Some PCI devices may have memory mapped in a BAR space that's intended for use in Peer-to-Peer transactions. In order to enable such transactions the memory must be registered with ZONE_DEVICE pages so it can be used by DMA interfaces in existing drivers. A kernel interface is provided so that other subsystems can find and allocate chunks of P2P memory as necessary to facilitate transfers between two PCI peers. Depending on hardware, this may reduce the bandwidth of the transfer but would significantly reduce pressure on system memory. This may be desirable in many cases: for example a system could be designed with a small CPU connected to a PCI switch by a small number of lanes which would maximize the number of lanes available to connect to NVME devices. The interface requires a user driver to collect a list of client devices involved in the transaction with the pci_p2pmem_add_client*() functions then call pci_p2pmem_find() to obtain any suitable P2P memory. Once this is done the list is bound to the memory and the calling driver is free to add and remove clients as necessary. The ACS bits on the downstream switch port will be managed for all the registered clients. The code is designed to only utilize the p2pmem device if all the devices involved in a transfer are behind the same PCI switch. This is because using P2P transactions through the PCI root complex can have performance limitations or, worse, might not work at all. Finding out how well a particular RC supports P2P transfers is non-trivial. Additionally, the benefits of P2P transfers that go through the RC is limited to only reducing DRAM usage. This commit includes significant rework and feedback from Christoph Hellwig. Signed-off-by: Christoph Hellwig <hch@xxxxxx> Signed-off-by: Logan Gunthorpe <logang@xxxxxxxxxxxx> --- drivers/pci/Kconfig | 14 ++ drivers/pci/Makefile | 1 + drivers/pci/p2p.c | 549 +++++++++++++++++++++++++++++++++++++++++++++++ include/linux/memremap.h | 18 ++ include/linux/pci-p2p.h | 85 ++++++++ include/linux/pci.h | 4 + 6 files changed, 671 insertions(+) create mode 100644 drivers/pci/p2p.c create mode 100644 include/linux/pci-p2p.h diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig index bda151788f3f..188ea94cfe2e 100644 --- a/drivers/pci/Kconfig +++ b/drivers/pci/Kconfig @@ -123,6 +123,20 @@ config PCI_PASID If unsure, say N. +config PCI_P2P + bool "PCI Peer to Peer transfer support" + depends on ZONE_DEVICE + select GENERIC_ALLOCATOR + help + Enableѕ drivers to do PCI peer to peer transactions to and from + bars that are exposed to other devices in the same domain. + + Many PCIe root complexes do not support P2P transactions and + it's hard to tell which support it with good performance, so + at this time you will need a PCIe switch. + + If unsure, say N. + config PCI_LABEL def_bool y if (DMI || ACPI) select NLS diff --git a/drivers/pci/Makefile b/drivers/pci/Makefile index c7819b973df7..749858201400 100644 --- a/drivers/pci/Makefile +++ b/drivers/pci/Makefile @@ -23,6 +23,7 @@ obj-$(CONFIG_PCI_MSI) += msi.o obj-$(CONFIG_PCI_ATS) += ats.o obj-$(CONFIG_PCI_IOV) += iov.o +obj-$(CONFIG_PCI_P2P) += p2p.o # # ACPI Related PCI FW Functions diff --git a/drivers/pci/p2p.c b/drivers/pci/p2p.c new file mode 100644 index 000000000000..aa465ac9273d --- /dev/null +++ b/drivers/pci/p2p.c @@ -0,0 +1,549 @@ +/* + * Peer 2 Peer Memory support. + * + * Copyright (c) 2016-2017, Microsemi Corporation + * Copyright (c) 2017, Christoph Hellwig. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + */ + +#include <linux/pci-p2p.h> +#include <linux/module.h> +#include <linux/slab.h> +#include <linux/genalloc.h> +#include <linux/memremap.h> +#include <linux/percpu-refcount.h> + +struct pci_p2p { + struct percpu_ref devmap_ref; + struct completion devmap_ref_done; + struct gen_pool *pool; + bool published; +}; + +static void pci_p2pmem_percpu_release(struct percpu_ref *ref) +{ + struct pci_p2p *p2p = + container_of(ref, struct pci_p2p, devmap_ref); + + complete_all(&p2p->devmap_ref_done); +} + +static void pci_p2pmem_percpu_kill(void *data) +{ + struct percpu_ref *ref = data; + + if (percpu_ref_is_dying(ref)) + return; + + percpu_ref_kill(ref); +} + +static void pci_p2pmem_release(void *data) +{ + struct pci_dev *pdev = data; + + wait_for_completion(&pdev->p2p->devmap_ref_done); + percpu_ref_exit(&pdev->p2p->devmap_ref); + + gen_pool_destroy(pdev->p2p->pool); + pdev->p2p = NULL; +} + +static int pci_p2pmem_setup(struct pci_dev *pdev) +{ + int error = -ENOMEM; + struct pci_p2p *p2p; + + p2p = devm_kzalloc(&pdev->dev, sizeof(*p2p), GFP_KERNEL); + if (!p2p) + return -ENOMEM; + + p2p->pool = gen_pool_create(PAGE_SHIFT, dev_to_node(&pdev->dev)); + if (!p2p->pool) + goto out; + + init_completion(&p2p->devmap_ref_done); + error = percpu_ref_init(&p2p->devmap_ref, + pci_p2pmem_percpu_release, 0, GFP_KERNEL); + if (error) + goto out_pool_destroy; + + percpu_ref_switch_to_atomic_sync(&p2p->devmap_ref); + + error = devm_add_action_or_reset(&pdev->dev, pci_p2pmem_release, pdev); + if (error) + goto out_pool_destroy; + + pdev->p2p = p2p; + + return 0; + +out_pool_destroy: + gen_pool_destroy(p2p->pool); +out: + devm_kfree(&pdev->dev, p2p); + return error; +} + +/** + * pci_p2pmem_add_resource - add memory for use as p2p memory + * @pci: the device to add the memory to + * @bar: PCI bar to add + * @size: size of the memory to add, may be zero to use the whole bar + * @offset: offset into the PCI bar + * + * The memory will be given ZONE_DEVICE struct pages so that it may + * be used with any dma request. + */ +int pci_p2pmem_add_resource(struct pci_dev *pdev, int bar, size_t size, + u64 offset) +{ + struct dev_pagemap *pgmap; + void *addr; + int error; + + if (WARN_ON(offset >= pci_resource_len(pdev, bar))) + return -EINVAL; + + if (!size) + size = pci_resource_len(pdev, bar) - offset; + + if (WARN_ON(size + offset > pci_resource_len(pdev, bar))) + return -EINVAL; + + if (!pdev->p2p) { + error = pci_p2pmem_setup(pdev); + if (error) + return error; + } + + pgmap = devm_kzalloc(&pdev->dev, sizeof(*pgmap), GFP_KERNEL); + if (!pgmap) + return -ENOMEM; + + pgmap->res.start = pci_resource_start(pdev, bar) + offset; + pgmap->res.end = pgmap->res.start + size - 1; + pgmap->ref = &pdev->p2p->devmap_ref; + pgmap->type = MEMORY_DEVICE_PCI_P2P; + + addr = devm_memremap_pages(&pdev->dev, pgmap); + if (IS_ERR(addr)) + return PTR_ERR(addr); + + error = gen_pool_add_virt(pdev->p2p->pool, (uintptr_t)addr, + pci_bus_address(pdev, bar) + offset, + resource_size(&pgmap->res), dev_to_node(&pdev->dev)); + if (error) + return error; + + error = devm_add_action_or_reset(&pdev->dev, pci_p2pmem_percpu_kill, + &pdev->p2p->devmap_ref); + if (error) + return error; + + dev_info(&pdev->dev, "added %zdB of p2p memory\n", size); + + return 0; +} +EXPORT_SYMBOL_GPL(pci_p2pmem_add_resource); + +static struct pci_dev *find_parent_pci_dev(struct device *dev) +{ + struct device *parent; + + dev = get_device(dev); + + while (dev) { + if (dev_is_pci(dev)) + return to_pci_dev(dev); + + parent = get_device(dev->parent); + put_device(dev); + dev = parent; + } + + return NULL; +} + +/* + * If a device is behind a switch, we try to find the upstream bridge + * port of the switch. This requires two calls to pci_upstream_bridge: + * one for the upstream port on the switch, one on the upstream port + * for the next level in the hierarchy. Because of this, devices connected + * to the root port will be rejected. + */ +static struct pci_dev *get_upstream_switch_port(struct pci_dev *pdev) +{ + struct pci_dev *up1, *up2; + + if (!pdev) + return NULL; + + up1 = pci_dev_get(pci_upstream_bridge(pdev)); + if (!up1) + return NULL; + + up2 = pci_dev_get(pci_upstream_bridge(up1)); + pci_dev_put(up1); + + return up2; +} + +static bool __upstream_bridges_match(struct pci_dev *upstream, + struct pci_dev *client) +{ + struct pci_dev *dma_up; + bool ret = true; + + dma_up = get_upstream_switch_port(client); + + if (!dma_up) { + dev_dbg(&client->dev, "not a pci device behind a switch\n"); + ret = false; + goto out; + } + + if (upstream != dma_up) { + dev_dbg(&client->dev, + "does not reside on the same upstream bridge\n"); + ret = false; + goto out; + } + +out: + pci_dev_put(dma_up); + return ret; +} + +static bool upstream_bridges_match(struct pci_dev *pdev, + struct pci_dev *client) +{ + struct pci_dev *upstream; + bool ret; + + upstream = get_upstream_switch_port(pdev); + if (!upstream) { + dev_warn(&pdev->dev, "not behind a pci switch\n"); + return false; + } + + ret = __upstream_bridges_match(upstream, client); + + pci_dev_put(upstream); + + return ret; +} + +struct pci_p2pmem_client { + struct list_head list; + struct pci_dev *client; + struct pci_dev *p2pmem; +}; + +/** + * pci_p2pmem_add_client - allocate a new element in a client device list + * @head: list head of p2pmem clients + * @dev: device to add to the list + * + * This adds @dev to a list of clients used by a p2pmem device. + * This list should be passed to p2pmem_find(). Once p2pmem_find() has + * been called successfully, the list will be bound to a specific p2pmem + * device and new clients can only be added to the list if they are + * supported by that p2pmem device. + * + * The caller is expected to have a lock which protects @head as necessary + * so that none of the pci_p2pmem functions can be called concurrently + * on that list. + * + * Returns 0 if the client was successfully added. + */ +int pci_p2pmem_add_client(struct list_head *head, struct device *dev) +{ + struct pci_p2pmem_client *item, *new_item; + struct pci_dev *p2pmem = NULL; + struct pci_dev *client; + int ret; + + client = find_parent_pci_dev(dev); + if (!client) { + dev_warn(dev, + "cannot be used for p2p as it is not a pci device\n"); + return -ENODEV; + } + + item = list_first_entry_or_null(head, struct pci_p2pmem_client, list); + if (item && item->p2pmem) { + p2pmem = item->p2pmem; + + if (!upstream_bridges_match(p2pmem, client)) { + ret = -EXDEV; + goto put_client; + } + } + + new_item = kzalloc(sizeof(*new_item), GFP_KERNEL); + if (!new_item) { + ret = -ENOMEM; + goto put_client; + } + + new_item->client = client; + new_item->p2pmem = pci_dev_get(p2pmem); + + list_add_tail(&new_item->list, head); + + return 0; + +put_client: + pci_dev_put(client); + return ret; +} +EXPORT_SYMBOL_GPL(pci_p2pmem_add_client); + +static void pci_p2pmem_client_free(struct pci_p2pmem_client *item) +{ + list_del(&item->list); + pci_dev_put(item->client); + pci_dev_put(item->p2pmem); + kfree(item); +} + +/** + * pci_p2pmem_remove_client - remove and free a new p2pmem client + * @head: list head of p2pmem clients + * @dev: device to remove from the list + * + * This removes @dev from a list of clients used by a p2pmem device. + * The caller is expected to have a lock which protects @head as necessary + * so that none of the pci_p2pmem functions can be called concurrently + * on that list. + */ +void pci_p2pmem_remove_client(struct list_head *head, struct device *dev) +{ + struct pci_p2pmem_client *pos, *tmp; + struct pci_dev *pdev; + + pdev = find_parent_pci_dev(dev); + if (!pdev) + return; + + list_for_each_entry_safe(pos, tmp, head, list) { + if (pos->client != pdev) + continue; + + pci_p2pmem_client_free(pos); + } + + pci_dev_put(pdev); +} +EXPORT_SYMBOL_GPL(pci_p2pmem_remove_client); + +/** + * pci_p2pmem_client_list_free - free an entire list of p2pmem clients + * @head: list head of p2pmem clients + * + * This removes all devices in a list of clients used by a p2pmem device. + * The caller is expected to have a lock which protects @head as necessary + * so that none of the pci_p2pmem functions can be called concurrently + * on that list. + */ +void pci_p2pmem_client_list_free(struct list_head *head) +{ + struct pci_p2pmem_client *pos, *tmp; + + list_for_each_entry_safe(pos, tmp, head, list) + pci_p2pmem_client_free(pos); +} +EXPORT_SYMBOL_GPL(pci_p2pmem_client_list_free); + +static bool upstream_bridges_match_list(struct pci_dev *pdev, + struct list_head *head) +{ + struct pci_p2pmem_client *pos; + struct pci_dev *upstream; + bool ret; + + upstream = get_upstream_switch_port(pdev); + if (!upstream) { + dev_warn(&pdev->dev, "not behind a pci switch\n"); + return false; + } + + list_for_each_entry(pos, head, list) { + ret = __upstream_bridges_match(upstream, pos->client); + if (!ret) + break; + } + + pci_dev_put(upstream); + return ret; +} + +/** + * pci_p2pmem_find - find a p2p mem device compatible with the specified device + * @dev: list of device to check (NULL-terminated) + * + * For now, we only support cases where the devices that will transfer to the + * p2pmem device are on the same switch. This cuts out cases that may work but + * is safest for the user. + * + * Returns a pointer to the PCI device with a reference taken (use pci_dev_put + * to return the reference) or NULL if no compatible device is found. + */ +struct pci_dev *pci_p2pmem_find(struct list_head *clients) +{ + struct pci_dev *pdev = NULL; + struct pci_p2pmem_client *pos; + + while ((pdev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, pdev))) { + if (!pdev->p2p || !pdev->p2p->published) + continue; + + if (!upstream_bridges_match_list(pdev, clients)) + continue; + + list_for_each_entry(pos, clients, list) + pos->p2pmem = pdev; + + return pdev; + } + + return NULL; +} +EXPORT_SYMBOL_GPL(pci_p2pmem_find); + +/** + * pci_alloc_p2p_mem - allocate p2p memory + * @pdev: the device to allocate memory from + * @size: number of bytes to allocate + * + * Returns the allocated memory or NULL on error. + */ +void *pci_alloc_p2pmem(struct pci_dev *pdev, size_t size) +{ + void *ret; + + if (unlikely(!pdev->p2p)) + return NULL; + + if (unlikely(!percpu_ref_tryget_live(&pdev->p2p->devmap_ref))) + return NULL; + + ret = (void *)(uintptr_t)gen_pool_alloc(pdev->p2p->pool, size); + + if (unlikely(!ret)) + percpu_ref_put(&pdev->p2p->devmap_ref); + + return