diff options
Diffstat (limited to 'drivers/ntb/ntb_transport.c')
| -rw-r--r-- | drivers/ntb/ntb_transport.c | 1029 |
1 files changed, 622 insertions, 407 deletions
diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c index e9bf2f47b61a..efe3ad4122f2 100644 --- a/drivers/ntb/ntb_transport.c +++ b/drivers/ntb/ntb_transport.c @@ -5,6 +5,7 @@ * GPL LICENSE SUMMARY * * Copyright(c) 2012 Intel Corporation. All rights reserved. + * Copyright (C) 2015 EMC Corporation. All Rights Reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of version 2 of the GNU General Public License as @@ -13,6 +14,7 @@ * BSD LICENSE * * Copyright(c) 2012 Intel Corporation. All rights reserved. + * Copyright (C) 2015 EMC Corporation. All Rights Reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions @@ -40,7 +42,7 @@ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * - * Intel PCIe NTB Linux driver + * PCIe NTB Transport Linux driver * * Contact Information: * Jon Mason <jon.mason@intel.com> @@ -56,11 +58,25 @@ #include <linux/pci.h> #include <linux/slab.h> #include <linux/types.h> -#include "ntb_hw.h" +#include <linux/uaccess.h> +#include "linux/ntb.h" +#include "linux/ntb_transport.h" -#define NTB_TRANSPORT_VERSION 3 +#define NTB_TRANSPORT_VERSION 4 +#define NTB_TRANSPORT_VER "4" +#define NTB_TRANSPORT_NAME "ntb_transport" +#define NTB_TRANSPORT_DESC "Software Queue-Pair Transport over NTB" -static unsigned int transport_mtu = 0x401E; +MODULE_DESCRIPTION(NTB_TRANSPORT_DESC); +MODULE_VERSION(NTB_TRANSPORT_VER); +MODULE_LICENSE("Dual BSD/GPL"); +MODULE_AUTHOR("Intel Corporation"); + +static unsigned long max_mw_size; +module_param(max_mw_size, ulong, 0644); +MODULE_PARM_DESC(max_mw_size, "Limit size of large memory windows"); + +static unsigned int transport_mtu = 0x10000; module_param(transport_mtu, uint, 0644); MODULE_PARM_DESC(transport_mtu, "Maximum size of NTB transport packets"); @@ -72,10 +88,16 @@ static unsigned int copy_bytes = 1024; module_param(copy_bytes, uint, 0644); MODULE_PARM_DESC(copy_bytes, "Threshold under which NTB will use the CPU to copy instead of DMA"); +static bool use_dma; +module_param(use_dma, bool, 0644); +MODULE_PARM_DESC(use_dma, "Use DMA engine to perform large data copy"); + +static struct dentry *nt_debugfs_dir; + struct ntb_queue_entry { /* ntb_queue list reference */ struct list_head entry; - /* pointers to data to be transfered */ + /* pointers to data to be transferred */ void *cb_data; void *buf; unsigned int len; @@ -94,14 +116,16 @@ struct ntb_rx_info { }; struct ntb_transport_qp { - struct ntb_transport *transport; - struct ntb_device *ndev; + struct ntb_transport_ctx *transport; + struct ntb_dev *ndev; void *cb_data; struct dma_chan *dma_chan; bool client_ready; - bool qp_link; + bool link_is_up; + u8 qp_num; /* Only 64 QP's are allowed. 0-63 */ + u64 qp_bit; struct ntb_rx_info __iomem *rx_info; struct ntb_rx_info *remote_rx_info; @@ -127,6 +151,7 @@ struct ntb_transport_qp { unsigned int rx_max_entry; unsigned int rx_max_frame; dma_cookie_t last_cookie; + struct tasklet_struct rxc_db_work; void (*event_handler)(void *data, int status); struct delayed_work link_work; @@ -153,33 +178,44 @@ struct ntb_transport_qp { }; struct ntb_transport_mw { - size_t size; + phys_addr_t phys_addr; + resource_size_t phys_size; + resource_size_t xlat_align; + resource_size_t xlat_align_size; + void __iomem *vbase; + size_t xlat_size; + size_t buff_size; void *virt_addr; dma_addr_t dma_addr; }; struct ntb_transport_client_dev { struct list_head entry; + struct ntb_transport_ctx *nt; struct device dev; }; -struct ntb_transport { +struct ntb_transport_ctx { struct list_head entry; struct list_head client_devs; - struct ntb_device *ndev; - struct ntb_transport_mw *mw; - struct ntb_transport_qp *qps; - unsigned int max_qps; - unsigned long qp_bitmap; - bool transport_link; + struct ntb_dev *ndev; + + struct ntb_transport_mw *mw_vec; + struct ntb_transport_qp *qp_vec; + unsigned int mw_count; + unsigned int qp_count; + u64 qp_bitmap; + u64 qp_bitmap_free; + + bool link_is_up; struct delayed_work link_work; struct work_struct link_cleanup; }; enum { - DESC_DONE_FLAG = 1 << 0, - LINK_DOWN_FLAG = 1 << 1, + DESC_DONE_FLAG = BIT(0), + LINK_DOWN_FLAG = BIT(1), }; struct ntb_payload_header { @@ -200,68 +236,69 @@ enum { MAX_SPAD, }; -#define QP_TO_MW(ndev, qp) ((qp) % ntb_max_mw(ndev)) +#define dev_client_dev(__dev) \ + container_of((__dev), struct ntb_transport_client_dev, dev) + +#define drv_client(__drv) \ + container_of((__drv), struct ntb_transport_client, driver) + +#define QP_TO_MW(nt, qp) ((qp) % nt->mw_count) #define NTB_QP_DEF_NUM_ENTRIES 100 #define NTB_LINK_DOWN_TIMEOUT 10 -static int ntb_match_bus(struct device *dev, struct device_driver *drv) +static void ntb_transport_rxc_db(unsigned long data); +static const struct ntb_ctx_ops ntb_transport_ops; +static struct ntb_client ntb_transport_client; + +static int ntb_transport_bus_match(struct device *dev, + struct device_driver *drv) { return !strncmp(dev_name(dev), drv->name, strlen(drv->name)); } -static int ntb_client_probe(struct device *dev) +static int ntb_transport_bus_probe(struct device *dev) { - const struct ntb_client *drv = container_of(dev->driver, - struct ntb_client, driver); - struct pci_dev *pdev = container_of(dev->parent, struct pci_dev, dev); + const struct ntb_transport_client *client; int rc = -EINVAL; get_device(dev); - if (drv && drv->probe) - rc = drv->probe(pdev); + + client = drv_client(dev->driver); + rc = client->probe(dev); if (rc) put_device(dev); return rc; } -static int ntb_client_remove(struct device *dev) +static int ntb_transport_bus_remove(struct device *dev) { - const struct ntb_client *drv = container_of(dev->driver, - struct ntb_client, driver); - struct pci_dev *pdev = container_of(dev->parent, struct pci_dev, dev); + const struct ntb_transport_client *client; - if (drv && drv->remove) - drv->remove(pdev); + client = drv_client(dev->driver); + client->remove(dev); put_device(dev); return 0; } -static struct bus_type ntb_bus_type = { - .name = "ntb_bus", - .match = ntb_match_bus, - .probe = ntb_client_probe, - .remove = ntb_client_remove, +static struct bus_type ntb_transport_bus = { + .name = "ntb_transport", + .match = ntb_transport_bus_match, + .probe = ntb_transport_bus_probe, + .remove = ntb_transport_bus_remove, }; static LIST_HEAD(ntb_transport_list); -static int ntb_bus_init(struct ntb_transport *nt) +static int ntb_bus_init(struct ntb_transport_ctx *nt) { - if (list_empty(&ntb_transport_list)) { - int rc = bus_register(&ntb_bus_type); - if (rc) - return rc; - } - list_add(&nt->entry, &ntb_transport_list); - return 0; } -static void ntb_bus_remove(struct ntb_transport *nt) +static void ntb_bus_remove(struct ntb_transport_ctx *nt) { struct ntb_transport_client_dev *client_dev, *cd; @@ -273,29 +310,26 @@ static void ntb_bus_remove(struct ntb_transport *nt) } list_del(&nt->entry); - - if (list_empty(&ntb_transport_list)) - bus_unregister(&ntb_bus_type); } -static void ntb_client_release(struct device *dev) +static void ntb_transport_client_release(struct device *dev) { struct ntb_transport_client_dev *client_dev; - client_dev = container_of(dev, struct ntb_transport_client_dev, dev); + client_dev = dev_client_dev(dev); kfree(client_dev); } /** - * ntb_unregister_client_dev - Unregister NTB client device + * ntb_transport_unregister_client_dev - Unregister NTB client device * @device_name: Name of NTB client device * * Unregister an NTB client device with the NTB transport layer */ -void ntb_unregister_client_dev(char *device_name) +void ntb_transport_unregister_client_dev(char *device_name) { struct ntb_transport_client_dev *client, *cd; - struct ntb_transport *nt; + struct ntb_transport_ctx *nt; list_for_each_entry(nt, &ntb_transport_list, entry) list_for_each_entry_safe(client, cd, &nt->client_devs, entry) @@ -305,18 +339,19 @@ void ntb_unregister_client_dev(char *device_name) device_unregister(&client->dev); } } -EXPORT_SYMBOL_GPL(ntb_unregister_client_dev); +EXPORT_SYMBOL_GPL(ntb_transport_unregister_client_dev); /** - * ntb_register_client_dev - Register NTB client device + * ntb_transport_register_client_dev - Register NTB client device * @device_name: Name of NTB client device * * Register an NTB client device with the NTB transport layer */ -int ntb_register_client_dev(char *device_name) +int ntb_transport_register_client_dev(char *device_name) { struct ntb_transport_client_dev *client_dev; - struct ntb_transport *nt; + struct ntb_transport_ctx *nt; + int node; int rc, i = 0; if (list_empty(&ntb_transport_list)) @@ -325,8 +360,10 @@ int ntb_register_client_dev(char *device_name) list_for_each_entry(nt, &ntb_transport_list, entry) { struct device *dev; - client_dev = kzalloc(sizeof(struct ntb_transport_client_dev), - GFP_KERNEL); + node = dev_to_node(&nt->ndev->dev); + + client_dev = kzalloc_node(sizeof(*client_dev), + GFP_KERNEL, node); if (!client_dev) { rc = -ENOMEM; goto err; @@ -336,9 +373,9 @@ int ntb_register_client_dev(char *device_name) /* setup and register client devices */ dev_set_name(dev, "%s%d", device_name, i); - dev->bus = &ntb_bus_type; - dev->release = ntb_client_release; - dev->parent = &ntb_query_pdev(nt->ndev)->dev; + dev->bus = &ntb_transport_bus; + dev->release = ntb_transport_client_release; + dev->parent = &nt->ndev->dev; rc = device_register(dev); if (rc) { @@ -353,44 +390,44 @@ int ntb_register_client_dev(char *device_name) return 0; err: - ntb_unregister_client_dev(device_name); + ntb_transport_unregister_client_dev(device_name); return rc; } -EXPORT_SYMBOL_GPL(ntb_register_client_dev); +EXPORT_SYMBOL_GPL(ntb_transport_register_client_dev); /** - * ntb_register_client - Register NTB client driver + * ntb_transport_register_client - Register NTB client driver * @drv: NTB client driver to be registered * * Register an NTB client driver with the NTB transport layer * * RETURNS: An appropriate -ERRNO error value on error, or zero for success. */ -int ntb_register_client(struct ntb_client *drv) +int ntb_transport_register_client(struct ntb_transport_client *drv) { - drv->driver.bus = &ntb_bus_type; + drv->driver.bus = &ntb_transport_bus; if (list_empty(&ntb_transport_list)) return -ENODEV; return driver_register(&drv->driver); } -EXPORT_SYMBOL_GPL(ntb_register_client); +EXPORT_SYMBOL_GPL(ntb_transport_register_client); /** - * ntb_unregister_client - Unregister NTB client driver + * ntb_transport_unregister_client - Unregister NTB client driver * @drv: NTB client driver to be unregistered * * Unregister an NTB client driver with the NTB transport layer * * RETURNS: An appropriate -ERRNO error value on error, or zero for success. */ -void ntb_unregister_client(struct ntb_client *drv) +void ntb_transport_unregister_client(struct ntb_transport_client *drv) { driver_unregister(&drv->driver); } -EXPORT_SYMBOL_GPL(ntb_unregister_client); +EXPORT_SYMBOL_GPL(ntb_transport_unregister_client); static ssize_t debugfs_read(struct file *filp, char __user *ubuf, size_t count, loff_t *offp) @@ -452,8 +489,8 @@ static ssize_t debugfs_read(struct file *filp, char __user *ubuf, size_t count, "tx_max_entry - \t%u\n", qp->tx_max_entry); out_offset += snprintf(buf + out_offset, out_count - out_offset, - "\nQP Link %s\n", (qp->qp_link == NTB_LINK_UP) ? - "Up" : "Down"); + "\nQP Link %s\n", + qp->link_is_up ? "Up" : "Down"); if (out_offset > out_count) out_offset = out_count; @@ -497,26 +534,31 @@ out: return entry; } -static void ntb_transport_setup_qp_mw(struct ntb_transport *nt, - unsigned int qp_num) +static int ntb_transport_setup_qp_mw(struct ntb_transport_ctx *nt, + unsigned int qp_num) { - struct ntb_transport_qp *qp = &nt->qps[qp_num]; + struct ntb_transport_qp *qp = &nt->qp_vec[qp_num]; + struct ntb_transport_mw *mw; unsigned int rx_size, num_qps_mw; - u8 mw_num, mw_max; + unsigned int mw_num, mw_count, qp_count; unsigned int i; - mw_max = ntb_max_mw(nt->ndev); - mw_num = QP_TO_MW(nt->ndev, qp_num); + mw_count = nt->mw_count; + qp_count = nt->qp_count; + + mw_num = QP_TO_MW(nt, qp_num); + mw = &nt->mw_vec[mw_num]; - WARN_ON(nt->mw[mw_num].virt_addr == NULL); + if (!mw->virt_addr) + return -ENOMEM; - if (nt->max_qps % mw_max && mw_num + 1 < nt->max_qps / mw_max) - num_qps_mw = nt->max_qps / mw_max + 1; + if (qp_count % mw_count && mw_num + 1 < qp_count / mw_count) + num_qps_mw = qp_count / mw_count + 1; else - num_qps_mw = nt->max_qps / mw_max; + num_qps_mw = qp_count / mw_count; - rx_size = (unsigned int) nt->mw[mw_num].size / num_qps_mw; - qp->rx_buff = nt->mw[mw_num].virt_addr + qp_num / mw_max * rx_size; + rx_size = (unsigned int)mw->xlat_size / num_qps_mw; + qp->rx_buff = mw->virt_addr + rx_size * qp_num / mw_count; rx_size -= sizeof(struct ntb_rx_info); qp->remote_rx_info = qp->rx_buff + rx_size; @@ -530,49 +572,63 @@ static void ntb_transport_setup_qp_mw(struct ntb_transport *nt, /* setup the hdr offsets with 0's */ for (i = 0; i < qp->rx_max_entry; i++) { - void *offset = qp->rx_buff + qp->rx_max_frame * (i + 1) - - sizeof(struct ntb_payload_header); + void *offset = (qp->rx_buff + qp->rx_max_frame * (i + 1) - + sizeof(struct ntb_payload_header)); memset(offset, 0, sizeof(struct ntb_payload_header)); } qp->rx_pkts = 0; qp->tx_pkts = 0; qp->tx_index = 0; + + return 0; } -static void ntb_free_mw(struct ntb_transport *nt, int num_mw) +static void ntb_free_mw(struct ntb_transport_ctx *nt, int num_mw) { - struct ntb_transport_mw *mw = &nt->mw[num_mw]; - struct pci_dev *pdev = ntb_query_pdev(nt->ndev); + struct ntb_transport_mw *mw = &nt->mw_vec[num_mw]; + struct pci_dev *pdev = nt->ndev->pdev; if (!mw->virt_addr) return; - dma_free_coherent(&pdev->dev, mw->size, mw->virt_addr, mw->dma_addr); + ntb_mw_clear_trans(nt->ndev, num_mw); + dma_free_coherent(&pdev->dev, mw->buff_size, + mw->virt_addr, mw->dma_addr); + mw->xlat_size = 0; + mw->buff_size = 0; mw->virt_addr = NULL; } -static int ntb_set_mw(struct ntb_transport *nt, int num_mw, unsigned int size) +static int ntb_set_mw(struct ntb_transport_ctx *nt, int num_mw, + unsigned int size) { - struct ntb_transport_mw *mw = &nt->mw[num_mw]; - struct pci_dev *pdev = ntb_query_pdev(nt->ndev); + struct ntb_transport_mw *mw = &nt->mw_vec[num_mw]; + struct pci_dev *pdev = nt->ndev->pdev; + unsigned int xlat_size, buff_size; + int rc; + + xlat_size = round_up(size, mw->xlat_align_size); + buff_size = round_up(size, mw->xlat_align); /* No need to re-setup */ - if (mw->size == ALIGN(size, 4096)) + if (mw->xlat_size == xlat_size) return 0; - if (mw->size != 0) + if (mw->buff_size) ntb_free_mw(nt, num_mw); - /* Alloc memory for receiving data. Must be 4k aligned */ - mw->size = ALIGN(size, 4096); + /* Alloc memory for receiving data. Must be aligned */ + mw->xlat_size = xlat_size; + mw->buff_size = buff_size; - mw->virt_addr = dma_alloc_coherent(&pdev->dev, mw->size, &mw->dma_addr, - GFP_KERNEL); + mw->virt_addr = dma_alloc_coherent(&pdev->dev, buff_size, + &mw->dma_addr, GFP_KERNEL); if (!mw->virt_addr) { - mw->size = 0; - dev_err(&pdev->dev, "Unable to allocate MW buffer of size %d\n", - (int) mw->size); + mw->xlat_size = 0; + mw->buff_size = 0; + dev_err(&pdev->dev, "Unable to alloc MW buff of size %d\n", + buff_size); return -ENOMEM; } @@ -582,34 +638,58 @@ static int ntb_set_mw(struct ntb_transport *nt, int num_mw, unsigned int size) * is a requirement of the hardware. It is recommended to setup CMA * for BAR sizes equal or greater than 4MB. */ - if (!IS_ALIGNED(mw->dma_addr, mw->size)) { - dev_err(&pdev->dev, "DMA memory %pad not aligned to BAR size\n", + if (!IS_ALIGNED(mw->dma_addr, mw->xlat_align)) { + dev_err(&pdev->dev, "DMA memory %pad is not aligned\n", &mw->dma_addr); ntb_free_mw(nt, num_mw); return -ENOMEM; } /* Notify HW the memory location of the receive buffer */ - ntb_set_mw_addr(nt->ndev, num_mw, mw->dma_addr); + rc = ntb_mw_set_trans(nt->ndev, num_mw, mw->dma_addr, mw->xlat_size); + if (rc) { + dev_err(&pdev->dev, "Unable to set mw%d translation", num_mw); + ntb_free_mw(nt, num_mw); + return -EIO; + } return 0; } +static void ntb_qp_link_down_reset(struct ntb_transport_qp *qp) +{ + qp->link_is_up = false; + + qp->tx_index = 0; + qp->rx_index = 0; + qp->rx_bytes = 0; + qp->rx_pkts = 0; + qp->rx_ring_empty = 0; + qp->rx_err_no_buf = 0; + qp->rx_err_oflow = 0; + qp->rx_err_ver = 0; + qp->rx_memcpy = 0; + qp->rx_async = 0; + qp->tx_bytes = 0; + qp->tx_pkts = 0; + qp->tx_ring_full = 0; + qp->tx_err_no_buf = 0; + qp->tx_memcpy = 0; + qp->tx_async = 0; +} + static void ntb_qp_link_cleanup(struct ntb_transport_qp *qp) { - struct ntb_transport *nt = qp->transport; - struct pci_dev *pdev = ntb_query_pdev(nt->ndev); + struct ntb_transport_ctx *nt = qp->transport; + struct pci_dev *pdev = nt->ndev->pdev; - if (qp->qp_link == NTB_LINK_DOWN) { - cancel_delayed_work_sync(&qp->link_work); - return; - } + dev_info(&pdev->dev, "qp %d: Link Cleanup\n", qp->qp_num); - if (qp->event_handler) - qp->event_handler(qp->cb_data, NTB_LINK_DOWN); + cancel_delayed_work_sync(&qp->link_work); + ntb_qp_link_down_reset(qp); - dev_info(&pdev->dev, "qp %d: Link Down\n", qp->qp_num); - qp->qp_link = NTB_LINK_DOWN; + if (qp->event_handler) + qp->event_handler(qp->cb_data, qp->link_is_up); } static void ntb_qp_link_cleanup_work(struct work_struct *work) @@ -617,11 +697,11 @@ static void ntb_qp_link_cleanup_work(struct work_struct *work) struct ntb_transport_qp *qp = container_of(work, struct ntb_transport_qp, link_cleanup); - struct ntb_transport *nt = qp->transport; + struct ntb_transport_ctx *nt = qp->transport; ntb_qp_link_cleanup(qp); - if (nt->transport_link == NTB_LINK_UP) + if (nt->link_is_up) schedule_delayed_work(&qp->link_work, msecs_to_jiffies(NTB_LINK_DOWN_TIMEOUT)); } @@ -631,180 +711,132 @@ static void ntb_qp_link_down(struct ntb_transport_qp *qp) schedule_work(&qp->link_cleanup); } -static void ntb_transport_link_cleanup(struct