus address
  2) If the data path goes through the host bridge, it should be mapped
     normally with an IOMMU IOVA.
  3) It is not possible for the two devices to communicate and thus
     the mapping operation should fail (and it will return -EREMOTEIO).

Similar to dma-direct, the sg_dma_mark_pci_p2pdma() flag is used to
indicate bus address segments. On unmap, P2PDMA segments are skipped
over when determining the start and end IOVA addresses.

With this change, the flags variable in the dma_map_ops is set to
DMA_F_PCI_P2PDMA_SUPPORTED to indicate support for P2PDMA pages.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
---
 drivers/iommu/dma-iommu.c | 68 +++++++++++++++++++++++++++++++++++----
 1 file changed, 61 insertions(+), 7 deletions(-)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 09f6e1c0f9c0..ef86f2b573d1 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -20,6 +20,7 @@
 #include <linux/iommu.h>
 #include <linux/iova.h>
 #include <linux/irq.h>
+#include <linux/memremap.h>
 #include <linux/mm.h>
 #include <linux/mutex.h>
 #include <linux/pci.h>
@@ -1045,6 +1046,16 @@ static int __finalise_sg(struct device *dev, struct scatterlist *sg, int nents,
 		sg_dma_address(s) = DMA_MAPPING_ERROR;
 		sg_dma_len(s) = 0;
 
+		if (is_pci_p2pdma_page(sg_page(s)) && !s_iova_len) {
+			if (i > 0)
+				cur = sg_next(cur);
+
+			pci_p2pdma_map_bus_segment(s, cur);
+			count++;
+			cur_len = 0;
+			continue;
+		}
+
 		/*
 		 * Now fill in the real DMA data. If...
 		 * - there is a valid output segment to append to
@@ -1141,6 +1152,8 @@ static int iommu_dma_map_sg(struct device *dev, struct scatterlist *sg,
 	struct iova_domain *iovad = &cookie->iovad;
 	struct scatterlist *s, *prev = NULL;
 	int prot = dma_info_to_prot(dir, dev_is_dma_coherent(dev), attrs);
+	struct dev_pagemap *pgmap = NULL;
+	enum pci_p2pdma_map_type map_type;
 	dma_addr_t iova;
 	size_t iova_len = 0;
 	unsigned long mask = dma_get_seg_boundary(dev);
@@ -1176,6 +1189,35 @@ static int iommu_dma_map_sg(struct device *dev, struct scatterlist *sg,
 		s_length = iova_align(iovad, s_length + s_iova_off);
 		s->length = s_length;
 
+		if (is_pci_p2pdma_page(sg_page(s))) {
+			if (sg_page(s)->pgmap != pgmap) {
+				pgmap = sg_page(s)->pgmap;
+				map_type = pci_p2pdma_map_type(pgmap, dev);
+			}
+
+			switch (map_type) {
+			case PCI_P2PDMA_MAP_BUS_ADDR:
+				/*
+				 * A zero length will be ignored by
+				 * iommu_map_sg() and then can be detected
+				 * in __finalise_sg() to actually map the
+				 * bus address.
+				 */
+				s->length = 0;
+				continue;
+			case PCI_P2PDMA_MAP_THRU_HOST_BRIDGE:
+				/*
+				 * Mapping through host bridge should be
+				 * mapped with regular IOVAs, thus we
+				 * do nothing here and continue below.
+				 */
+				break;
+			default:
+				ret = -EREMOTEIO;
+				goto out_restore_sg;
+			}
+		}
+
 		/*
 		 * Due to the alignment of our single IOVA allocation, we can
 		 * depend on these assumptions about the segment boundary mask:
@@ -1198,6 +1240,9 @@ static int iommu_dma_map_sg(struct device *dev, struct scatterlist *sg,
 		prev = s;
 	}
 
+	if (!iova_len)
+		return __finalise_sg(dev, sg, nents, 0);
+
 	iova = iommu_dma_alloc_iova(domain, iova_len, dma_get_mask(dev), dev);
 	if (!iova) {
 		ret = -ENOMEM;
@@ -1219,7 +1264,7 @@ static int iommu_dma_map_sg(struct device *dev, struct scatterlist *sg,
 out_restore_sg:
 	__invalidate_sg(sg, nents);
 out:
-	if (ret != -ENOMEM)
+	if (ret != -ENOMEM && ret != -EREMOTEIO)
 		return -EINVAL;
 	return ret;
 }
@@ -1227,7 +1272,7 @@ static int iommu_dma_map_sg(struct device *dev, struct scatterlist *sg,
 static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
 		int nents, enum dma_data_direction dir, unsigned long attrs)
 {
-	dma_addr_t start, end;
+	dma_addr_t end, start = DMA_MAPPING_ERROR;
 	struct scatterlist *tmp;
 	int i;
 
@@ -1243,14 +1288,22 @@ static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
 	 * The scatterlist segments are mapped into a single
 	 * contiguous IOVA allocation, so this is incredibly easy.
 	 */
-	start = sg_dma_address(sg);
-	for_each_sg(sg_next(sg), tmp, nents - 1, i) {
+	for_each_sg(sg, tmp, nents, i) {
+		if (sg_is_dma_bus_address(tmp)) {
+			sg_dma_unmark_bus_address(tmp);
+			continue;
+		}
 		if (sg_dma_len(tmp) == 0)
 			break;
-		sg = tmp;
+
+		if (start == DMA_MAPPING_ERROR)
+			start = sg_dma_address(tmp);
+
+		end = sg_dma_address(tmp) + sg_dma_len(tmp);
 	}
-	end = sg_dma_address(sg) + sg_dma_len(sg);
-	__iommu_dma_unmap(dev, start, end - start);
+
+	if (start != DMA_MAPPING_ERROR)
+		__iommu_dma_unmap(dev, start, end - start);
 }
 
 static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
@@ -1443,6 +1496,7 @@ static unsigned long iommu_dma_get_merge_boundary(struct device *dev)
 }
 
 static const struct dma_map_ops iommu_dma_ops = {
+	.flags			= DMA_F_PCI_P2PDMA_SUPPORTED,
 	.alloc			= iommu_dma_alloc,
 	.free			= iommu_dma_free,
 	.alloc_pages		= dma_common_alloc_pages,
-- 
2.30.2

[PATCH v6 09/21] nvme-pci: check DMA ops when indicating support for PCI P2PDMA

[Logan Gunthorpe]

Introduce a supports_pci_p2pdma() operation in nvme_ctrl_ops to
replace the fixed NVME_F_PCI_P2PDMA flag such that the dma_map_ops
flags can be checked for PCI P2PDMA support.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
---
 drivers/nvme/host/core.c |  3 ++-
 drivers/nvme/host/nvme.h |  2 +-
 drivers/nvme/host/pci.c  | 11 +++++++++--
 3 files changed, 12 insertions(+), 4 deletions(-)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index efb85c6d8e2d..bbc276dda49f 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -3912,7 +3912,8 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid,
 		blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, ns->queue);
 
 	blk_queue_flag_set(QUEUE_FLAG_NONROT, ns->queue);
-	if (ctrl->ops->flags & NVME_F_PCI_P2PDMA)
+	if (ctrl->ops->supports_pci_p2pdma &&
+	    ctrl->ops->supports_pci_p2pdma(ctrl))
 		blk_queue_flag_set(QUEUE_FLAG_PCI_P2PDMA, ns->queue);
 
 	ns->ctrl = ctrl;
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 1393bbf82d71..7d97bfb2a9e2 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -489,7 +489,6 @@ struct nvme_ctrl_ops {
 	unsigned int flags;
 #define NVME_F_FABRICS			(1 << 0)
 #define NVME_F_METADATA_SUPPORTED	(1 << 1)
-#define NVME_F_PCI_P2PDMA		(1 << 2)
 	int (*reg_read32)(struct nvme_ctrl *ctrl, u32 off, u32 *val);
 	int (*reg_write32)(struct nvme_ctrl *ctrl, u32 off, u32 val);
 	int (*reg_read64)(struct nvme_ctrl *ctrl, u32 off, u64 *val);
@@ -497,6 +496,7 @@ struct nvme_ctrl_ops {
 	void (*submit_async_event)(struct nvme_ctrl *ctrl);
 	void (*delete_ctrl)(struct nvme_ctrl *ctrl);
 	int (*get_address)(struct nvme_ctrl *ctrl, char *buf, int size);
+	bool (*supports_pci_p2pdma)(struct nvme_ctrl *ctrl);
 };
 
 /*
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index d817ca17463e..fec4c7191310 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -2969,17 +2969,24 @@ static int nvme_pci_get_address(struct nvme_ctrl *ctrl, char *buf, int size)
 	return snprintf(buf, size, "%s\n", dev_name(&pdev->dev));
 }
 
+static bool nvme_pci_supports_pci_p2pdma(struct nvme_ctrl *ctrl)
+{
+	struct nvme_dev *dev = to_nvme_dev(ctrl);
+
+	return dma_pci_p2pdma_supported(dev->dev);
+}
+
 static const struct nvme_ctrl_ops nvme_pci_ctrl_ops = {
 	.name			= "pcie",
 	.module			= THIS_MODULE,
-	.flags			= NVME_F_METADATA_SUPPORTED |
-				  NVME_F_PCI_P2PDMA,
+	.flags			= NVME_F_METADATA_SUPPORTED,
 	.reg_read32		= nvme_pci_reg_read32,
 	.reg_write32		= nvme_pci_reg_write32,
 	.reg_read64		= nvme_pci_reg_read64,
 	.free_ctrl		= nvme_pci_free_ctrl,
 	.submit_async_event	= nvme_pci_submit_async_event,
 	.get_address		= nvme_pci_get_address,
+	.supports_pci_p2pdma	= nvme_pci_supports_pci_p2pdma,
 };
 
 static int nvme_dev_map(struct nvme_dev *dev)
-- 
2.30.2

[PATCH v6 10/21] nvme-pci: convert to using dma_map_sgtable()

[Logan Gunthorpe]

The dma_map operations now support P2PDMA pages directly. So remove
the calls to pci_p2pdma_[un]map_sg_attrs() and replace them with calls
to dma_map_sgtable().

dma_map_sgtable() returns more complete error codes than dma_map_sg()
and allows differentiating EREMOTEIO errors in case an unsupported
P2PDMA transfer is requested. When this happens, return BLK_STS_TARGET
so the request isn't retried.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
---
 drivers/nvme/host/pci.c | 69 +++++++++++++++++------------------------
 1 file changed, 29 insertions(+), 40 deletions(-)

diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index fec4c7191310..07412116d4d1 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -230,11 +230,10 @@ struct nvme_iod {
 	bool use_sgl;
 	int aborted;
 	int npages;		/* In the PRP list. 0 means small pool in use */
-	int nents;		/* Used in scatterlist */
 	dma_addr_t first_dma;
 	unsigned int dma_len;	/* length of single DMA segment mapping */
 	dma_addr_t meta_dma;
-	struct scatterlist *sg;
+	struct sg_table sgt;
 };
 
 static inline unsigned int nvme_dbbuf_size(struct nvme_dev *dev)
@@ -524,7 +523,7 @@ static void nvme_commit_rqs(struct blk_mq_hw_ctx *hctx)
 static void **nvme_pci_iod_list(struct request *req)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
-	return (void **)(iod->sg + blk_rq_nr_phys_segments(req));
+	return (void **)(iod->sgt.sgl + blk_rq_nr_phys_segments(req));
 }
 
 static inline bool nvme_pci_use_sgls(struct nvme_dev *dev, struct request *req)
@@ -576,17 +575,6 @@ static void nvme_free_sgls(struct nvme_dev *dev, struct request *req)
 	}
 }
 