ret; +} +EXPORT_SYMBOL_GPL(pci_alloc_p2pmem); + +/** + * pci_free_p2pmem - allocate p2p memory + * @pdev: the device the memory was allocated from + * @addr: address of the memory that was allocated + * @size: number of bytes that was allocated + */ +void pci_free_p2pmem(struct pci_dev *pdev, void *addr, size_t size) +{ + gen_pool_free(pdev->p2p->pool, (uintptr_t)addr, size); + percpu_ref_put(&pdev->p2p->devmap_ref); +} +EXPORT_SYMBOL_GPL(pci_free_p2pmem); + +/** + * pci_virt_to_bus - return the pci bus address for a given virtual + * address obtained with pci_alloc_p2pmem + * @pdev: the device the memory was allocated from + * @addr: address of the memory that was allocated + */ +pci_bus_addr_t pci_p2pmem_virt_to_bus(struct pci_dev *pdev, void *addr) +{ + if (!addr) + return 0; + if (!pdev->p2p) + return 0; + + return gen_pool_virt_to_phys(pdev->p2p->pool, (unsigned long)addr); +} +EXPORT_SYMBOL_GPL(pci_p2pmem_virt_to_bus); + +/** + * pci_p2pmem_alloc_sgl - allocate p2p memory in an sgl + * @pdev: the device to allocate memory from + * @sgl: the allocated sgl + * @nents: the number of sgs in the list + * @length: number of bytes to allocate + * + * Returns 0 on success + */ +int pci_p2pmem_alloc_sgl(struct pci_dev *pdev, struct scatterlist **sgl, + unsigned int *nents, u32 length) +{ + struct scatterlist *sg; + void *addr; + + sg = kzalloc(sizeof(*sg), GFP_KERNEL); + if (!sg) + return -ENOMEM; + + sg_init_table(sg, 1); + + addr = pci_alloc_p2pmem(pdev, length); + if (!addr) + goto out_free_sg; + + sg_set_buf(sg, addr, length); + *sgl = sg; + *nents = 1; + return 0; + +out_free_sg: + kfree(sg); + return -ENOMEM; +} +EXPORT_SYMBOL_GPL(pci_p2pmem_alloc_sgl); + +/** + * pci_p2pmem_free_sgl - free an sgl allocated by pci_p2pmem_alloc_sgl + * @pdev: the device to allocate memory from + * @sgl: the allocated sgl + * @nents: the number of sgs in the list + */ +void pci_p2pmem_free_sgl(struct pci_dev *pdev, struct scatterlist *sgl, + unsigned int nents) +{ + struct scatterlist *sg; + int count; + + if (!sgl || !nents) + return; + + for_each_sg(sgl, sg, nents, count) + pci_free_p2pmem(pdev, sg_virt(sg), sg->length); + kfree(sgl); +} +EXPORT_SYMBOL_GPL(pci_p2pmem_free_sgl); + +/** + * pci_p2pmem_publish - publish the p2p memory for use by other devices + * with pci_p2pmem_find + * @pdev: the device with p2p memory to publish + * @publish: set to true to publish the memory, false to unpublish it + */ +void pci_p2pmem_publish(struct pci_dev *pdev, bool publish) +{ + if (WARN_ON(publish && !pdev->p2p)) + return; + + pdev->p2p->published = publish; +} +EXPORT_SYMBOL_GPL(pci_p2pmem_publish); diff --git a/include/linux/memremap.h b/include/linux/memremap.h index 7b4899c06f49..c17a6d167d48 100644 --- a/include/linux/memremap.h +++ b/include/linux/memremap.h @@ -53,11 +53,16 @@ struct vmem_altmap { * driver can hotplug the device memory using ZONE_DEVICE and with that memory * type. Any page of a process can be migrated to such memory. However no one * should be allow to pin such memory so that it can always be evicted. + * + * MEMORY_DEVICE_PCI_P2P: + * Device memory residing in a PCI BAR intended for use with Peer-to-Peer + * transactions. */ enum memory_type { MEMORY_DEVICE_HOST = 0, MEMORY_DEVICE_PRIVATE, MEMORY_DEVICE_PUBLIC, + MEMORY_DEVICE_PCI_P2P, }; /* @@ -161,6 +166,19 @@ static inline void vmem_altmap_free(struct vmem_altmap *altmap, } #endif /* CONFIG_ZONE_DEVICE */ +#ifdef CONFIG_PCI_P2P +static inline bool is_pci_p2p_page(const struct page *page) +{ + return