ntb_transport *nt) +static void ntb_transport_link_cleanup(struct ntb_transport_ctx *nt) { + struct ntb_transport_qp *qp; + u64 qp_bitmap_alloc; int i; + qp_bitmap_alloc = nt->qp_bitmap & ~nt->qp_bitmap_free; + /* Pass along the info to any clients */ - for (i = 0; i < nt->max_qps; i++) - if (!test_bit(i, &nt->qp_bitmap)) - ntb_qp_link_cleanup(&nt->qps[i]); + for (i = 0; i < nt->qp_count; i++) + if (qp_bitmap_alloc & BIT_ULL(i)) { + qp = &nt->qp_vec[i]; + ntb_qp_link_cleanup(qp); + cancel_work_sync(&qp->link_cleanup); + cancel_delayed_work_sync(&qp->link_work); + } - if (nt->transport_link == NTB_LINK_DOWN) + if (!nt->link_is_up) cancel_delayed_work_sync(&nt->link_work); - else - nt->transport_link = NTB_LINK_DOWN; /* The scratchpad registers keep the values if the remote side * goes down, blast them now to give them a sane value the next * time they are accessed */ for (i = 0; i < MAX_SPAD; i++) - ntb_write_local_spad(nt->ndev, i, 0); + ntb_spad_write(nt->ndev, i, 0); } static void ntb_transport_link_cleanup_work(struct work_struct *work) { - struct ntb_transport *nt = container_of(work, struct ntb_transport, - link_cleanup); + struct ntb_transport_ctx *nt = + container_of(work, struct ntb_transport_ctx, link_cleanup); ntb_transport_link_cleanup(nt); } -static void ntb_transport_event_callback(void *data, enum ntb_hw_event event) +static void ntb_transport_event_callback(void *data) { - struct ntb_transport *nt = data; + struct ntb_transport_ctx *nt = data; - switch (event) { - case NTB_EVENT_HW_LINK_UP: + if (ntb_link_is_up(nt->ndev, NULL, NULL) == 1) schedule_delayed_work(&nt->link_work, 0); - break; - case NTB_EVENT_HW_LINK_DOWN: + else schedule_work(&nt->link_cleanup); - break; - default: - BUG(); - } } static void ntb_transport_link_work(struct work_struct *work) { - struct ntb_transport *nt = container_of(work, struct ntb_transport, - link_work.work); - struct ntb_device *ndev = nt->ndev; - struct pci_dev *pdev = ntb_query_pdev(ndev); + struct ntb_transport_ctx *nt = + container_of(work, struct ntb_transport_ctx, link_work.work); + struct ntb_dev *ndev = nt->ndev; + struct pci_dev *pdev = ndev->pdev; + resource_size_t size; u32 val; - int rc, i; + int rc, i, spad; /* send the local info, in the opposite order of the way we read it */ - for (i = 0; i < ntb_max_mw(ndev); i++) { - rc = ntb_write_remote_spad(ndev, MW0_SZ_HIGH + (i * 2), - ntb_get_mw_size(ndev, i) >> 32); - if (rc) { - dev_err(&pdev->dev, "Error writing %u to remote spad %d\n", - (u32)(ntb_get_mw_size(ndev, i) >> 32), - MW0_SZ_HIGH + (i * 2)); - goto out; - } + for (i = 0; i < nt->mw_count; i++) { + size = nt->mw_vec[i].phys_size; - rc = ntb_write_remote_spad(ndev, MW0_SZ_LOW + (i * 2), - (u32) ntb_get_mw_size(ndev, i)); - if (rc) { - dev_err(&pdev->dev, "Error writing %u to remote spad %d\n", - (u32) ntb_get_mw_size(ndev, i), - MW0_SZ_LOW + (i * 2)); - goto out; - } - } + if (max_mw_size && size > max_mw_size) + size = max_mw_size; - rc = ntb_write_remote_spad(ndev, NUM_MWS, ntb_max_mw(ndev)); - if (rc) { - dev_err(&pdev->dev, "Error writing %x to remote spad %d\n", - ntb_max_mw(ndev), NUM_MWS); - goto out; - } + spad = MW0_SZ_HIGH + (i * 2); + ntb_peer_spad_write(ndev, spad, (u32)(size >> 32)); - rc = ntb_write_remote_spad(ndev, NUM_QPS, nt->max_qps); - if (rc) { - dev_err(&pdev->dev, "Error writing %x to remote spad %d\n", - nt->max_qps, NUM_QPS); - goto out; + spad = MW0_SZ_LOW + (i * 2); + ntb_peer_spad_write(ndev, spad, (u32)size); } - rc = ntb_write_remote_spad(ndev, VERSION, NTB_TRANSPORT_VERSION); - if (rc) { - dev_err(&pdev->dev, "Error writing %x to remote spad %d\n", - NTB_TRANSPORT_VERSION, VERSION); - goto out; - } + ntb_peer_spad_write(ndev, NUM_MWS, nt->mw_count); - /* Query the remote side for its info */ - rc = ntb_read_remote_spad(ndev, VERSION, &val); - if (rc) { - dev_err(&pdev->dev, "Error reading remote spad %d\n", VERSION); - goto out; - } + ntb_peer_spad_write(ndev, NUM_QPS, nt->qp_count); - if (val != NTB_TRANSPORT_VERSION) - goto out; - dev_dbg(&pdev->dev, "Remote version = %d\n", val); + ntb_peer_spad_write(ndev, VERSION, NTB_TRANSPORT_VERSION); - rc = ntb_read_remote_spad(ndev, NUM_QPS, &val); - if (rc) { - dev_err(&pdev->dev, "Error reading remote spad %d\n", NUM_QPS); + /* Query the remote side for its info */ + val = ntb_spad_read(ndev, VERSION); + dev_dbg(&pdev->dev, "Remote version = %d\n", val); + if (val != NTB_TRANSPORT_VERSION) goto out; - } - if (val != nt->max_qps) - goto out; + val = ntb_spad_read(ndev, NUM_QPS); dev_dbg(&pdev->dev, "Remote max number of qps = %d\n", val); - - rc = ntb_read_remote_spad(ndev, NUM_MWS, &val); - if (rc) { - dev_err(&pdev->dev, "Error reading remote spad %d\n", NUM_MWS); + if (val != nt->qp_count) goto out; - } - if (val != ntb_max_mw(ndev)) - goto out; + val = ntb_spad_read(ndev, NUM_MWS); dev_dbg(&pdev->dev, "Remote number of mws = %d\n", val); + if (val != nt->mw_count) + goto out; - for (i = 0; i < ntb_max_mw(ndev); i++) { + for (i = 0; i < nt->mw_count; i++) { u64 val64; - rc = ntb_read_remote_spad(ndev, MW0_SZ_HIGH + (i * 2), &val); - if (rc) { - dev_err(&pdev->dev, "Error reading remote spad %d\n", - MW0_SZ_HIGH + (i * 2)); - goto out1; - } - - val64 = (u64) val << 32; - - rc = ntb_read_remote_spad(ndev, MW0_SZ_LOW + (i * 2), &val); - if (rc) { - dev_err(&pdev->dev, "Error reading remote spad %d\n", - MW0_SZ_LOW + (i * 2)); - goto out1; - } + val = ntb_spad_read(ndev, MW0_SZ_HIGH + (i * 2)); + val64 = (u64)val << 32; + val = ntb_spad_read(ndev, MW0_SZ_LOW + (i * 2)); val64 |= val; - dev_dbg(&pdev->dev, "Remote MW%d size = %llu\n", i, val64); + dev_dbg(&pdev->dev, "Remote MW%d size = %#llx\n", i, val64); rc = ntb_set_mw(nt, i, val64); if (rc) goto out1; } - nt->transport_link = NTB_LINK_UP; + nt->link_is_up = true; - for (i = 0; i < nt->max_qps; i++) { - struct ntb_transport_qp *qp = &nt->qps[i]; + for (i = 0; i < nt->qp_count; i++) { + struct ntb_transport_qp *qp = &nt->qp_vec[i]; ntb_transport_setup_qp_mw(nt, i); - if (qp->client_ready == NTB_LINK_UP) + if (qp->client_ready) schedule_delayed_work(&qp->link_work, 0); } return; out1: - for (i = 0; i < ntb_max_mw(ndev); i++) + for (i = 0; i < nt->mw_count; i++) ntb_free_mw(nt, i); out: - if (ntb_hw_link_status(ndev)) + if (ntb_link_is_up(ndev, NULL, NULL) == 1) schedule_delayed_work(&nt->link_work, msecs_to_jiffies(NTB_LINK_DOWN_TIMEOUT)); } @@ -814,73 +846,73 @@ static void ntb_qp_link_work(struct work_struct *work) struct ntb_transport_qp *qp = container_of(work, struct ntb_transport_qp, link_work.work); - struct pci_dev *pdev = ntb_query_pdev(qp->ndev); - struct ntb_transport *nt = qp->transport; - int rc, val; + struct pci_dev *pdev = qp->ndev->pdev; + struct ntb_transport_ctx *nt = qp->transport; + int val; - WARN_ON(nt->transport_link != NTB_LINK_UP); + WARN_ON(!nt->link_is_up); - rc = ntb_read_local_spad(nt->ndev, QP_LINKS, &val); - if (rc) { - dev_err(&pdev->dev, "Error reading spad %d\n", QP_LINKS); - return; - } + val = ntb_spad_read(nt->ndev, QP_LINKS); - rc = ntb_write_remote_spad(nt->ndev, QP_LINKS, val | 1 << qp->qp_num); - if (rc) - dev_err(&pdev->dev, "Error writing %x to remote spad %d\n", - val | 1 << qp->qp_num, QP_LINKS); + ntb_peer_spad_write(nt->ndev, QP_LINKS, val | BIT(qp->qp_num)); /* query remote spad for qp ready bits */ - rc = ntb_read_remote_spad(nt->ndev, QP_LINKS, &val); - if (rc) - dev_err(&pdev->dev, "Error reading remote spad %d\n", QP_LINKS); - - dev_dbg(&pdev->dev, "Remote QP link status = %x\n", val); + ntb_peer_spad_read(nt->ndev, QP_LINKS); + dev_dbg_ratelimited(&pdev->dev, "Remote QP link status = %x\n", val); /* See if the remote side is up */ - if (1 << qp->qp_num & val) { - qp->qp_link = NTB_LINK_UP; - + if (val & BIT(qp->qp_num)) { dev_info(&pdev->dev, "qp %d: Link Up\n", qp->qp_num); + qp->link_is_up = true; + if (qp->event_handler) - qp->event_handler(qp->cb_data, NTB_LINK_UP); - } else if (nt->transport_link == NTB_LINK_UP) + qp->event_handler(qp->cb_data, qp->link_is_up); + } else if (nt->link_is_up) schedule_delayed_work(&qp->link_work, msecs_to_jiffies(NTB_LINK_DOWN_TIMEOUT)); } -static int ntb_transport_init_queue(struct ntb_transport *nt, +static int ntb_transport_init_queue(struct ntb_transport_ctx *nt, unsigned int qp_num) { struct ntb_transport_qp *qp; + struct ntb_transport_mw *mw; + phys_addr_t mw_base; + resource_size_t mw_size; unsigned int num_qps_mw, tx_size; - u8 mw_num, mw_max; + unsigned int mw_num, mw_count, qp_count; u64 qp_offset; - mw_max = ntb_max_mw(nt->ndev); - mw_num = QP_TO_MW(nt->ndev, qp_num); + mw_count = nt->mw_count; + qp_count = nt->qp_count; - qp = &nt->qps[qp_num]; + mw_num = QP_TO_MW(nt, qp_num); + mw = &nt->mw_vec[mw_num]; + + qp = &nt->qp_vec[qp_num]; qp->qp_num = qp_num; qp->transport = nt; qp->ndev = nt->ndev; - qp->qp_link = NTB_LINK_DOWN; - qp->client_ready = NTB_LINK_DOWN; + qp->client_ready = false; qp->event_handler = NULL; + ntb_qp_link_down_reset(qp); - if (nt->max_qps % mw_max && mw_num + 1 < nt->max_qps / mw_max) - num_qps_mw = nt->max_qps / mw_max + 1; + if (qp_count % mw_count && mw_num + 1 < qp_count / mw_count) + num_qps_mw = qp_count / mw_count + 1; else - num_qps_mw = nt->max_qps / mw_max; + num_qps_mw = qp_count / mw_count; + + mw_base = nt->mw_vec[mw_num].phys_addr; + mw_size = nt->mw_vec[mw_num].phys_size; - tx_size = (unsigned int) ntb_get_mw_size(qp->ndev, mw_num) / num_qps_mw; - qp_offset = qp_num / mw_max * tx_size; - qp->tx_mw = ntb_get_mw_vbase(nt->ndev, mw_num) + qp_offset; + tx_size = (unsigned int)mw_size / num_qps_mw; + qp_offset = tx_size * qp_num / mw_count; + + qp->tx_mw = nt->mw_vec[mw_num].vbase + qp_offset; if (!qp->tx_mw) return -EINVAL; - qp->tx_mw_phys = ntb_get_mw_base(qp->ndev, mw_num) + qp_offset; + qp->tx_mw_phys = mw_base + qp_offset; if (!qp->tx_mw_phys) return -EINVAL; @@ -891,16 +923,19 @@ static int ntb_transport_init_queue(struct ntb_transport *nt, qp->tx_max_frame = min(transport_mtu, tx_size / 2); qp->tx_max_entry = tx_size / qp->tx_max_frame; - if (ntb_query_debugfs(nt->ndev)) { + if (nt_debugfs_dir) { char debugfs_name[4]; snprintf(debugfs_name, 4, "qp%d", qp_num); qp->debugfs_dir = debugfs_create_dir(debugfs_name, - ntb_query_debugfs(nt->ndev)); + nt_debugfs_dir); qp->debugfs_stats = debugfs_create_file("stats", S_IRUSR, qp->debugfs_dir, qp, &ntb_qp_debugfs_stats); + } else { + qp->debugfs_dir = NULL; + qp->debugfs_stats = NULL; } INIT_DELAYED_WORK(&qp->link_work, ntb_qp_link_work); @@ -914,46 +949,89 @@ static int ntb_transport_init_queue(struct ntb_transport *nt, INIT_LIST_HEAD(&qp->rx_free_q); INIT_LIST_HEAD(&qp->tx_free_q); + tasklet_init(&qp->rxc_db_work, ntb_transport_rxc_db, + (unsigned long)qp); + return 0; } -int ntb_transport_init(struct pci_dev *pdev) +static int ntb_transport_probe(struct ntb_client *self, struct ntb_dev *ndev) { - struct ntb_transport *nt; + struct ntb_transport_ctx *nt; + struct ntb_transport_mw *mw; + unsigned int mw_count, qp_count; + u64 qp_bitmap; + int node; int rc, i; - nt = kzalloc(sizeof(struct ntb_transport), GFP_KERNEL); + if (ntb_db_is_unsafe(ndev)) + dev_dbg(&ndev->dev, + "doorbell is unsafe, proceed anyway...\n"); + if (ntb_spad_is_unsafe(ndev)) + dev_dbg(&ndev->dev, + "scratchpad is unsafe, proceed anyway...