-static void nvme_unmap_sg(struct nvme_dev *dev, struct request *req)
-{
-	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
-
-	if (is_pci_p2pdma_page(sg_page(iod->sg)))
-		pci_p2pdma_unmap_sg(dev->dev, iod->sg, iod->nents,
-				    rq_dma_dir(req));
-	else
-		dma_unmap_sg(dev->dev, iod->sg, iod->nents, rq_dma_dir(req));
-}
-
 static void nvme_unmap_data(struct nvme_dev *dev, struct request *req)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
@@ -597,9 +585,10 @@ static void nvme_unmap_data(struct nvme_dev *dev, struct request *req)
 		return;
 	}
 
-	WARN_ON_ONCE(!iod->nents);
+	WARN_ON_ONCE(!iod->sgt.nents);
+
+	dma_unmap_sgtable(dev->dev, &iod->sgt, rq_dma_dir(req), 0);
 
-	nvme_unmap_sg(dev, req);
 	if (iod->npages == 0)
 		dma_pool_free(dev->prp_small_pool, nvme_pci_iod_list(req)[0],
 			      iod->first_dma);
@@ -607,7 +596,7 @@ static void nvme_unmap_data(struct nvme_dev *dev, struct request *req)
 		nvme_free_sgls(dev, req);
 	else
 		nvme_free_prps(dev, req);
-	mempool_free(iod->sg, dev->iod_mempool);
+	mempool_free(iod->sgt.sgl, dev->iod_mempool);
 }
 
 static void nvme_print_sgl(struct scatterlist *sgl, int nents)
@@ -630,7 +619,7 @@ static blk_status_t nvme_pci_setup_prps(struct nvme_dev *dev,
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	struct dma_pool *pool;
 	int length = blk_rq_payload_bytes(req);
-	struct scatterlist *sg = iod->sg;
+	struct scatterlist *sg = iod->sgt.sgl;
 	int dma_len = sg_dma_len(sg);
 	u64 dma_addr = sg_dma_address(sg);
 	int offset = dma_addr & (NVME_CTRL_PAGE_SIZE - 1);
@@ -703,16 +692,16 @@ static blk_status_t nvme_pci_setup_prps(struct nvme_dev *dev,
 		dma_len = sg_dma_len(sg);
 	}
 done:
-	cmnd->dptr.prp1 = cpu_to_le64(sg_dma_address(iod->sg));
+	cmnd->dptr.prp1 = cpu_to_le64(sg_dma_address(iod->sgt.sgl));
 	cmnd->dptr.prp2 = cpu_to_le64(iod->first_dma);
 	return BLK_STS_OK;
 free_prps:
 	nvme_free_prps(dev, req);
 	return BLK_STS_RESOURCE;
 bad_sgl:
-	WARN(DO_ONCE(nvme_print_sgl, iod->sg, iod->nents),
+	WARN(DO_ONCE(nvme_print_sgl, iod->sgt.sgl, iod->sgt.nents),
 			"Invalid SGL for payload:%d nents:%d\n",
-			blk_rq_payload_bytes(req), iod->nents);
+			blk_rq_payload_bytes(req), iod->sgt.nents);
 	return BLK_STS_IOERR;
 }
 
@@ -738,12 +727,13 @@ static void nvme_pci_sgl_set_seg(struct nvme_sgl_desc *sge,
 }
 
 static blk_status_t nvme_pci_setup_sgls(struct nvme_dev *dev,
-		struct request *req, struct nvme_rw_command *cmd, int entries)
+		struct request *req, struct nvme_rw_command *cmd)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	struct dma_pool *pool;
 	struct nvme_sgl_desc *sg_list;
-	struct scatterlist *sg = iod->sg;
+	struct scatterlist *sg = iod->sgt.sgl;
+	unsigned int entries = iod->sgt.nents;
 	dma_addr_t sgl_dma;
 	int i = 0;
 
@@ -841,7 +831,7 @@ static blk_status_t nvme_map_data(struct nvme_dev *dev, struct request *req,
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	blk_status_t ret = BLK_STS_RESOURCE;
-	int nr_mapped;
+	int rc;
 
 	if (blk_rq_nr_phys_segments(req) == 1) {
 		struct bio_vec bv = req_bvec(req);
@@ -859,26 +849,25 @@ static blk_status_t nvme_map_data(struct nvme_dev *dev, struct request *req,
 	}
 
 	iod->dma_len = 0;
-	iod->sg = mempool_alloc(dev->iod_mempool, GFP_ATOMIC);
-	if (!iod->sg)
+	iod->sgt.sgl = mempool_alloc(dev->iod_mempool, GFP_ATOMIC);
+	if (!iod->sgt.sgl)
 		return BLK_STS_RESOURCE;
-	sg_init_table(iod->sg, blk_rq_nr_phys_segments(req));
-	iod->nents = blk_rq_map_sg(req->q, req, iod->sg);
-	if (!iod->nents)
+	sg_init_table(iod->sgt.sgl, blk_rq_nr_phys_segments(req));
+	iod->sgt.orig_nents = blk_rq_map_sg(req->q, req, iod->sgt.sgl);
+	if (!iod->sgt.orig_nents)
 		goto out_free_sg;
 
-	if (is_pci_p2pdma_page(sg_page(iod->sg)))
-		nr_mapped = pci_p2pdma_map_sg_attrs(dev->dev, iod->sg,
-				iod->nents, rq_dma_dir(req), DMA_ATTR_NO_WARN);
-	else
-		nr_mapped = dma_map_sg_attrs(dev->dev, iod->sg, iod->nents,
-					     rq_dma_dir(req), DMA_ATTR_NO_WARN);
-	if (!nr_mapped)
+	rc = dma_map_sgtable(dev->dev, &iod->sgt, rq_dma_dir(req),
+			     DMA_ATTR_NO_WARN);
+	if (rc) {
+		if (rc == -EREMOTEIO)
+			ret = BLK_STS_TARGET;
 		goto out_free_sg;
+	}
 
 	iod->use_sgl = nvme_pci_use_sgls(dev, req);
 	if (iod->use_sgl)
-		ret = nvme_pci_setup_sgls(dev, req, &cmnd->rw, nr_mapped);
+		ret = nvme_pci_setup_sgls(dev, req, &cmnd->rw);
 	else
 		ret = nvme_pci_setup_prps(dev, req, &cmnd->rw);
 	if (ret != BLK_STS_OK)
@@ -886,9 +875,9 @@ static blk_status_t nvme_map_data(struct nvme_dev *dev, struct request *req,
 	return BLK_STS_OK;
 
 out_unmap_sg:
-	nvme_unmap_sg(dev, req);
+	dma_unmap_sgtable(dev->dev, &iod->sgt, rq_dma_dir(req), 0);
 out_free_sg:
-	mempool_free(iod->sg, dev->iod_mempool);
+	mempool_free(iod->sgt.sgl, dev->iod_mempool);
 	return ret;
 }
 
@@ -912,7 +901,7 @@ static blk_status_t nvme_prep_rq(struct nvme_dev *dev, struct request *req)
 
 	iod->aborted = 0;
 	iod->npages = -1;
-	iod->nents = 0;
+	iod->sgt.nents = 0;
 
 	ret = nvme_setup_cmd(req->q->queuedata, req);
 	if (ret)
-- 
2.30.2

[PATCH v6 11/21] RDMA/core: introduce ib_dma_pci_p2p_dma_supported()

[Logan Gunthorpe]

Introduce the helper function ib_dma_pci_p2p_dma_supported() to check
if a given ib_device can be used in P2PDMA transfers. This ensures
the ib_device is not using virt_dma and also that the underlying
dma_device supports P2PDMA.

Use the new helper in nvme-rdma to replace the existing check for
ib_uses_virt_dma(). Adding the dma_pci_p2pdma_supported() check allows
switching away from pci_p2pdma_[un]map_sg().

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Max Gurtovoy <mgurtovoy@nvidia.com>
---
 drivers/nvme/target/rdma.c |  2 +-
 include/rdma/ib_verbs.h    | 11 +++++++++++
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index 2fab0b219b25..12258f87ccc8 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -415,7 +415,7 @@ static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev,
 	if (ib_dma_mapping_error(ndev->device, r->send_sge.addr))
 		goto out_free_rsp;
 
-	if (!ib_uses_virt_dma(ndev->device))
+	if (ib_dma_pci_p2p_dma_supported(ndev->device))
 		r->req.p2p_client = &ndev->device->dev;
 	r->send_sge.length = sizeof(*r->req.cqe);
 	r->send_sge.lkey = ndev->pd->local_dma_lkey;
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 69d883f7fb41..79609ab73014 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -4003,6 +4003,17 @@ static inline bool ib_uses_virt_dma(struct ib_device *dev)
 	return IS_ENABLED(CONFIG_INFINIBAND_VIRT_DMA) && !dev->dma_device;
 }
 
+/*
+ * Check if a IB device's underlying DMA mapping supports P2PDMA transfers.
+ */
+static inline bool ib_dma_pci_p2p_dma_supported(struct ib_device *dev)
+{
+	if (ib_uses_virt_dma(dev))
+		return false;
+
+	return dma_pci_p2pdma_supported(dev->dma_device);
+}
+
 /**
  * ib_dma_mapping_error - check a DMA addr for error
  * @dev: The device for which the dma_addr was created
-- 
2.30.2

[PATCH v6 12/21] RDMA/rw: drop pci_p2pdma_[un]map_sg()

[Logan Gunthorpe]

dma_map_sg() now supports the use of P2PDMA pages so pci_p2pdma_map_sg()
is no longer necessary and may be dropped. This means the
rdma_rw_[un]map_sg() helpers are no longer necessary. Remove it all.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
---
 drivers/infiniband/core/rw.c | 45 ++++++++----------------------------
 1 file changed, 9 insertions(+), 36 deletions(-)

diff --git a/drivers/infiniband/core/rw.c b/drivers/infiniband/core/rw.c
index 4d98f931a13d..8367974b7998 100644
--- a/drivers/infiniband/core/rw.c
+++ b/drivers/infiniband/core/rw.c
@@ -274,33 +274,6 @@ static int rdma_rw_init_single_wr(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
 	return 1;
 }
 
-static void rdma_rw_unmap_sg(struct ib_device *dev, struct scatterlist *sg,
-			     u32 sg_cnt, enum dma_data_direction dir)
-{
-	if (is_pci_p2pdma_page(sg_page(sg)))
-		pci_p2pdma_unmap_sg(dev->dma_device, sg, sg_cnt, dir);
-	else
-		ib_dma_unmap_sg(dev, sg, sg_cnt, dir);
-}
-
-static int rdma_rw_map_sgtable(struct ib_device *dev, struct sg_table *sgt,
-			       enum dma_data_direction dir)
-{
-	int nents;
-
-	if (is_pci_p2pdma_page(sg_page(sgt->sgl))) {
-		if (WARN_ON_ONCE(ib_uses_virt_dma(dev)))
-			return 0;
-		nents = pci_p2pdma_map_sg(dev->dma_device, sgt->sgl,
-					  sgt->orig_nents, dir);
-		if (!nents)
-			return -EIO;
-		sgt->nents = nents;
-		return 0;
-	}
-	return ib_dma_map_sgtable_attrs(dev, sgt, dir, 0);
-}
-
 /**
  * rdma_rw_ctx_init - initialize a RDMA READ/WRITE context
  * @ctx:	context to initialize
@@ -327,7 +300,7 @@ int rdma_rw_ctx_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u32 port_num,
 	};
 	int ret;
 
-	ret = rdma_rw_map_sgtable(dev, &sgt, dir);
+	ret = ib_dma_map_sgtable_attrs(dev, &sgt, dir, 0);
 	if (ret)
 		return ret;
 	sg_cnt = sgt.nents;
@@ -366,7 +339,7 @@ int rdma_rw_ctx_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u32 port_num,
 	return ret;
 
 out_unmap_sg:
-	rdma_rw_unmap_sg(dev, sgt.sgl, sgt.orig_nents, dir);
+	ib_dma_unmap_sgtable_attrs(dev, &sgt, dir, 0);
 	return ret;
 }
 EXPORT_SYMBOL(rdma_rw_ctx_init);
@@ -414,12 +387,12 @@ int rdma_rw_ctx_signature_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
 		return -EINVAL;
 	}
 