is_zone_device_page(page) && + page->pgmap->type == MEMORY_DEVICE_PCI_P2P; +} +#else +static inline bool is_pci_p2p_page(const struct page *page) +{ + return false; +} +#endif + #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC) static inline bool is_device_private_page(const struct page *page) { diff --git a/include/linux/pci-p2p.h b/include/linux/pci-p2p.h new file mode 100644 index 000000000000..f811c97a5886 --- /dev/null +++ b/include/linux/pci-p2p.h @@ -0,0 +1,85 @@ +#ifndef _LINUX_PCI_P2P_H +#define _LINUX_PCI_P2P_H +/* + * Copyright (c) 2016-2017, Microsemi Corporation + * Copyright (c) 2017, Christoph Hellwig. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + */ + +#include <linux/pci.h> + +struct block_device; +struct scatterlist; + +#ifdef CONFIG_PCI_P2P +int pci_p2pmem_add_resource(struct pci_dev *pdev, int bar, size_t size, + u64 offset); +int pci_p2pmem_add_client(struct list_head *head, struct device *dev); +void pci_p2pmem_remove_client(struct list_head *head, struct device *dev); +void pci_p2pmem_client_list_free(struct list_head *head); +struct pci_dev *pci_p2pmem_find(struct list_head *clients); +void *pci_alloc_p2pmem(struct pci_dev *pdev, size_t size); +void pci_free_p2pmem(struct pci_dev *pdev, void *addr, size_t size); +pci_bus_addr_t pci_p2pmem_virt_to_bus(struct pci_dev *pdev, void *addr); +int pci_p2pmem_alloc_sgl(struct pci_dev *pdev, struct scatterlist **sgl, + unsigned int *nents, u32 length); +void pci_p2pmem_free_sgl(struct pci_dev *pdev, struct scatterlist *sgl, + unsigned int nents); +void pci_p2pmem_publish(struct pci_dev *pdev, bool publish); +#else /* CONFIG_PCI_P2P */ +static inline int pci_p2pmem_add_resource(struct pci_dev *pdev, int bar, + size_t size, u64 offset) +{ + return 0; +} +static inline int pci_p2pmem_add_client(struct list_head *head, + struct device *dev) +{ + return 0; +} +static inline void pci_p2pmem_remove_client(struct list_head *head, + struct device *dev) +{ +} +static inline void pci_p2pmem_client_list_free(struct list_head *head) +{ +} +static inline struct pci_dev *pci_p2pmem_find(struct list_head *clients) +{ + return NULL; +} +static inline void *pci_alloc_p2pmem(struct pci_dev *pdev, size_t size) +{ + return NULL; +} +static inline void pci_free_p2pmem(struct pci_dev *pdev, void *addr, + size_t size) +{ +} +static inline pci_bus_addr_t pci_p2pmem_virt_to_bus(struct pci_dev *pdev, + void *addr) +{ + return 0; +} +static inline int pci_p2pmem_alloc_sgl(struct pci_dev *pdev, + struct scatterlist **sgl, unsigned int *nents, u32 length) +{ + return -ENODEV; +} +static inline void pci_p2pmem_free_sgl(struct pci_dev *pdev, + struct scatterlist *sgl, unsigned int nents) +{ +} +static inline void pci_p2pmem_publish(struct pci_dev *pdev, bool publish) +{ +} +#endif /* CONFIG_PCI_P2P */ +#endif /* _LINUX_PCI_P2P_H */ diff --git a/include/linux/pci.h b/include/linux/pci.h index c170c9250c8b..047aea679e87 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -279,6 +279,7 @@ struct pcie_link_state; struct pci_vpd; struct pci_sriov; struct pci_ats; +struct pci_p2p; /* * The pci_dev structure is used to describe PCI devices. @@ -432,6 +433,9 @@ struct pci_dev { #ifdef CONFIG_PCI_PASID u16 pasid_features; #endif +#ifdef CONFIG_PCI_P2P + struct pci_p2p *p2p; +#endif phys_addr_t rom; /* Physical address of ROM if it's not from the BAR */ size_t romlen; /* Length of ROM if it's not from the BAR */ char *driver_override; /* Driver name to force a match */ -- 2.11.0