\n"); + + node = dev_to_node(&ndev->dev); + + nt = kzalloc_node(sizeof(*nt), GFP_KERNEL, node); if (!nt) return -ENOMEM; - nt->ndev = ntb_register_transport(pdev, nt); - if (!nt->ndev) { - rc = -EIO; + nt->ndev = ndev; + + mw_count = ntb_mw_count(ndev); + + nt->mw_count = mw_count; + + nt->mw_vec = kzalloc_node(mw_count * sizeof(*nt->mw_vec), + GFP_KERNEL, node); + if (!nt->mw_vec) { + rc = -ENOMEM; goto err; } - nt->mw = kcalloc(ntb_max_mw(nt->ndev), sizeof(struct ntb_transport_mw), - GFP_KERNEL); - if (!nt->mw) { - rc = -ENOMEM; - goto err1; + for (i = 0; i < mw_count; i++) { + mw = &nt->mw_vec[i]; + + rc = ntb_mw_get_range(ndev, i, &mw->phys_addr, &mw->phys_size, + &mw->xlat_align, &mw->xlat_align_size); + if (rc) + goto err1; + + mw->vbase = ioremap_wc(mw->phys_addr, mw->phys_size); + if (!mw->vbase) { + rc = -ENOMEM; + goto err1; + } + + mw->buff_size = 0; + mw->xlat_size = 0; + mw->virt_addr = NULL; + mw->dma_addr = 0; } - if (max_num_clients) - nt->max_qps = min(ntb_max_cbs(nt->ndev), max_num_clients); - else - nt->max_qps = min(ntb_max_cbs(nt->ndev), ntb_max_mw(nt->ndev)); + qp_bitmap = ntb_db_valid_mask(ndev); + + qp_count = ilog2(qp_bitmap); + if (max_num_clients && max_num_clients < qp_count) + qp_count = max_num_clients; + else if (mw_count < qp_count) + qp_count = mw_count; + + qp_bitmap &= BIT_ULL(qp_count) - 1; - nt->qps = kcalloc(nt->max_qps, sizeof(struct ntb_transport_qp), - GFP_KERNEL); - if (!nt->qps) { + nt->qp_count = qp_count; + nt->qp_bitmap = qp_bitmap; + nt->qp_bitmap_free = qp_bitmap; + + nt->qp_vec = kzalloc_node(qp_count * sizeof(*nt->qp_vec), + GFP_KERNEL, node); + if (!nt->qp_vec) { rc = -ENOMEM; goto err2; } - nt->qp_bitmap = ((u64) 1 << nt->max_qps) - 1; - - for (i = 0; i < nt->max_qps; i++) { + for (i = 0; i < qp_count; i++) { rc = ntb_transport_init_queue(nt, i); if (rc) goto err3; @@ -962,8 +1040,7 @@ int ntb_transport_init(struct pci_dev *pdev) INIT_DELAYED_WORK(&nt->link_work, ntb_transport_link_work); INIT_WORK(&nt->link_cleanup, ntb_transport_link_cleanup_work); - rc = ntb_register_event_callback(nt->ndev, - ntb_transport_event_callback); + rc = ntb_set_ctx(ndev, nt, &ntb_transport_ops); if (rc) goto err3; @@ -972,51 +1049,61 @@ int ntb_transport_init(struct pci_dev *pdev) if (rc) goto err4; - if (ntb_hw_link_status(nt->ndev)) - schedule_delayed_work(&nt->link_work, 0); + nt->link_is_up = false; + ntb_link_enable(ndev, NTB_SPEED_AUTO, NTB_WIDTH_AUTO); + ntb_link_event(ndev); return 0; err4: - ntb_unregister_event_callback(nt->ndev); + ntb_clear_ctx(ndev); err3: - kfree(nt->qps); + kfree(nt->qp_vec); err2: - kfree(nt->mw); + kfree(nt->mw_vec); err1: - ntb_unregister_transport(nt->ndev); + while (i--) { + mw = &nt->mw_vec[i]; + iounmap(mw->vbase); + } err: kfree(nt); return rc; } -void ntb_transport_free(void *transport) +static void ntb_transport_free(struct ntb_client *self, struct ntb_dev *ndev) { - struct ntb_transport *nt = transport; - struct ntb_device *ndev = nt->ndev; + struct ntb_transport_ctx *nt = ndev->ctx; + struct ntb_transport_qp *qp; + u64 qp_bitmap_alloc; int i; ntb_transport_link_cleanup(nt); + cancel_work_sync(&nt->link_cleanup); + cancel_delayed_work_sync(&nt->link_work); + + qp_bitmap_alloc = nt->qp_bitmap & ~nt->qp_bitmap_free; /* verify that all the qp's are freed */ - for (i = 0; i < nt->max_qps; i++) { - if (!test_bit(i, &nt->qp_bitmap)) - ntb_transport_free_queue(&nt->qps[i]); - debugfs_remove_recursive(nt->qps[i].debugfs_dir); + for (i = 0; i < nt->qp_count; i++) { + qp = &nt->qp_vec[i]; + if (qp_bitmap_alloc & BIT_ULL(i)) + ntb_transport_free_queue(qp); + debugfs_remove_recursive(qp->debugfs_dir); } - ntb_bus_remove(nt); + ntb_link_disable(ndev); + ntb_clear_ctx(ndev); - cancel_delayed_work_sync(&nt->link_work); - - ntb_unregister_event_callback(ndev); + ntb_bus_remove(nt); - for (i = 0; i < ntb_max_mw(ndev); i++) + for (i = nt->mw_count; i--; ) { ntb_free_mw(nt, i); + iounmap(nt->mw_vec[i].vbase); + } - kfree(nt->qps); - kfree(nt->mw); - ntb_unregister_transport(ndev); + kfree(nt->qp_vec); + kfree(nt->mw_vec); kfree(nt); } @@ -1028,15 +1115,13 @@ static void ntb_rx_copy_callback(void *data) unsigned int len = entry->len; struct ntb_payload_header *hdr = entry->rx_hdr; - /* Ensure that the data is fully copied out before clearing the flag */ - wmb(); hdr->flags = 0; iowrite32(entry->index, &qp->rx_info->entry); ntb_list_add(&qp->ntb_rx_free_q_lock, &entry->entry, &qp->rx_free_q); - if (qp->rx_handler && qp->client_ready == NTB_LINK_UP) + if (qp->rx_handler && qp->client_ready) qp->rx_handler(qp, qp->cb_data, cb_data, len); } @@ -1047,6 +1132,9 @@ static void ntb_memcpy_rx(struct ntb_queue_entry *entry, void *offset) memcpy(buf, offset, len); + /* Ensure that the data is fully copied out before clearing the flag */ + wmb(); + ntb_rx_copy_callback(entry); } @@ -1071,8 +1159,8 @@ static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset, goto err_wait; device = chan->device; - pay_off = (size_t) offset & ~PAGE_MASK; - buff_off = (size_t) buf & ~PAGE_MASK; + pay_off = (size_t)offset & ~PAGE_MASK; + buff_off = (size_t)buf & ~PAGE_MASK; if (!is_dma_copy_aligned(device, pay_off, buff_off, len)) goto err_wait; @@ -1138,86 +1226,103 @@ static int ntb_process_rxc(struct ntb_transport_qp *qp) struct ntb_payload_header *hdr; struct ntb_queue_entry *entry; void *offset; + int rc; offset = qp->rx_buff + qp->rx_max_frame * qp->rx_index; hdr = offset + qp->rx_max_frame - sizeof(struct ntb_payload_header); - entry = ntb_list_rm(&qp->ntb_rx_pend_q_lock, &qp->rx_pend_q); - if (!entry) { - dev_dbg(&ntb_query_pdev(qp->ndev)->dev, - "no buffer - HDR ver %u, len %d, flags %x\n", - hdr->ver, hdr->len, hdr->flags); - qp->rx_err_no_buf++; - return -ENOMEM; - } + dev_dbg(&qp->ndev->pdev->dev, "qp %d: RX ver %u len %d flags %x\n", + qp->qp_num, hdr->ver, hdr->len, hdr->flags); if (!