-	ret = rdma_rw_map_sgtable(dev, &sgt, dir);
+	ret = ib_dma_map_sgtable_attrs(dev, &sgt, dir, 0);
 	if (ret)
 		return ret;
 
 	if (prot_sg_cnt) {
-		ret = rdma_rw_map_sgtable(dev, &prot_sgt, dir);
+		ret = ib_dma_map_sgtable_attrs(dev, &prot_sgt, dir, 0);
 		if (ret)
 			goto out_unmap_sg;
 	}
@@ -486,9 +459,9 @@ int rdma_rw_ctx_signature_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
 	kfree(ctx->reg);
 out_unmap_prot_sg:
 	if (prot_sgt.nents)
-		rdma_rw_unmap_sg(dev, prot_sgt.sgl, prot_sgt.orig_nents, dir);
+		ib_dma_unmap_sgtable_attrs(dev, &prot_sgt, dir, 0);
 out_unmap_sg:
-	rdma_rw_unmap_sg(dev, sgt.sgl, sgt.orig_nents, dir);
+	ib_dma_unmap_sgtable_attrs(dev, &sgt, dir, 0);
 	return ret;
 }
 EXPORT_SYMBOL(rdma_rw_ctx_signature_init);
@@ -621,7 +594,7 @@ void rdma_rw_ctx_destroy(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
 		break;
 	}
 
-	rdma_rw_unmap_sg(qp->pd->device, sg, sg_cnt, dir);
+	ib_dma_unmap_sg(qp->pd->device, sg, sg_cnt, dir);
 }
 EXPORT_SYMBOL(rdma_rw_ctx_destroy);
 
@@ -649,8 +622,8 @@ void rdma_rw_ctx_destroy_signature(struct rdma_rw_ctx *ctx, struct ib_qp *qp,
 	kfree(ctx->reg);
 
 	if (prot_sg_cnt)
-		rdma_rw_unmap_sg(qp->pd->device, prot_sg, prot_sg_cnt, dir);
-	rdma_rw_unmap_sg(qp->pd->device, sg, sg_cnt, dir);
+		ib_dma_unmap_sg(qp->pd->device, prot_sg, prot_sg_cnt, dir);
+	ib_dma_unmap_sg(qp->pd->device, sg, sg_cnt, dir);
 }
 EXPORT_SYMBOL(rdma_rw_ctx_destroy_signature);
 
-- 
2.30.2

[PATCH v6 13/21] PCI/P2PDMA: Remove pci_p2pdma_[un]map_sg()

[Logan Gunthorpe]

This interface is superseded by support in dma_map_sg() which now supports
heterogeneous scatterlists. There are no longer any users, so remove it.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Max Gurtovoy <mgurtovoy@nvidia.com>
---
 drivers/pci/p2pdma.c       | 66 --------------------------------------
 include/linux/pci-p2pdma.h | 27 ----------------
 2 files changed, 93 deletions(-)

diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
index 9032c2ed2cdf..4d3cab9da748 100644
--- a/drivers/pci/p2pdma.c
+++ b/drivers/pci/p2pdma.c
@@ -880,72 +880,6 @@ enum pci_p2pdma_map_type pci_p2pdma_map_type(struct dev_pagemap *pgmap,
 	return type;
 }
 
-static int __pci_p2pdma_map_sg(struct pci_p2pdma_pagemap *p2p_pgmap,
-		struct device *dev, struct scatterlist *sg, int nents)
-{
-	struct scatterlist *s;
-	int i;
-
-	for_each_sg(sg, s, nents, i) {
-		s->dma_address = sg_phys(s) + p2p_pgmap->bus_offset;
-		sg_dma_len(s) = s->length;
-	}
-
-	return nents;
-}
-
-/**
- * pci_p2pdma_map_sg_attrs - map a PCI peer-to-peer scatterlist for DMA
- * @dev: device doing the DMA request
- * @sg: scatter list to map
- * @nents: elements in the scatterlist
- * @dir: DMA direction
- * @attrs: DMA attributes passed to dma_map_sg() (if called)
- *
- * Scatterlists mapped with this function should be unmapped using
- * pci_p2pdma_unmap_sg_attrs().
- *
- * Returns the number of SG entries mapped or 0 on error.
- */
-int pci_p2pdma_map_sg_attrs(struct device *dev, struct scatterlist *sg,
-		int nents, enum dma_data_direction dir, unsigned long attrs)
-{
-	struct pci_p2pdma_pagemap *p2p_pgmap =
-		to_p2p_pgmap(sg_page(sg)->pgmap);
-
-	switch (pci_p2pdma_map_type(sg_page(sg)->pgmap, dev)) {
-	case PCI_P2PDMA_MAP_THRU_HOST_BRIDGE:
-		return dma_map_sg_attrs(dev, sg, nents, dir, attrs);
-	case PCI_P2PDMA_MAP_BUS_ADDR:
-		return __pci_p2pdma_map_sg(p2p_pgmap, dev, sg, nents);
-	default:
-		/* Mapping is not Supported */
-		return 0;
-	}
-}
-EXPORT_SYMBOL_GPL(pci_p2pdma_map_sg_attrs);
-
-/**
- * pci_p2pdma_unmap_sg_attrs - unmap a PCI peer-to-peer scatterlist that was
- *	mapped with pci_p2pdma_map_sg()
- * @dev: device doing the DMA request
- * @sg: scatter list to map
- * @nents: number of elements returned by pci_p2pdma_map_sg()
- * @dir: DMA direction
- * @attrs: DMA attributes passed to dma_unmap_sg() (if called)
- */
-void pci_p2pdma_unmap_sg_attrs(struct device *dev, struct scatterlist *sg,
-		int nents, enum dma_data_direction dir, unsigned long attrs)
-{
-	enum pci_p2pdma_map_type map_type;
-
-	map_type = pci_p2pdma_map_type(sg_page(sg)->pgmap, dev);
-
-	if (map_type == PCI_P2PDMA_MAP_THRU_HOST_BRIDGE)
-		dma_unmap_sg_attrs(dev, sg, nents, dir, attrs);
-}
-EXPORT_SYMBOL_GPL(pci_p2pdma_unmap_sg_attrs);
-
 /**
  * pci_p2pdma_map_segment - map an sg segment determining the mapping type
  * @state: State structure that should be declared outside of the for_each_sg()
diff --git a/include/linux/pci-p2pdma.h b/include/linux/pci-p2pdma.h
index 8318a97c9c61..2c07aa6b7665 100644
--- a/include/linux/pci-p2pdma.h
+++ b/include/linux/pci-p2pdma.h
@@ -30,10 +30,6 @@ struct scatterlist *pci_p2pmem_alloc_sgl(struct pci_dev *pdev,
 					 unsigned int *nents, u32 length);
 void pci_p2pmem_free_sgl(struct pci_dev *pdev, struct scatterlist *sgl);
 void pci_p2pmem_publish(struct pci_dev *pdev, bool publish);
-int pci_p2pdma_map_sg_attrs(struct device *dev, struct scatterlist *sg,
-		int nents, enum dma_data_direction dir, unsigned long attrs);
-void pci_p2pdma_unmap_sg_attrs(struct device *dev, struct scatterlist *sg,
-		int nents, enum dma_data_direction dir, unsigned long attrs);
 int pci_p2pdma_enable_store(const char *page, struct pci_dev **p2p_dev,
 			    bool *use_p2pdma);
 ssize_t pci_p2pdma_enable_show(char *page, struct pci_dev *p2p_dev,
@@ -83,17 +79,6 @@ static inline void pci_p2pmem_free_sgl(struct pci_dev *pdev,
 static inline void pci_p2pmem_publish(struct pci_dev *pdev, bool publish)
 {
 }
-static inline int pci_p2pdma_map_sg_attrs(struct device *dev,
-		struct scatterlist *sg, int nents, enum dma_data_direction dir,
-		unsigned long attrs)
-{
-	return 0;
-}
-static inline void pci_p2pdma_unmap_sg_attrs(struct device *dev,
-		struct scatterlist *sg, int nents, enum dma_data_direction dir,
-		unsigned long attrs)
-{
-}
 static inline int pci_p2pdma_enable_store(const char *page,
 		struct pci_dev **p2p_dev, bool *use_p2pdma)
 {
@@ -119,16 +104,4 @@ static inline struct pci_dev *pci_p2pmem_find(struct device *client)
 	return pci_p2pmem_find_many(&client, 1);
 }
 
-static inline int pci_p2pdma_map_sg(struct device *dev, struct scatterlist *sg,
-				    int nents, enum dma_data_direction dir)
-{
-	return pci_p2pdma_map_sg_attrs(dev, sg, nents, dir, 0);
-}
-
-static inline void pci_p2pdma_unmap_sg(struct device *dev,
-		struct scatterlist *sg, int nents, enum dma_data_direction dir)
-{
-	pci_p2pdma_unmap_sg_attrs(dev, sg, nents, dir, 0);
-}
-
 #endif /* _LINUX_PCI_P2P_H */
-- 
2.30.2

[PATCH v6 14/21] mm: introduce FOLL_PCI_P2PDMA to gate getting PCI P2PDMA pages

[Logan Gunthorpe]

GUP Callers that expect PCI P2PDMA pages can now set FOLL_PCI_P2PDMA to
allow obtaining P2PDMA pages. If GUP is called without the flag and a
P2PDMA page is found, it will return an error.

FOLL_PCI_P2PDMA cannot be set if FOLL_LONGTERM is set.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 include/linux/mm.h |  1 +
 mm/gup.c           | 22 +++++++++++++++++++++-
 2 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index e34edb775334..14ef41af8b77 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2936,6 +2936,7 @@ struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
 #define FOLL_SPLIT_PMD	0x20000	/* split huge pmd before returning */
 #define FOLL_PIN	0x40000	/* pages must be released via unpin_user_page */
 #define FOLL_FAST_ONLY	0x80000	/* gup_fast: prevent fall-back to slow gup */
+#define FOLL_PCI_P2PDMA	0x100000 /* allow returning PCI P2PDMA pages */
 
 /*
  * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
diff --git a/mm/gup.c b/mm/gup.c
index f598a037eb04..0af6f802ca38 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -490,6 +490,12 @@ static struct page *follow_page_pte(struct vm_area_struct *vma,
 			page = pte_page(pte);
 		else
 			goto no_page;
+
+		if (unlikely(!(flags & FOLL_PCI_P2PDMA) &&
+			     is_pci_p2pdma_page(page))) {
+			page = ERR_PTR(-EREMOTEIO);
+			goto out;
+		}
 	} else if (unlikely(!page)) {
 		if (flags & FOLL_DUMP) {
 			/* Avoid special (like zero) pages in core dumps */
@@ -919,6 +925,9 @@ static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
 	if ((gup_flags & FOLL_LONGTERM) && vma_is_fsdax(vma))
 		return -EOPNOTSUPP;
 
+	if ((gup_flags & FOLL_LONGTERM) && (gup_flags & FOLL_PCI_P2PDMA))
+		return -EOPNOTSUPP;
+
 	if (vma_is_secretmem(vma))
 		return -EFAULT;
 
@@ -2184,6 +2193,10 @@ static int gup_pte_range(pmd_t pmd, unsigned long addr, unsigned long end,
 		VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
 		page = pte_page(pte);
 
+		if (unlikely(pte_devmap(pte) && !(flags & FOLL_PCI_P2PDMA) &&
+			     is_pci_p2pdma_page(page)))
+			goto pte_unmap;
+
 		folio = try_grab_folio(page, 1, flags);
 		if (!folio)
 			goto pte_unmap;
@@ -2258,6 +2271,12 @@ static int __gup_device_huge(unsigned long pfn, unsigned long addr,
 			undo_dev_pagemap(nr, nr_start, flags, pages);
 			break;
 		}
+
+		if (!(flags & FOLL_PCI_P2PDMA) && is_pci_p2pdma_page(page)) {
+			undo_dev_pagemap(nr, nr_start, flags, pages);
+			break;
+		}
+
 		SetPageReferenced(page);
 		pages[*nr] = page;
 		if (unlikely(!try_grab_page(page, flags))) {
@@ -2729,7 +2748,8 @@ static int internal_get_user_pages_fast(unsigned long start,
 
 	if (WARN_ON_ONCE(gup_flags & ~(FOLL_WRITE | FOLL_LONGTERM |
 				       FOLL_FORCE | FOLL_PIN | FOLL_GET |
-				       FOLL_FAST_ONLY | FOLL_NOFAULT)))
+				       FOLL_FAST_ONLY | FOLL_NOFAULT |
+				       FOLL_PCI_P2PDMA)))
 		return -EINVAL;
 
 	if (gup_flags & FOLL_PIN)
-- 
2.30.2

[PATCH v6 15/21] iov_iter: introduce iov_iter_get_pages_[alloc_]flags()

[Logan Gunthorpe]

Add iov_iter_get_pages_flags() and iov_iter_get_pages_alloc_flags()
which take a flags argument that is passed to get_user_pages_fast().

This is so that FOLL_PCI_P2PDMA can be passed when appropriate.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 include/linux/uio.h |  6 ++++++
 lib/iov_iter.c      | 25 +++++++++++++++++++------
 2 files changed, 25 insertions(+), 6 deletions(-)

diff --git a/include/linux/uio.h b/include/linux/uio.h
index 739285fe5a2f..ddf9e4cf4a59 100644
--- a/include/linux/uio.h
+++ b/include/linux/uio.h
@@ -232,8 +232,14 @@ void iov_iter_pipe(struct iov_iter *i, unsigned int direction, struct pipe_inode
 void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count);
 void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray,
 		     loff_t start, size_t count);
+ssize_t iov_iter_get_pages_flags(struct iov_iter *i, struct page **pages,
+		size_t maxsize, unsigned maxpages, size_t *start,
+		unsigned int gup_flags);
 ssize_t iov_iter_get_pages(struct iov_iter *i, struct page **pages,
 			size_t maxsize, unsigned maxpages, size_t *start);
+ssize_t iov_iter_get_pages_alloc_flags(struct iov_iter *i,
+		struct page ***pages, size_t maxsize, size_t *start,
+		unsigned int gup_flags);
 ssize_t iov_iter_get_pages_alloc(struct iov_iter *i, struct page ***pages,
 			size_t maxsize, size_t *start);
 int iov_iter_npages(const struct iov_iter *i, int maxpages);
diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index 6dd5330f7a99..9bf6e3af5120 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -1515,9 +1515,9 @@ static struct page *first_bvec_segment(const struct iov_iter *i,
 	return page;
 }
 
-ssize_t iov_iter_get_pages(struct iov_iter *i,
+ssize_t iov_iter_get_pages_flags(struct iov_iter *i,
 		   struct page **pages, size_t maxsize, unsigned maxpages,
-		   size_t *start)
+		   size_t *start, unsigned int gup_flags)
 {
 	size_t len;
 	int n, res;
@@ -1528,7 +1528,6 @@ ssize_t iov_iter_get_pages(struct iov_iter *i,
 		return 0;
 
 	if (likely(iter_is_iovec(i))) {
-		unsigned int gup_flags = 0;
 		unsigned long addr;
 
 		if (iov_iter_rw(i) != WRITE)
@@ -1558,6 +1557,13 @@ ssize_t iov_iter_get_pages(struct iov_iter *i,
 		return iter_xarray_get_pages(i, pages, maxsize, maxpages, start);
 	return -EFAULT;
 }
+EXPORT_SYMBOL_GPL(iov_iter_get_pages_flags);
+
+ssize_t iov_iter_get_pages(struct iov_iter *i, struct page **pages,
+			   size_t maxsize, unsigned maxpages, size_t *start)
+{
+	return iov_iter_get_pages_flags(i, pages, maxsize, maxpages, start, 0);
+}
 EXPORT_SYMBOL(iov_iter_get_pages);
 
 static struct page **get_pages_array(size_t n)
@@ -1640,9 +1646,9 @@ static ssize_t iter_xarray_get_pages_alloc(struct iov_iter *i,
 	return actual;
 }
 
-ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
+ssize_t iov_iter_get_pages_alloc_flags(struct iov_iter *i,
 		   struct page ***pages, size_t maxsize,
-		   size_t *start)
+		   size_t *start, unsigned int gup_flags)
 {
 	struct page **p;
 	size_t len;
@@ -1654,7 +1660,6 @@ ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
 		return 0;
 
 	if (likely(iter_is_iovec(i))) {
-		unsigned int gup_flags = 0;
 		unsigned long addr;
 
 		if (iov_iter_rw(i) != WRITE)
@@ -1667,6 +1672,7 @@ ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
 		p = get_pages_array(n);
 		if (!p)
 			return -ENOMEM;
+
 		res = get_user_pages_fast(addr, n, gup_flags, p);
 		if (unlikely(res <= 0)) {
 			kvfree(p);
@@ -1694,6 +1700,13 @@ ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
 		return iter_xarray_get_pages_alloc(i, pages, maxsize, start);
 	return -EFAULT;
 }
+EXPORT_SYMBOL_GPL(iov_iter_get_pages_alloc_flags);
+
+ssize_t iov_iter_get_pages_alloc(struct iov_iter *i, struct page ***pages,
+				 size_t maxsize, size_t *start)
+{
+	return iov_iter_get_pages_alloc_flags(i, pages, maxsize, start, 0);
+}
 EXPORT_SYMBOL(iov_iter_get_pages_alloc);
 
 size_t csum_and_copy_from_iter(void *addr, size_t bytes, __wsum *csum,
-- 
2.30.2

[PATCH v6 16/21] block: add check when merging zone device pages

[Logan Gunthorpe]

Consecutive zone device pages should not be merged into the same sgl
or bvec segment with other types of pages or if they belong to different
pgmaps. Otherwise getting the pgmap of a given segment is not possible
without scanning the entire segment. This helper returns true either if
both pages are not zone device pages or both pages are zone device
pages with the same pgmap.

Add a helper to determine if zone device pages are mergeable and use
this helper in page_is_mergeable().

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 block/bio.c        |  2 ++
 include/linux/mm.h | 23 +++++++++++++++++++++++
 2 files changed, 25 insertions(+)

diff --git a/block/bio.c b/block/bio.c
index cdd7b2915c53..3406c0450db3 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -834,6 +834,8 @@ static inline bool page_is_mergeable(const struct bio_vec *bv,
 		return false;
 	if (xen_domain() && !xen_biovec_phys_mergeable(bv, page))
 		return false;
+	if (!zone_device_pages_have_same_pgmap(bv->bv_page, page))
+		return false;
 
 	*same_page = ((vec_end_addr & PAGE_MASK) == page_addr);
 	if (*same_page)
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 14ef41af8b77..fb2264a17e4a 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1108,6 +1108,24 @@ static inline bool is_zone_device_page(const struct page *page)
 {
 	return page_zonenum(page) == ZONE_DEVICE;
 }
+
+/*
+ * Consecutive zone device pages should not be merged into the same sgl
+ * or bvec segment with other types of pages or if they belong to different
+ * pgmaps. Otherwise getting the pgmap of a given segment is not possible
+ * without scanning the entire segment. This helper returns true either if
+ * both pages are not zone device pages or both pages are zone device pages
+ * with the same pgmap.
+ */
+static inline bool zone_device_pages_have_same_pgmap(const struct page *a,
+						     const struct page *b)
+{
+	if (is_zone_device_page(a) != is_zone_device_page(b))
+		return false;
+	if (!is_zone_device_page(a))
+		return true;
+	return a->pgmap == b->pgmap;
+}
 extern void memmap_init_zone_device(struct zone *, unsigned long,
 				    unsigned long, struct dev_pagemap *);
 #else
@@ -1115,6 +1133,11 @@ static inline bool is_zone_device_page(const struct page *page)
 {
 	return false;
 }
+static inline bool zone_device_pages_have_same_pgmap(const struct page *a,
+						     const struct page *b)
+{
+	return true;
+}
 #endif
 
 static inline bool folio_is_zone_device(const struct folio *folio)
-- 
2.30.2

[PATCH v6 17/21] lib/scatterlist: add check when merging zone device pages

[Logan Gunthorpe]

Consecutive zone device pages should not be merged into the same sgl
or bvec segment with other types of pages or if they belong to different
pgmaps. Otherwise getting the pgmap of a given segment is not possible
without scanning the entire segment. This helper returns true either if
both pages are not zone device pages or both pages are zone device
pages with the same pgmap.

Factor out the check for page mergability into a pages_are_mergable()
helper and add a check with zone_device_pages_are_mergeable().

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 lib/scatterlist.c | 25 +++++++++++++++----------
 1 file changed, 15 insertions(+), 10 deletions(-)

diff --git a/lib/scatterlist.c b/lib/scatterlist.c
index d5e82e4a57ad..af53a0984f76 100644
--- a/lib/scatterlist.c
+++ b/lib/scatterlist.c
@@ -410,6 +410,15 @@ static struct scatterlist *get_next_sg(struct sg_append_table *table,
 	return new_sg;
 }
 
+static bool pages_are_mergeable(struct page *a, struct page *b)
+{
+	if (page_to_pfn(a) != page_to_pfn(b) + 1)
+		return false;
+	if (!zone_device_pages_have_same_pgmap(a, b))
+		return false;
+	return true;
+}
+
 /**
  * sg_alloc_append_table_from_pages - Allocate and initialize an append sg
  *                                    table from an array of pages
@@ -447,6 +456,7 @@ int sg_alloc_append_table_from_pages(struct sg_append_table *sgt_append,
 	unsigned int chunks, cur_page, seg_len, i, prv_len = 0;
 	unsigned int added_nents = 0;
 	struct scatterlist *s = sgt_append->prv;
+	struct page *last_pg;
 
 	/*
 	 * The algorithm below requires max_segment to be aligned to PAGE_SIZE
@@ -460,21 +470,17 @@ int sg_alloc_append_table_from_pages(struct sg_append_table *sgt_append,
 		return -EOPNOTSUPP;
 
 	if (sgt_append->prv) {
-		unsigned long paddr =
-			(page_to_pfn(sg_page(sgt_append->prv)) * PAGE_SIZE +
-			 sgt_append->prv->offset + sgt_append->prv->length) /
-			PAGE_SIZE;
-
 		if (WARN_ON(offset))
 			return -EINVAL;
 
 		/* Merge contiguous pages into the last SG */
 		prv_len = sgt_append->prv->length;
-		while (n_pages && page_to_pfn(pages[0]) == paddr) {
+		last_pg = sg_page(sgt_append->prv);
+		while (n_pages && pages_are_mergeable(last_pg, pages[0])) {
 			if (sgt_append->prv->length + PAGE_SIZE > max_segment)
 				break;
 			sgt_append->prv->length += PAGE_SIZE;
-			paddr++;
+			last_pg = pages[0];
 			pages++;
 			n_pages--;
 		}
@@ -488,7 +494,7 @@ int sg_alloc_append_table_from_pages(struct sg_append_table *sgt_append,
 	for (i = 1; i < n_pages; i++) {
 		seg_len += PAGE_SIZE;
 		if (seg_len >= max_segment ||
-		    page_to_pfn(pages[i]) != page_to_pfn(pages[i - 1]) + 1) {
+		    !pages_are_mergeable(pages[i], pages[i - 1])) {
 			chunks++;
 			seg_len = 0;
 		}
@@ -504,8 +510,7 @@ int sg_alloc_append_table_from_pages(struct sg_append_table *sgt_append,
 		for (j = cur_page + 1; j < n_pages; j++) {
 			seg_len += PAGE_SIZE;
 			if (seg_len >= max_segment ||
-			    page_to_pfn(pages[j]) !=
-			    page_to_pfn(pages[j - 1]) + 1)
+			    !pages_are_mergeable(pages[j], pages[j - 1]))
 				break;
 		}
 