(hdr->flags & DESC_DONE_FLAG)) { - ntb_list_add(&qp->ntb_rx_pend_q_lock, &entry->entry, - &qp->rx_pend_q); + dev_dbg(&qp->ndev->pdev->dev, "done flag not set\n"); qp->rx_ring_empty++; return -EAGAIN; } - if (hdr->ver != (u32) qp->rx_pkts) { - dev_dbg(&ntb_query_pdev(qp->ndev)->dev, - "qp %d: version mismatch, expected %llu - got %u\n", - qp->qp_num, qp->rx_pkts, hdr->ver); - ntb_list_add(&qp->ntb_rx_pend_q_lock, &entry->entry, - &qp->rx_pend_q); + if (hdr->flags & LINK_DOWN_FLAG) { + dev_dbg(&qp->ndev->pdev->dev, "link down flag set\n"); + ntb_qp_link_down(qp); + hdr->flags = 0; + return -EAGAIN; + } + + if (hdr->ver != (u32)qp->rx_pkts) { + dev_dbg(&qp->ndev->pdev->dev, + "version mismatch, expected %llu - got %u\n", + qp->rx_pkts, hdr->ver); qp->rx_err_ver++; return -EIO; } - if (hdr->flags & LINK_DOWN_FLAG) { - ntb_qp_link_down(qp); + entry = ntb_list_rm(&qp->ntb_rx_pend_q_lock, &qp->rx_pend_q); + if (!entry) { + dev_dbg(&qp->ndev->pdev->dev, "no receive buffer\n"); + qp->rx_err_no_buf++; + rc = -ENOMEM; goto err; } - dev_dbg(&ntb_query_pdev(qp->ndev)->dev, - "rx offset %u, ver %u - %d payload received, buf size %d\n", - qp->rx_index, hdr->ver, hdr->len, entry->len); - - qp->rx_bytes += hdr->len; - qp->rx_pkts++; - if (hdr->len > entry->len) { - qp->rx_err_oflow++; - dev_dbg(&ntb_query_pdev(qp->ndev)->dev, - "RX overflow! Wanted %d got %d\n", + dev_dbg(&qp->ndev->pdev->dev, + "receive buffer overflow! Wanted %d got %d\n", hdr->len, entry->len); + qp->rx_err_oflow++; + rc = -EIO; goto err; } + dev_dbg(&qp->ndev->pdev->dev, + "RX OK index %u ver %u size %d into buf size %d\n", + qp->rx_index, hdr->ver, hdr->len, entry->len); + + qp->rx_bytes += hdr->len; + qp->rx_pkts++; + entry->index = qp->rx_index; entry->rx_hdr = hdr; ntb_async_rx(entry, offset, hdr->len); -out: qp->rx_index++; qp->rx_index %= qp->rx_max_entry; return 0; err: - ntb_list_add(&qp->ntb_rx_pend_q_lock, &entry->entry, &qp->rx_pend_q); - /* Ensure that the data is fully copied out before clearing the flag */ - wmb(); + /* FIXME: if this syncrhonous update of the rx_index gets ahead of + * asyncrhonous ntb_rx_copy_callback of previous entry, there are three + * scenarios: + * + * 1) The peer might miss this update, but observe the update + * from the memcpy completion callback. In this case, the buffer will + * not be freed on the peer to be reused for a different packet. The + * successful rx of a later packet would clear the condition, but the + * condition could persist if several rx fail in a row. + * + * 2) The peer may observe this update before the asyncrhonous copy of + * prior packets is completed. The peer may overwrite the buffers of + * the prior packets before they are copied. + * + * 3) Both: the peer may observe the update, and then observe the index + * decrement by the asynchronous completion callback. Who knows what + * badness that will cause. + */ hdr->flags = 0; iowrite32(qp->rx_index, &qp->rx_info->entry); - goto out; + return rc; } -static int ntb_transport_rxc_db(void *data, int db_num) +static void ntb_transport_rxc_db(unsigned long data) { - struct ntb_transport_qp *qp = data; + struct ntb_transport_qp *qp = (void *)data; int rc, i; - dev_dbg(&ntb_query_pdev(qp->ndev)->dev, "%s: doorbell %d received\n", - __func__, db_num); + dev_dbg(&qp->ndev->pdev->dev, "%s: doorbell %d received\n", + __func__, qp->qp_num); /* Limit the number of packets processed in a single interrupt to * provide fairness to others @@ -1231,7 +1336,21 @@ static int ntb_transport_rxc_db(void *data, int db_num) if (qp->dma_chan) dma_async_issue_pending(qp->dma_chan); - return i; + if (i == qp->rx_max_entry) { + /* there is more work to do */ + tasklet_schedule(&qp->rxc_db_work); + } else if (ntb_db_read(qp->ndev) & BIT_ULL(qp->qp_num)) { + /* the doorbell bit is set: clear it */ + ntb_db_clear(qp->ndev, BIT_ULL(qp->qp_num)); + /* ntb_db_read ensures ntb_db_clear write is committed */ + ntb_db_read(qp->ndev); + + /* an interrupt may have arrived between finishing + * ntb_process_rxc and clearing the doorbell bit: + * there might be some more work to do. + */ + tasklet_schedule(&qp->rxc_db_work); + } } static void ntb_tx_copy_callback(void *data) @@ -1240,11 +1359,9 @@ static void ntb_tx_copy_callback(void *data) struct ntb_transport_qp *qp = entry->qp; struct ntb_payload_header __iomem *hdr = entry->tx_hdr; - /* Ensure that the data is fully copied out before setting the flags */ - wmb(); iowrite32(entry->flags | DESC_DONE_FLAG, &hdr->flags); - ntb_ring_doorbell(qp->ndev, qp->qp_num); + ntb_peer_db_set(qp->ndev, BIT_ULL(qp->qp_num)); /* The entry length can only be zero if the packet is intended to be a * "link down" or similar. Since no payload is being sent in these @@ -1263,7 +1380,18 @@ static void ntb_tx_copy_callback(void *data) static void ntb_memcpy_tx(struct ntb_queue_entry *entry, void __iomem *offset) { +#ifdef ARCH_HAS_NOCACHE_UACCESS + /* + * Using non-temporal mov to improve performance on non-cached + * writes, even though we aren't actually copying from user space. + */ + __copy_from_user_inatomic_nocache(offset, entry->buf, entry->len); +#else memcpy_toio(offset, entry->buf, entry->len); +#endif + + /* Ensure that the data is fully copied out before setting the flags */ + wmb(); ntb_tx_copy_callback(entry); } @@ -1288,7 +1416,7 @@ static void ntb_async_tx(struct ntb_transport_qp *qp, entry->tx_hdr = hdr; iowrite32(entry->len, &hdr->len); - iowrite32((u32) qp->tx_pkts, &hdr->ver); + iowrite32((u32)qp->tx_pkts, &hdr->ver); if (!chan) goto err; @@ -1298,8 +1426,8 @@ static void ntb_async_tx(struct ntb_transport_qp *qp, device = chan->device; dest = qp->tx_mw_phys + qp->tx_max_frame * qp->tx_index; - buff_off = (size_t) buf & ~PAGE_MASK; - dest_off = (size_t) dest & ~PAGE_MASK; + buff_off = (size_t)buf & ~PAGE_MASK; + dest_off = (size_t)dest & ~PAGE_MASK; if (!is_dma_copy_aligned(device, buff_off, dest_off, len)) goto err; @@ -1347,9 +1475,6 @@ err: static int ntb_process_tx(struct ntb_transport_qp *qp, struct ntb_queue_entry *entry) { - dev_dbg(&ntb_query_pdev(qp->ndev)->dev, "%lld - tx %u, entry len %d flags %x buff %p\n", - qp->tx_pkts, qp->tx_index, entry->len, entry->flags, - entry->buf); if (qp->tx_index == qp->remote_rx_info->entry) { qp->tx_ring_full++; return -EAGAIN; @@ -1376,15 +1501,14 @@ static int ntb_process_tx(struct ntb_transport_qp *qp, static void ntb_send_link_down(struct ntb_transport_qp *qp) { - struct pci_dev *pdev = ntb_query_pdev(qp->ndev); + struct pci_dev *pdev = qp->ndev->pdev; struct ntb_queue_entry *entry; int i, rc; - if (qp->qp_link == NTB_LINK_DOWN) + if (!qp->link_is_up) return; - qp->qp_link = NTB_LINK_DOWN; - dev_info(&pdev->dev, "qp %d: Link Down\n", qp->qp_num); + dev_info(&pdev->dev, "qp %d: Send