-- 
2.30.2

[PATCH v6 18/21] block: set FOLL_PCI_P2PDMA in __bio_iov_iter_get_pages()

[Logan Gunthorpe]

When a bio's queue supports PCI P2PDMA, set FOLL_PCI_P2PDMA for
iov_iter_get_pages_flags(). This allows PCI P2PDMA pages to be passed
from userspace and enables the O_DIRECT path in iomap based filesystems
and direct to block devices.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 block/bio.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/block/bio.c b/block/bio.c
index 3406c0450db3..271a720a6dc1 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -1149,6 +1149,7 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt;
 	struct page **pages = (struct page **)bv;
 	bool same_page = false;
+	unsigned int flags = 0;
 	ssize_t size, left;
 	unsigned len, i;
 	size_t offset;
@@ -1161,7 +1162,12 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 	BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
 	pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
 
-	size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset);
+	if (bio->bi_bdev && bio->bi_bdev->bd_disk &&
+	    blk_queue_pci_p2pdma(bio->bi_bdev->bd_disk->queue))
+		flags |= FOLL_PCI_P2PDMA;
+
+	size = iov_iter_get_pages_flags(iter, pages, LONG_MAX, nr_pages,
+					&offset, flags);
 	if (unlikely(size <= 0))
 		return size ? size : -EFAULT;
 
-- 
2.30.2

[PATCH v6 19/21] block: set FOLL_PCI_P2PDMA in bio_map_user_iov()

[Logan Gunthorpe]

When a bio's queue supports PCI P2PDMA, set FOLL_PCI_P2PDMA for
iov_iter_get_pages_flags(). This allows PCI P2PDMA pages to be
passed from userspace and enables the NVMe passthru requests to
use P2PDMA pages.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 block/blk-map.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/block/blk-map.c b/block/blk-map.c
index c7f71d83eff1..85baf922a0e8 100644
--- a/block/blk-map.c
+++ b/block/blk-map.c
@@ -234,6 +234,7 @@ static int bio_map_user_iov(struct request *rq, struct iov_iter *iter,
 		gfp_t gfp_mask)
 {
 	unsigned int max_sectors = queue_max_hw_sectors(rq->q);
+	unsigned int flags = 0;
 	struct bio *bio;
 	int ret;
 	int j;
@@ -246,13 +247,17 @@ static int bio_map_user_iov(struct request *rq, struct iov_iter *iter,
 		return -ENOMEM;
 	bio->bi_opf |= req_op(rq);
 
+	if (blk_queue_pci_p2pdma(rq->q))
+		flags |= FOLL_PCI_P2PDMA;
+
 	while (iov_iter_count(iter)) {
 		struct page **pages;
 		ssize_t bytes;
 		size_t offs, added = 0;
 		int npages;
 
-		bytes = iov_iter_get_pages_alloc(iter, &pages, LONG_MAX, &offs);
+		bytes = iov_iter_get_pages_alloc_flags(iter, &pages, LONG_MAX,
+						       &offs, flags);
 		if (unlikely(bytes <= 0)) {
 			ret = bytes ? bytes : -EFAULT;
 			goto out_unmap;
-- 
2.30.2

[PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Logan Gunthorpe]

Introduce pci_mmap_p2pmem() which is a helper to allocate and mmap
a hunk of p2pmem into userspace.

Pages are allocated from the genalloc in bulk with their reference
count set to one. They are returned to the genalloc when the page is put
through p2pdma_page_free() (the reference count is once again set to 1
in free_zone_device_page()).

The VMA does not take a reference to the pages when they are inserted
with vmf_insert_mixed() (which is necessary for zone device pages) so
the backing P2P memory is stored in a structures in vm_private_data.

A pseudo mount is used to allocate an inode for each PCI device. The
inode's address_space is used in the file doing the mmap so that all
VMAs are collected and can be unmapped if the PCI device is unbound.
After unmapping, the VMAs are iterated through and their pages are
put so the device can continue to be unbound. An active flag is used
to signal to VMAs not to allocate any further P2P memory once the
removal process starts. The flag is synchronized with concurrent
access with an RCU lock.

The VMAs and inode will survive after the unbind of the device, but no
pages will be present in the VMA and a subsequent access will result
in a SIGBUS error.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
---
 drivers/pci/p2pdma.c       | 340 ++++++++++++++++++++++++++++++++++++-
 include/linux/pci-p2pdma.h |  11 ++
 include/uapi/linux/magic.h |   1 +
 3 files changed, 350 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/p2pdma.c b/drivers/pci/p2pdma.c
index 4d3cab9da748..cce4c7b6dd75 100644
--- a/drivers/pci/p2pdma.c
+++ b/drivers/pci/p2pdma.c
@@ -17,14 +17,19 @@
 #include <linux/genalloc.h>
 #include <linux/memremap.h>
 #include <linux/percpu-refcount.h>
+#include <linux/pfn_t.h>
+#include <linux/pseudo_fs.h>
 #include <linux/random.h>
 #include <linux/seq_buf.h>
 #include <linux/xarray.h>
+#include <uapi/linux/magic.h>
 
 struct pci_p2pdma {
 	struct gen_pool *pool;
 	bool p2pmem_published;
 	struct xarray map_types;
+	struct inode *inode;
+	bool active;
 };
 
 struct pci_p2pdma_pagemap {
@@ -33,6 +38,17 @@ struct pci_p2pdma_pagemap {
 	u64 bus_offset;
 };
 
+struct pci_p2pdma_map {
+	struct kref ref;
+	struct rcu_head rcu;
+	struct pci_dev *pdev;
+	struct inode *inode;
+	size_t len;
+
+	spinlock_t kaddr_lock;
+	void *kaddr;
+};
+
 static struct pci_p2pdma_pagemap *to_p2p_pgmap(struct dev_pagemap *pgmap)
 {
 	return container_of(pgmap, struct pci_p2pdma_pagemap, pgmap);
@@ -101,6 +117,38 @@ static const struct attribute_group p2pmem_group = {
 	.name = "p2pmem",
 };
 
+/*
+ * P2PDMA internal mount
+ * Fake an internal VFS mount-point in order to allocate struct address_space
+ * mappings to remove VMAs on unbind events.
+ */
+static int pci_p2pdma_fs_cnt;
+static struct vfsmount *pci_p2pdma_fs_mnt;
+
+static int pci_p2pdma_fs_init_fs_context(struct fs_context *fc)
+{
+	return init_pseudo(fc, P2PDMA_MAGIC) ? 0 : -ENOMEM;
+}
+
+static struct file_system_type pci_p2pdma_fs_type = {
+	.name = "p2dma",
+	.owner = THIS_MODULE,
+	.init_fs_context = pci_p2pdma_fs_init_fs_context,
+	.kill_sb = kill_anon_super,
+};
+
+static void p2pdma_page_free(struct page *page)
+{
+	struct pci_p2pdma_pagemap *pgmap = to_p2p_pgmap(page->pgmap);
+
+	gen_pool_free(pgmap->provider->p2pdma->pool,
+		      (uintptr_t)page_to_virt(page), PAGE_SIZE);
+}
+
+static const struct dev_pagemap_ops p2pdma_pgmap_ops = {
+	.page_free = p2pdma_page_free,
+};
+
 static void pci_p2pdma_release(void *data)
 {
 	struct pci_dev *pdev = data;
@@ -117,6 +165,9 @@ static void pci_p2pdma_release(void *data)
 	gen_pool_destroy(p2pdma->pool);
 	sysfs_remove_group(&pdev->dev.kobj, &p2pmem_group);
 	xa_destroy(&p2pdma->map_types);
+
+	iput(p2pdma->inode);
+	simple_release_fs(&pci_p2pdma_fs_mnt, &pci_p2pdma_fs_cnt);
 }
 
 static int pci_p2pdma_setup(struct pci_dev *pdev)
@@ -134,17 +185,32 @@ static int pci_p2pdma_setup(struct pci_dev *pdev)
 	if (!p2p->pool)
 		goto out;
 
-	error = devm_add_action_or_reset(&pdev->dev, pci_p2pdma_release, pdev);
+	error = simple_pin_fs(&pci_p2pdma_fs_type, &pci_p2pdma_fs_mnt,
+			      &pci_p2pdma_fs_cnt);
 	if (error)
 		goto out_pool_destroy;
 
+	p2p->inode = alloc_anon_inode(pci_p2pdma_fs_mnt->mnt_sb);
+	if (IS_ERR(p2p->inode)) {
+		error = -ENOMEM;
+		goto out_unpin_fs;
+	}
+
+	error = devm_add_action_or_reset(&pdev->dev, pci_p2pdma_release, pdev);
+	if (error)
+		goto out_put_inode;
+
 	error = sysfs_create_group(&pdev->dev.kobj, &p2pmem_group);
 	if (error)
-		goto out_pool_destroy;
+		goto out_put_inode;
 
 	rcu_assign_pointer(pdev->p2pdma, p2p);
 	return 0;
 
+out_put_inode:
+	iput(p2p->inode);
+out_unpin_fs:
+	simple_release_fs(&pci_p2pdma_fs_mnt, &pci_p2pdma_fs_cnt);
 out_pool_destroy:
 	gen_pool_destroy(p2p->pool);
 out:
@@ -152,6 +218,54 @@ static int pci_p2pdma_setup(struct pci_dev *pdev)
 	return error;
 }
 
+static void pci_p2pdma_map_free_pages(struct pci_p2pdma_map *pmap)
+{
+	int i;
+
+	if (!pmap->kaddr)
+		return;
+
+	for (i = 0; i < pmap->len; i += PAGE_SIZE)
+		put_page(virt_to_page(pmap->kaddr + i));
+
+	pmap->kaddr = NULL;
+}
+
+static void pci_p2pdma_free_mappings(struct address_space *mapping)
+{
+	struct vm_area_struct *vma;
+
+	i_mmap_lock_write(mapping);
+	if (RB_EMPTY_ROOT(&mapping->i_mmap.rb_root))
+		goto out;
+
+	vma_interval_tree_foreach(vma, &mapping->i_mmap, 0, -1)
+		pci_p2pdma_map_free_pages(vma->vm_private_data);
+
+out:
+	i_mmap_unlock_write(mapping);
+}
+
+static void pci_p2pdma_unmap_mappings(void *data)
+{
+	struct pci_dev *pdev = data;
+	struct pci_p2pdma *p2pdma = rcu_dereference_protected(pdev->p2pdma, 1);
+
+	/* Ensure no new pages can be allocated in mappings */
+	p2pdma->active = false;
+	synchronize_rcu();
+
+	unmap_mapping_range(p2pdma->inode->i_mapping, 0, 0, 1);
+
+	/*
+	 * On some architectures, TLB flushes are done with call_rcu()
+	 * so to ensure GUP fast is done with the pages, call synchronize_rcu()
+	 * before freeing them.
+	 */
+	synchronize_rcu();
+	pci_p2pdma_free_mappings(p2pdma->inode->i_mapping);
+}
+
 /**
  * pci_p2pdma_add_resource - add memory for use as p2p memory
  * @pdev: the device to add the memory to
@@ -198,6 +312,7 @@ int pci_p2pdma_add_resource(struct pci_dev *pdev, int bar, size_t size,
 	pgmap->range.end = pgmap->range.start + size - 1;
 	pgmap->nr_range = 1;
 	pgmap->type = MEMORY_DEVICE_PCI_P2PDMA;
+	pgmap->ops = &p2pdma_pgmap_ops;
 
 	p2p_pgmap->provider = pdev;
 	p2p_pgmap->bus_offset = pci_bus_address(pdev, bar) -
@@ -209,6 +324,11 @@ int pci_p2pdma_add_resource(struct pci_dev *pdev, int bar, size_t size,
 		goto pgmap_free;
 	}
 
+	error = devm_add_action_or_reset(&pdev->dev, pci_p2pdma_unmap_mappings,
+					 pdev);
+	if (error)
+		goto pages_free;
+
 	p2pdma = rcu_dereference_protected(pdev->p2pdma, 1);
 	error = gen_pool_add_owner(p2pdma->pool, (unsigned long)addr,
 			pci_bus_address(pdev, bar) + offset,
@@ -217,6 +337,7 @@ int pci_p2pdma_add_resource(struct pci_dev *pdev, int bar, size_t size,
 	if (error)
 		goto pages_free;
 
+	p2pdma->active = true;
 	pci_info(pdev, "added peer-to-peer DMA memory %#llx-%#llx\n",
 		 pgmap->range.start, pgmap->range.end);
 
@@ -1018,3 +1139,218 @@ ssize_t pci_p2pdma_enable_show(char *page, struct pci_dev *p2p_dev,
 	return sprintf(page, "%s\n", pci_name(p2p_dev));
 }
 EXPORT_SYMBOL_GPL(pci_p2pdma_enable_show);
+
+static struct pci_p2pdma_map *pci_p2pdma_map_alloc(struct pci_dev *pdev,
+						   size_t len)
+{
+	struct pci_p2pdma_map *pmap;
+
+	pmap = kzalloc(sizeof(*pmap), GFP_KERNEL);
+	if (!pmap)
+		return NULL;
+
+	kref_init(&pmap->ref);
+	pmap->pdev = pci_dev_get(pdev);
+	pmap->len = len;
+
+	return pmap;
+}
+
+static void pci_p2pdma_map_free(struct rcu_head *rcu)
+{
+	struct pci_p2pdma_map *pmap =
+		container_of(rcu, struct pci_p2pdma_map, rcu);
+
+	pci_p2pdma_map_free_pages(pmap);
+	kfree(pmap);
+}
+
+static void pci_p2pdma_map_release(struct kref *ref)
+{
+	struct pci_p2pdma_map *pmap =
+		container_of(ref, struct pci_p2pdma_map, ref);
+
+	iput(pmap->inode);
+	simple_release_fs(&pci_p2pdma_fs_mnt, &pci_p2pdma_fs_cnt);
+	pci_dev_put(pmap->pdev);
+
+	if (pmap->kaddr) {
+		/*
+		 * Make sure to wait for the TLB flush (which some
+		 * architectures do using call_rcu()) before returning the
+		 * pages to the genalloc. This ensures the pages are not reused
+		 * before GUP-fast is finished with them. So the mapping is
+		 * freed using call_rcu() seeing adding synchronize_rcu() to
+		 * the munmap path can cause long delays on large systems
+		 * during process cleanup.
+		 */
+		call_rcu(&pmap->rcu, pci_p2pdma_map_free);
+		return;
+	}
+
+	/*
+	 * If there are no pages, just free the object immediately. There
+	 * are no more references to it so there is nothing that can race
+	 * with adding the pages.
+	 */
+	pci_p2pdma_map_free(&pmap->rcu);
+}
+
+static void pci_p2pdma_vma_open(struct vm_area_struct *vma)
+{
+	struct pci_p2pdma_map *pmap = vma->vm_private_data;
+
+	kref_get(&pmap->ref);
+}
+
+static void pci_p2pdma_vma_close(struct vm_area_struct *vma)
+{
+	struct pci_p2pdma_map *pmap = vma->vm_private_data;
+
+	kref_put(&pmap->ref, pci_p2pdma_map_release);
+}
+
+static bool pci_p2pdma_vma_alloc(struct pci_p2pdma_map *pmap)
+{
+	struct pci_p2pdma *p2pdma;
+	bool ret = true;
+	void *kaddr;
+
+	spin_lock(&pmap->kaddr_lock);
+	if (pmap->kaddr)
+		goto out_spin_unlock;
+
+	rcu_read_lock();
+	ret = false;
+	p2pdma = rcu_dereference(pmap->pdev->p2pdma);
+	if (!p2pdma)
+		goto out_rcu_unlock;
+
+	if (!p2pdma->active)
+		goto out_rcu_unlock;
+
+	kaddr = (void *)gen_pool_alloc(p2pdma->pool, pmap->len);
+	if (!kaddr) {
+		pci_err(pmap->pdev,
+				"No free P2PDMA memory for userspace mmap\n");
+		goto out_rcu_unlock;
+	}
+
+	WRITE_ONCE(pmap->kaddr, kaddr);
+	ret = true;
+
+out_rcu_unlock:
+	rcu_read_unlock();
+out_spin_unlock:
+	spin_unlock(&pmap->kaddr_lock);
+	return ret;
+}
+
+static vm_fault_t pci_p2pdma_vma_fault(struct vm_fault *vmf)
+{
+	struct pci_p2pdma_map *pmap = vmf->vma->vm_private_data;
+	void *vaddr;
+	pfn_t pfn;
+
+	if (!READ_ONCE(pmap->kaddr)) {
+		if (!pci_p2pdma_vma_alloc(pmap))
+			return VM_FAULT_SIGBUS;
+	}
+
+	vaddr = pmap->kaddr + vmf->address - vmf->vma->vm_start;
+	pfn = phys_to_pfn_t(virt_to_phys(vaddr), PFN_DEV | PFN_MAP);
+
+	return vmf_insert_mixed(vmf->vma, vmf->address, pfn);
+}
+static const struct vm_operations_struct pci_p2pdma_vmops = {
+	.open = pci_p2pdma_vma_open,
+	.close = pci_p2pdma_vma_close,
+	.fault = pci_p2pdma_vma_fault,
+};
+
+/**
+ * pci_p2pdma_file_open - setup file mapping to store P2PMEM VMAs
+ * @pdev: the device to allocate memory from
+ * @vma: the userspace vma to map the memory to
+ *
+ * Set f_mapping of the file to the p2pdma inode so that mappings
+ * are can be torn down on device unbind.
+ *
+ * Returns 0 on success, or a negative error code on failure
+ */
+void pci_p2pdma_file_open(struct pci_dev *pdev, struct file *file)
+{
+	struct pci_p2pdma *p2pdma;
+
+	rcu_read_lock();
+	p2pdma = rcu_dereference(pdev->p2pdma);
+	if (p2pdma)
+		file->f_mapping = p2pdma->inode->i_mapping;
+	rcu_read_unlock();
+}
+EXPORT_SYMBOL_GPL(pci_p2pdma_file_open);
+
+/**
+ * pci_mmap_p2pmem - setup an mmap region to be backed with P2PDMA memory
+ *	that was registered with pci_p2pdma_add_resource()
+ * @pdev: the device to allocate memory from
+ * @vma: the userspace vma to map the memory to
+ *
+ * The file must call pci_p2pdma_mmap_file_open() in its open() operation.
+ *
+ * Returns 0 on success, or a negative error code on failure
+ */
+int pci_mmap_p2pmem(struct pci_dev *pdev, struct vm_area_struct *vma)
+{
+	struct pci_p2pdma_map *pmap;
+	struct pci_p2pdma *p2pdma;
+	int ret;
+
+	/* prevent private mappings from being established */
+	if ((vma->vm_flags & VM_MAYSHARE) != VM_MAYSHARE) {
+		pci_info_ratelimited(pdev,
+				     "%s: fail, attempted private mapping\n",
+				     current->comm);
+		return -EINVAL;
+	}
+
+	if (vma->vm_pgoff) {
+		pci_info_ratelimited(pdev,
+				     "%s: fail, attempted mapping with non-zero offset\n",
+				     current->comm);
+		return -EINVAL;
+	}
+
+	pmap = pci_p2pdma_map_alloc(pdev, vma->vm_end - vma->vm_start);
+	if (!pmap)
+		return -ENOMEM;
+
+	spin_lock_init(&pmap->kaddr_lock);
+	rcu_read_lock();
+	p2pdma = rcu_dereference(pdev->p2pdma);
+	if (!p2pdma) {
+		ret = -ENODEV;
+		goto out;
+	}
+
+	ret = simple_pin_fs(&pci_p2pdma_fs_type, &pci_p2pdma_fs_mnt,
+			    &pci_p2pdma_fs_cnt);
+	if (ret)
+		goto out;
+
+	ihold(p2pdma->inode);
+	pmap->inode = p2pdma->inode;
+	rcu_read_unlock();
+
+	vma->vm_flags |= VM_MIXEDMAP;
+	vma->vm_private_data = pmap;
+	vma->vm_ops = &pci_p2pdma_vmops;
+
+	return 0;
+
+out:
+	rcu_read_unlock();
+	kfree(pmap);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(pci_mmap_p2pmem);
diff --git a/include/linux/pci-p2pdma.h b/include/linux/pci-p2pdma.h
index 2c07aa6b7665..040d79126463 100644
--- a/include/linux/pci-p2pdma.h
+++ b/include/linux/pci-p2pdma.h
@@ -34,6 +34,8 @@ int pci_p2pdma_enable_store(const char *page, struct pci_dev **p2p_dev,
 			    bool *use_p2pdma);
 ssize_t pci_p2pdma_enable_show(char *page, struct pci_dev *p2p_dev,
 			       bool use_p2pdma);
+void pci_p2pdma_file_open(struct pci_dev *pdev, struct file *file);
+int pci_mmap_p2pmem(struct pci_dev *pdev, struct vm_area_struct *vma);
 #else /* CONFIG_PCI_P2PDMA */
 static inline int pci_p2pdma_add_resource(struct pci_dev *pdev, int bar,
 		size_t size, u64 offset)
@@ -90,6 +92,15 @@ static inline ssize_t pci_p2pdma_enable_show(char *page,
 {
 	return sprintf(page, "none\n");
 }
+static inline void pci_p2pdma_file_open(struct pci_dev *pdev,
+					struct file *file)
+{
+}
+static inline int pci_mmap_p2pmem(struct pci_dev *pdev,
+				  struct vm_area_struct *vma)
+{
+	return -EOPNOTSUPP;
+}
 #endif /* CONFIG_PCI_P2PDMA */
 
 
diff --git a/include/uapi/linux/magic.h b/include/uapi/linux/magic.h
index f724129c0425..59ba2e60dc03 100644
--- a/include/uapi/linux/magic.h
+++ b/include/uapi/linux/magic.h
@@ -95,6 +95,7 @@
 #define BPF_FS_MAGIC		0xcafe4a11
 #define AAFS_MAGIC		0x5a3c69f0
 #define ZONEFS_MAGIC		0x5a4f4653
+#define P2PDMA_MAGIC		0x70327064
 
 /* Since UDF 2.01 is ISO 13346 based... */
 #define UDF_SUPER_MAGIC		0x15013346
-- 
2.30.2

[PATCH v6 21/21] nvme-pci: allow mmaping the CMB in userspace

[Logan Gunthorpe]

Allow userspace to obtain CMB memory by mmaping the controller's
char device. The mmap call allocates and returns a hunk of CMB memory,
(the offset is ignored) so userspace does not have control over the
address within the CMB.

A VMA allocated in this way will only be usable by drivers that set
FOLL_PCI_P2PDMA when calling GUP. And inter-device support will be
checked the first time the pages are mapped for DMA.

Currently this is only supported by O_DIRECT to an PCI NVMe device
or through the NVMe passthrough IOCTL.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
---
 drivers/nvme/host/core.c | 15 +++++++++++++++
 drivers/nvme/host/nvme.h |  2 ++
 drivers/nvme/host/pci.c  | 17 +++++++++++++++++
 3 files changed, 34 insertions(+)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index bbc276dda49f..1fd3372c2c18 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -3114,6 +3114,10 @@ static int nvme_dev_open(struct inode *inode, struct file *file)
 	}
 
 	file->private_data = ctrl;
+
+	if (ctrl->ops->cdev_file_open)
+		ctrl->ops->cdev_file_open(ctrl, file);
+
 	return 0;
 }
 
@@ -3127,12 +3131,23 @@ static int nvme_dev_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
+static int nvme_dev_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	struct nvme_ctrl *ctrl = file->private_data;
+
+	if (!ctrl->ops->mmap_cmb)
+		return -ENODEV;
+
+	return ctrl->ops->mmap_cmb(ctrl, vma);
+}
+
 static const struct file_operations nvme_dev_fops = {
 	.owner		= THIS_MODULE,
 	.open		= nvme_dev_open,
 	.release	= nvme_dev_release,
 	.unlocked_ioctl	= nvme_dev_ioctl,
 	.compat_ioctl	= compat_ptr_ioctl,
+	.mmap		= nvme_dev_mmap,
 };
 
 static ssize_t nvme_sysfs_reset(struct device *dev,
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 7d97bfb2a9e2..24fbcd274c64 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -497,6 +497,8 @@ struct nvme_ctrl_ops {
 	void (*delete_ctrl)(struct nvme_ctrl *ctrl);
 	int (*get_address)(struct nvme_ctrl *ctrl, char *buf, int size);
 	bool (*supports_pci_p2pdma)(struct nvme_ctrl *ctrl);
+	void (*cdev_file_open)(struct nvme_ctrl *ctrl, struct file *file);
+	int (*mmap_cmb)(struct nvme_ctrl *ctrl, struct vm_area_struct *vma);
 };
 
 /*
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 07412116d4d1..5946244e0295 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -2965,6 +2965,21 @@ static bool nvme_pci_supports_pci_p2pdma(struct nvme_ctrl *ctrl)
 	return dma_pci_p2pdma_supported(dev->dev);
 }
 
+static void nvme_pci_cdev_file_open(struct nvme_ctrl *ctrl, struct file *file)
+{
+	struct pci_dev *pdev = to_pci_dev(to_nvme_dev(ctrl)->dev);
+
+	pci_p2pdma_file_open(pdev, file);
+}
+
+static int nvme_pci_mmap_cmb(struct nvme_ctrl *ctrl,
+			     struct vm_area_struct *vma)
+{
+	struct pci_dev *pdev = to_pci_dev(to_nvme_dev(ctrl)->dev);
+
+	return pci_mmap_p2pmem(pdev, vma);
+}
+
 static const struct nvme_ctrl_ops nvme_pci_ctrl_ops = {
 	.name			= "pcie",
 	.module			= THIS_MODULE,
@@ -2976,6 +2991,8 @@ static const struct nvme_ctrl_ops nvme_pci_ctrl_ops = {
 	.submit_async_event	= nvme_pci_submit_async_event,
 	.get_address		= nvme_pci_get_address,
 	.supports_pci_p2pdma	= nvme_pci_supports_pci_p2pdma,
+	.cdev_file_open		= nvme_pci_cdev_file_open,
+	.mmap_cmb		= nvme_pci_mmap_cmb,
 };
 
 static int nvme_dev_map(struct nvme_dev *dev)
-- 
2.30.2

Re: [PATCH v6 00/21] Userspace P2PDMA with O_DIRECT NVMe devices

[Chaitanya Kulkarni]

On 4/7/22 08:46, Logan Gunthorpe wrote:
> Hi,
> 
> This patchset continues my work to add userspace P2PDMA access using
> O_DIRECT NVMe devices. This posting contains some minor fixes and a
> rebase onto v5.18-rc1 which contains cleanup from Christoph around
> free_zone_device_page() that helps to enable this patchset. The
> previous posting was here[1].
> 
> The patchset enables userspace P2PDMA by allowing userspace to mmap()
> allocated chunks of the CMB. The resulting VMA can be passed only
> to O_DIRECT IO on NVMe backed files or block devices. A flag is added
> to GUP() in Patch <>, then Patches <> through <> wire this flag up based
> on whether the block queue indicates P2PDMA support. Patches <>
> through <> enable the CMB to be mapped into userspace by mmaping
> the nvme char device.
> 
> This is relatively straightforward, however the one significant
> problem is that, presently, pci_p2pdma_map_sg() requires a homogeneous
> SGL with all P2PDMA pages or all regular pages. Enhancing GUP to
> support enforcing this rule would require a huge hack that I don't
> expect would be all that pallatable. So the first 13 patches add
> support for P2PDMA pages to dma_map_sg[table]() to the dma-direct
> and dma-iommu implementations. Thus systems without an IOMMU plus
> Intel and AMD IOMMUs are supported. (Other IOMMU implementations would
> then be unsupported, notably ARM and PowerPC but support would be added
> when they convert to dma-iommu).
> 
> dma_map_sgtable() is preferred when dealing with P2PDMA memory as it
> will return -EREMOTEIO when the DMA device cannot map specific P2PDMA
> pages based on the existing rules in calc_map_type_and_dist().
> 
> The other issue is dma_unmap_sg() needs a flag to determine whether a
> given dma_addr_t was mapped regularly or as a PCI bus address. To allow
> this, a third flag is added to the page_link field in struct
> scatterlist. This effectively means support for P2PDMA will now depend
> on CONFIG_64BIT.
> 
> Feedback welcome.
> 
> This series is based on v5.18-rc1. A git branch is available here:
> 
>    https://github.com/sbates130272/linux-p2pmem/  p2pdma_user_cmb_v6
> 
> Thanks,
> 
> Logan
> 
> [1] lkml.kernel.org/r/20220128002614.6136-1-logang@deltatee.com
> 
> --


Do you have any plans to re-spin this ?

-ck

Re: [PATCH v6 00/21] Userspace P2PDMA with O_DIRECT NVMe devices

[Logan Gunthorpe]

On 2022-05-16 16:31, Chaitanya Kulkarni wrote:
> Do you have any plans to re-spin this ?

I didn't get any feedback this cycle, so there haven't been any changes.
I'll probably do a rebase and resend after the merge window.

Logan

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Jason Gunthorpe]

On Thu, Apr 07, 2022 at 09:47:16AM -0600, Logan Gunthorpe wrote:
> +static void pci_p2pdma_unmap_mappings(void *data)
> +{
> +	struct pci_dev *pdev = data;
> +	struct pci_p2pdma *p2pdma = rcu_dereference_protected(pdev->p2pdma, 1);
> +
> +	/* Ensure no new pages can be allocated in mappings */
> +	p2pdma->active = false;
> +	synchronize_rcu();
> +
> +	unmap_mapping_range(p2pdma->inode->i_mapping, 0, 0, 1);
> +
> +	/*
> +	 * On some architectures, TLB flushes are done with call_rcu()
> +	 * so to ensure GUP fast is done with the pages, call synchronize_rcu()
> +	 * before freeing them.
> +	 */
> +	synchronize_rcu();
> +	pci_p2pdma_free_mappings(p2pdma->inode->i_mapping);

With the series from Felix getting close this should get updated to
not set pte_devmap and use proper natural refcounting without any of
this stuff.

Jason

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Logan Gunthorpe]

On 2022-05-27 06:55, Jason Gunthorpe wrote:
> On Thu, Apr 07, 2022 at 09:47:16AM -0600, Logan Gunthorpe wrote:
>> +static void pci_p2pdma_unmap_mappings(void *data)
>> +{
>> +	struct pci_dev *pdev = data;
>> +	struct pci_p2pdma *p2pdma = rcu_dereference_protected(pdev->p2pdma, 1);
>> +
>> +	/* Ensure no new pages can be allocated in mappings */
>> +	p2pdma->active = false;
>> +	synchronize_rcu();
>> +
>> +	unmap_mapping_range(p2pdma->inode->i_mapping, 0, 0, 1);
>> +
>> +	/*
>> +	 * On some architectures, TLB flushes are done with call_rcu()
>> +	 * so to ensure GUP fast is done with the pages, call synchronize_rcu()
>> +	 * before freeing them.
>> +	 */
>> +	synchronize_rcu();
>> +	pci_p2pdma_free_mappings(p2pdma->inode->i_mapping);
> 
> With the series from Felix getting close this should get updated to
> not set pte_devmap and use proper natural refcounting without any of
> this stuff.

Can you send a link? I'm not sure what you are referring to.

Thanks,

Logan

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Jason Gunthorpe]

On Fri, May 27, 2022 at 09:35:07AM -0600, Logan Gunthorpe wrote:
> 
> 
> On 2022-05-27 06:55, Jason Gunthorpe wrote:
> > On Thu, Apr 07, 2022 at 09:47:16AM -0600, Logan Gunthorpe wrote:
> >> +static void pci_p2pdma_unmap_mappings(void *data)
> >> +{
> >> +	struct pci_dev *pdev = data;
> >> +	struct pci_p2pdma *p2pdma = rcu_dereference_protected(pdev->p2pdma, 1);
> >> +
> >> +	/* Ensure no new pages can be allocated in mappings */
> >> +	p2pdma->active = false;
> >> +	synchronize_rcu();
> >> +
> >> +	unmap_mapping_range(p2pdma->inode->i_mapping, 0, 0, 1);
> >> +
> >> +	/*
> >> +	 * On some architectures, TLB flushes are done with call_rcu()
> >> +	 * so to ensure GUP fast is done with the pages, call synchronize_rcu()
> >> +	 * before freeing them.
> >> +	 */
> >> +	synchronize_rcu();
> >> +	pci_p2pdma_free_mappings(p2pdma->inode->i_mapping);
> > 
> > With the series from Felix getting close this should get updated to
> > not set pte_devmap and use proper natural refcounting without any of
> > this stuff.
> 
> Can you send a link? I'm not sure what you are referring to.

IIRC this is the last part:

https://lore.kernel.org/linux-mm/20220524190632.3304-1-alex.sierra@amd.com/

And the earlier bit with Christoph's pieces looks like it might get
merged to v5.19..

The general idea is once pte_devmap is not set then all the
refcounting works the way it should. This is what all new ZONE_DEVICE
users should do..

Jason

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Logan Gunthorpe]

On 2022-05-27 13:03, Jason Gunthorpe wrote:
> On Fri, May 27, 2022 at 09:35:07AM -0600, Logan Gunthorpe wrote:
>>
>>
>> On 2022-05-27 06:55, Jason Gunthorpe wrote:
>>> On Thu, Apr 07, 2022 at 09:47:16AM -0600, Logan Gunthorpe wrote:
>>>> +static void pci_p2pdma_unmap_mappings(void *data)
>>>> +{
>>>> +	struct pci_dev *pdev = data;
>>>> +	struct pci_p2pdma *p2pdma = rcu_dereference_protected(pdev->p2pdma, 1);
>>>> +
>>>> +	/* Ensure no new pages can be allocated in mappings */
>>>> +	p2pdma->active = false;
>>>> +	synchronize_rcu();
>>>> +
>>>> +	unmap_mapping_range(p2pdma->inode->i_mapping, 0, 0, 1);
>>>> +
>>>> +	/*
>>>> +	 * On some architectures, TLB flushes are done with call_rcu()
>>>> +	 * so to ensure GUP fast is done with the pages, call synchronize_rcu()
>>>> +	 * before freeing them.
>>>> +	 */
>>>> +	synchronize_rcu();
>>>> +	pci_p2pdma_free_mappings(p2pdma->inode->i_mapping);
>>>
>>> With the series from Felix getting close this should get updated to
>>> not set pte_devmap and use proper natural refcounting without any of
>>> this stuff.
>>
>> Can you send a link? I'm not sure what you are referring to.
> 
> IIRC this is the last part:
> 
> https://lore.kernel.org/linux-mm/20220524190632.3304-1-alex.sierra@amd.com/
> 
> And the earlier bit with Christoph's pieces looks like it might get
> merged to v5.19..
> 
> The general idea is once pte_devmap is not set then all the
> refcounting works the way it should. This is what all new ZONE_DEVICE
> users should do..

Ok, I don't actually follow how those patches relate to this.

Based on your description I guess I don't need to set PFN_DEV and
perhaps not use vmf_insert_mixed()? And then just use vm_normal_page()?

But the refcounting of the pages seemed like it was already sane to me,
unless you mean that the code no longer has to synchronize_rcu() before
returning the pages... that would be spectacular and clean things up a
lot (plus fix an annoying issue where if you use then free all the
memory you can't allocate new memory for an indeterminate amount of time).

Logan

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Jason Gunthorpe]

On Fri, May 27, 2022 at 04:41:08PM -0600, Logan Gunthorpe wrote:
> > 
> > IIRC this is the last part:
> > 
> > https://lore.kernel.org/linux-mm/20220524190632.3304-1-alex.sierra@amd.com/
> > 
> > And the earlier bit with Christoph's pieces looks like it might get
> > merged to v5.19..
> > 
> > The general idea is once pte_devmap is not set then all the
> > refcounting works the way it should. This is what all new ZONE_DEVICE
> > users should do..
> 
> Ok, I don't actually follow how those patches relate to this.
> 
> Based on your description I guess I don't need to set PFN_DEV and

Yes

> perhaps not use vmf_insert_mixed()? And then just use vm_normal_page()?

I'm not sure ATM the best function to use, but yes, a function that
doesn't set PFN_DEV is needed here.
 
> But the refcounting of the pages seemed like it was already sane to me,
> unless you mean that the code no longer has to synchronize_rcu() before
> returning the pages... 

Right. It also doesn't need to call unmap range or keep track of the
inode, or do any of that stuff unless it really needs mmap revokation
semantics (which I doubt this use case does)

unmap range was only necessary because the refcounting is wrong -
since the pte's don't hold a ref on the page in PFN_DEV mode it is
necessary to wipe all the PTE explicitly before going ahead to
decrement the refcount on this path.

Just stuff the pages into the mmap, and your driver unprobe will
automatically block until all the mmaps are closed - no different than
having an open file descriptor or something.

Jason

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Logan Gunthorpe]

On 2022-06-01 18:00, Jason Gunthorpe wrote:
> On Fri, May 27, 2022 at 04:41:08PM -0600, Logan Gunthorpe wrote:
>>>
>>> IIRC this is the last part:
>>>
>>> https://lore.kernel.org/linux-mm/20220524190632.3304-1-alex.sierra@amd.com/
>>>
>>> And the earlier bit with Christoph's pieces looks like it might get
>>> merged to v5.19..
>>>
>>> The general idea is once pte_devmap is not set then all the
>>> refcounting works the way it should. This is what all new ZONE_DEVICE
>>> users should do..
>>
>> Ok, I don't actually follow how those patches relate to this.
>>
>> Based on your description I guess I don't need to set PFN_DEV and
> 
> Yes
> 
>> perhaps not use vmf_insert_mixed()? And then just use vm_normal_page()?
> 
> I'm not sure ATM the best function to use, but yes, a function that
> doesn't set PFN_DEV is needed here.
>  
>> But the refcounting of the pages seemed like it was already sane to me,
>> unless you mean that the code no longer has to synchronize_rcu() before
>> returning the pages... 
> 
> Right. It also doesn't need to call unmap range or keep track of the
> inode, or do any of that stuff unless it really needs mmap revokation
> semantics (which I doubt this use case does)
> 
> unmap range was only necessary because the refcounting is wrong -
> since the pte's don't hold a ref on the page in PFN_DEV mode it is
> necessary to wipe all the PTE explicitly before going ahead to
> decrement the refcount on this path.
> 
> Just stuff the pages into the mmap, and your driver unprobe will
> automatically block until all the mmaps are closed - no different than
> having an open file descriptor or something.

Oh is that what we want? With the current method the mmaps are unmapped
on unbind so that it doesn't block indefinitely. It seems more typical
for resources to be dropped quickly on unbind and processes that are
using them will get an error on next use.

Logan

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Jason Gunthorpe]

On Thu, Jun 02, 2022 at 10:16:10AM -0600, Logan Gunthorpe wrote:

> > Just stuff the pages into the mmap, and your driver unprobe will
> > automatically block until all the mmaps are closed - no different than
> > having an open file descriptor or something.
> 
> Oh is that what we want?

Yes, it is the typical case - eg if you have a sysfs file open unbind
hangs indefinitely. Many drivers can't unbind while they have open file
descriptors/etc.

A couple drivers go out of their way to allow unbinding while a live
userspace exists but this can get complicated. Usually there should be
a good reason.

The module will already be refcounted anyhow because the mmap points
to a char file which holds a module reference - meaning a simple rmmod
of the driver shouldn't work already..

Jason

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Logan Gunthorpe]

On 2022-06-02 10:30, Jason Gunthorpe wrote:
> On Thu, Jun 02, 2022 at 10:16:10AM -0600, Logan Gunthorpe wrote:
> 
>>> Just stuff the pages into the mmap, and your driver unprobe will
>>> automatically block until all the mmaps are closed - no different than
>>> having an open file descriptor or something.
>>
>> Oh is that what we want?
> 
> Yes, it is the typical case - eg if you have a sysfs file open unbind
> hangs indefinitely. Many drivers can't unbind while they have open file
> descriptors/etc.
> 
> A couple drivers go out of their way to allow unbinding while a live
> userspace exists but this can get complicated. Usually there should be
> a good reason.

This is not my experience. All the drivers I've worked with do not block
unbind with open file descriptors (at least for char devices). I know,
for example, that having a file descriptor open of /dev/nvmeX does not
cause unbinding to block. I figured this was the expectation as the
userspace process doing the unbind won't be able to be interrupted
seeing there's no way to fail on that path. Though, it certainly would
make things a lot easier if the unbind can block indefinitely as it
usually requires some complicated locking.

Do you have an example of this? What mechanisms are developers using to
block unbind with open file descriptors?

Logan

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Logan Gunthorpe]

On 2022-06-02 10:30, Jason Gunthorpe wrote:
> On Thu, Jun 02, 2022 at 10:16:10AM -0600, Logan Gunthorpe wrote:
> 
>>> Just stuff the pages into the mmap, and your driver unprobe will
>>> automatically block until all the mmaps are closed - no different than
>>> having an open file descriptor or something.
>>
>> Oh is that what we want?
> 
> Yes, it is the typical case - eg if you have a sysfs file open unbind
> hangs indefinitely. Many drivers can't unbind while they have open file
> descriptors/etc.
> 
> A couple drivers go out of their way to allow unbinding while a live
> userspace exists but this can get complicated. Usually there should be
> a good reason.
> 
> The module will already be refcounted anyhow because the mmap points
> to a char file which holds a module reference - meaning a simple rmmod
> of the driver shouldn't work already..

Also, I just tried it... If I open a sysfs file for an nvme device (ie.
/sys/class/nvme/nvme4/cntlid) and unbind the device, it does not block.
A subsequent read on that file descriptor returns ENODEV. Which is what
I would have expected.

Logan

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Jason Gunthorpe]

On Thu, Jun 02, 2022 at 10:45:55AM -0600, Logan Gunthorpe wrote:
> 
> 
> 
> On 2022-06-02 10:30, Jason Gunthorpe wrote:
> > On Thu, Jun 02, 2022 at 10:16:10AM -0600, Logan Gunthorpe wrote:
> > 
> >>> Just stuff the pages into the mmap, and your driver unprobe will
> >>> automatically block until all the mmaps are closed - no different than
> >>> having an open file descriptor or something.
> >>
> >> Oh is that what we want?
> > 
> > Yes, it is the typical case - eg if you have a sysfs file open unbind
> > hangs indefinitely. Many drivers can't unbind while they have open file
> > descriptors/etc.
> > 
> > A couple drivers go out of their way to allow unbinding while a live
> > userspace exists but this can get complicated. Usually there should be
> > a good reason.
> 
> This is not my experience. All the drivers I've worked with do not block
> unbind with open file descriptors (at least for char devices). I know,
> for example, that having a file descriptor open of /dev/nvmeX does not
> cause unbinding to block.

So there are lots of bugs in the kernel, and I've seen many drivers
that think calling cdev_device_del() is all they need to do - and then
happily allow cdev ioctl's/etc on a de-initialized driver struct.

Drivers that do take care of this usually have to put a lock around
all their fops to serialize against unbind. RDMA uses SRCU, iirc TPM
used a rwlock. But this is tricky and hurts fops performance.

I don't know what nvme did to protect against this, I didn't notice
an obvious lock.

> I figured this was the expectation as the userspace process doing
> the unbind won't be able to be interrupted seeing there's no way to
> fail on that path. Though, it certainly would make things a lot
> easier if the unbind can block indefinitely as it usually requires
> some complicated locking.

As I said, this is what sysfs does today and I don't see that ever
changing. If you userspace has a sysfs file open then the driver
unbind hangs until the file is closed.

So, doing as bad as sysfs seems like a reasonable baseline to me.

> Do you have an example of this? What mechanisms are developers using to
> block unbind with open file descriptors?

Sysfs maintains a refcount with a bias that is basically a fancied
rwlock. Most places use some kind of refcount triggering a
completion. Sleep on the completion until refcount is 0 on unbind kind
of thing.

Jason

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Jason Gunthorpe]

On Thu, Jun 02, 2022 at 10:49:15AM -0600, Logan Gunthorpe wrote:
> 
> 
> On 2022-06-02 10:30, Jason Gunthorpe wrote:
> > On Thu, Jun 02, 2022 at 10:16:10AM -0600, Logan Gunthorpe wrote:
> > 
> >>> Just stuff the pages into the mmap, and your driver unprobe will
> >>> automatically block until all the mmaps are closed - no different than
> >>> having an open file descriptor or something.
> >>
> >> Oh is that what we want?
> > 
> > Yes, it is the typical case - eg if you have a sysfs file open unbind
> > hangs indefinitely. Many drivers can't unbind while they have open file
> > descriptors/etc.
> > 
> > A couple drivers go out of their way to allow unbinding while a live
> > userspace exists but this can get complicated. Usually there should be
> > a good reason.
> > 
> > The module will already be refcounted anyhow because the mmap points
> > to a char file which holds a module reference - meaning a simple rmmod
> > of the driver shouldn't work already..
> 
> Also, I just tried it... If I open a sysfs file for an nvme device (ie.
> /sys/class/nvme/nvme4/cntlid) and unbind the device, it does not block.
> A subsequent read on that file descriptor returns ENODEV. Which is what
> I would have expected.

Oh interesting, this has been changed since years ago when I last
looked, the kernfs_get_active() is now more narrowed than it once
was. So manybe sysfs isn't the same concern it used to be!

Thanks,
Jason

Re: [PATCH v6 20/21] PCI/P2PDMA: Introduce pci_mmap_p2pmem()

[Logan Gunthorpe]

On 2022-06-02 11:28, Jason Gunthorpe wrote:
> On Thu, Jun 02, 2022 at 10:49:15AM -0600, Logan Gunthorpe wrote:
>>
>>
>> On 2022-06-02 10:30, Jason Gunthorpe wrote:
>>> On Thu, Jun 02, 2022 at 10:16:10AM -0600, Logan Gunthorpe wrote:
>>>
>>>>> Just stuff the pages into the mmap, and your driver unprobe will
>>>>> automatically block until all the mmaps are closed - no different than
>>>>> having an open file descriptor or something.
>>>>
>>>> Oh is that what we want?
>>>
>>> Yes, it is the typical case - eg if you have a sysfs file open unbind
>>> hangs indefinitely. Many drivers can't unbind while they have open file
>>> descriptors/etc.
>>>
>>> A couple drivers go out of their way to allow unbinding while a live
>>> userspace exists but this can get complicated. Usually there should be
>>> a good reason.
>>>
>>> The module will already be refcounted anyhow because the mmap points
>>> to a char file which holds a module reference - meaning a simple rmmod
>>> of the driver shouldn't work already..
>>
>> Also, I just tried it... If I open a sysfs file for an nvme device (ie.
>> /sys/class/nvme/nvme4/cntlid) and unbind the device, it does not block.
>> A subsequent read on that file descriptor returns ENODEV. Which is what
>> I would have expected.
> 
> Oh interesting, this has been changed since years ago when I last
> looked, the kernfs_get_active() is now more narrowed than it once
> was. So manybe sysfs isn't the same concern it used to be!

Yeah, so I really think *not* blocking unbind indefinitely is the better
approach here. It's what has always been done with device dax, etc.
mmaps in userspace processes get unmapped and will fault with SIGBUS on
next access and unbind will actually unbind the device relatively
promptly. Userspace processes can fail or try to handle the device going
away gracefully.

Logan