Link Down\n", qp->qp_num); for (i = 0; i < NTB_LINK_DOWN_TIMEOUT; i++) { entry = ntb_list_rm(&qp->ntb_tx_free_q_lock, &qp->tx_free_q); @@ -1405,6 +1529,13 @@ static void ntb_send_link_down(struct ntb_transport_qp *qp) if (rc) dev_err(&pdev->dev, "ntb: QP%d unable to send linkdown msg\n", qp->qp_num); + + ntb_qp_link_down_reset(qp); +} + +static bool ntb_dma_filter_fn(struct dma_chan *chan, void *node) +{ + return dev_to_node(&chan->dev->device) == (int)(unsigned long)node; } /** @@ -1422,18 +1553,25 @@ static void ntb_send_link_down(struct ntb_transport_qp *qp) * RETURNS: pointer to newly created ntb_queue, NULL on error. */ struct ntb_transport_qp * -ntb_transport_create_queue(void *data, struct pci_dev *pdev, +ntb_transport_create_queue(void *data, struct device *client_dev, const struct ntb_queue_handlers *handlers) { + struct ntb_dev *ndev; + struct pci_dev *pdev; + struct ntb_transport_ctx *nt; struct ntb_queue_entry *entry; struct ntb_transport_qp *qp; - struct ntb_transport *nt; + u64 qp_bit; unsigned int free_queue; - int rc, i; + dma_cap_mask_t dma_mask; + int node; + int i; - nt = ntb_find_transport(pdev); - if (!nt) - goto err; + ndev = dev_ntb(client_dev->parent); + pdev = ndev->pdev; + nt = ndev->ctx; + + node = dev_to_node(&ndev->dev); free_queue = ffs(nt->qp_bitmap); if (!free_queue) @@ -1442,23 +1580,31 @@ ntb_transport_create_queue(void *data, struct pci_dev *pdev, /* decrement free_queue to make it zero based */ free_queue--; - clear_bit(free_queue, &nt->qp_bitmap); + qp = &nt->qp_vec[free_queue]; + qp_bit = BIT_ULL(qp->qp_num); + + nt->qp_bitmap_free &= ~qp_bit; - qp = &nt->qps[free_queue]; qp->cb_data = data; qp->rx_handler = handlers->rx_handler; qp->tx_handler = handlers->tx_handler; qp->event_handler = handlers->event_handler; - dmaengine_get(); - qp->dma_chan = dma_find_channel(DMA_MEMCPY); - if (!qp->dma_chan) { - dmaengine_put(); - dev_info(&pdev->dev, "Unable to allocate DMA channel, using CPU instead\n"); + dma_cap_zero(dma_mask); + dma_cap_set(DMA_MEMCPY, dma_mask); + + if (use_dma) { + qp->dma_chan = dma_request_channel(dma_mask, ntb_dma_filter_fn, + (void *)(unsigned long)node); + if (!qp->dma_chan) + dev_info(&pdev->dev, "Unable to allocate DMA channel\n"); + } else { + qp->dma_chan = NULL; } + dev_dbg(&pdev->dev, "Using %s memcpy\n", qp->dma_chan ? "DMA" : "CPU"); for (i = 0; i < NTB_QP_DEF_NUM_ENTRIES; i++) { - entry = kzalloc(sizeof(struct ntb_queue_entry), GFP_ATOMIC); + entry = kzalloc_node(sizeof(*entry), GFP_ATOMIC, node); if (!entry) goto err1; @@ -1468,7 +1614,7 @@ ntb_transport_create_queue(void *data, struct pci_dev *pdev, } for (i = 0; i < NTB_QP_DEF_NUM_ENTRIES; i++) { - entry = kzalloc(sizeof(struct ntb_queue_entry), GFP_ATOMIC); + entry = kzalloc_node(sizeof(*entry), GFP_ATOMIC, node); if (!entry) goto err2; @@ -1477,10 +1623,8 @@ ntb_transport_create_queue(void *data, struct pci_dev *pdev, &qp->tx_free_q); } - rc = ntb_register_db_callback(qp->ndev, free_queue, qp, - ntb_transport_rxc_db); - if (rc) - goto err2; + ntb_db_clear(qp->ndev, qp_bit); + ntb_db_clear_mask(qp->ndev, qp_bit); dev_info(&pdev->dev, "NTB Transport QP %d created\n", qp->qp_num); @@ -1493,8 +1637,8 @@ err1: while ((entry = ntb_list_rm(&qp->ntb_rx_free_q_lock, &qp->rx_free_q))) kfree(entry); if (qp->dma_chan) - dmaengine_put(); - set_bit(free_queue, &nt->qp_bitmap); + dma_release_channel(qp->dma_chan); + nt->qp_bitmap_free |= qp_bit; err: return NULL; } @@ -1508,13 +1652,15 @@ EXPORT_SYMBOL_GPL(ntb_transport_create_queue); */ void ntb_transport_free_queue(struct ntb_transport_qp *qp) { + struct ntb_transport_ctx *nt = qp->transport; struct pci_dev *pdev; struct ntb_queue_entry *entry; + u64 qp_bit; if (!qp) return; - pdev = ntb_query_pdev(qp->ndev); + pdev = qp->ndev->pdev; if (qp->dma_chan) { struct dma_chan *chan = qp->dma_chan; @@ -1528,13 +1674,21 @@ void ntb_transport_free_queue(struct ntb_transport_qp *qp) */ dma_sync_wait(chan, qp->last_cookie); dmaengine_terminate_all(chan); - dmaengine_put(); + dma_release_channel(chan); } - ntb_unregister_db_callback(qp->ndev, qp->qp_num); + qp_bit = BIT_ULL(qp->qp_num); + + ntb_db_set_mask(qp->ndev, qp_bit); + tasklet_disable(&qp->rxc_db_work); cancel_delayed_work_sync(&qp->link_work); + qp->cb_data = NULL; + qp->rx_handler = NULL; + qp->tx_handler = NULL; + qp->event_handler = NULL; + while ((entry = ntb_list_rm(&qp->ntb_rx_free_q_lock, &qp->rx_free_q))) kfree(entry); @@ -1546,7 +1700,7 @@ void ntb_transport_free_queue(struct ntb_transport_qp *qp) while ((entry = ntb_list_rm(&qp->ntb_tx_free_q_lock, &qp->tx_free_q))) kfree(entry); - set_bit(qp->qp_num, &qp->transport->qp_bitmap); + nt->qp_bitmap_free |= qp_bit; dev_info(&pdev->dev, "NTB Transport QP %d freed\n", qp->qp_num); } @@ -1567,7 +1721,7 @@ void *ntb_transport_rx_remove(struct ntb_transport_qp *qp, unsigned int *len) struct ntb_queue_entry *entry; void *buf; - if (!qp || qp->client_ready == NTB_LINK_UP) + if (!qp || qp->client_ready) return NULL; entry = ntb_list_rm(&qp->ntb_rx_pend_q_lock, &qp->rx_pend_q); @@ -1636,7 +1790,7 @@ int ntb_transport_tx_enqueue(struct ntb_transport_qp *qp, void *cb, void *data, struct ntb_queue_entry *entry; int rc; - if (!qp || qp->qp_link != NTB_LINK_UP || !len) + if (!qp || !qp->link_is_up || !len) return -EINVAL; entry = ntb_list_rm(&qp->ntb_tx_free_q_lock, &qp->tx_free_q); @@ -1670,9 +1824,9 @@ void ntb_transport_link_up(struct ntb_transport_qp *qp) if (!qp) return; - qp->client_ready = NTB_LINK_UP; + qp->client_ready = true; - if (qp->transport->transport_link == NTB_LINK_UP) + if (qp->transport->link_is_up) schedule_delayed_work(&qp->link_work, 0); } EXPORT_SYMBOL_GPL(ntb_transport_link_up); @@ -1688,27 +1842,20 @@ EXPORT_SYMBOL_GPL(ntb_transport_link_up); void ntb_transport_link_down(struct ntb_transport_qp *qp) { struct pci_dev *pdev; - int rc, val; + int val; if (!qp) return; - pdev = ntb_query_pdev(qp->ndev); - qp->client_ready = NTB_LINK_DOWN; + pdev = qp->ndev->pdev; + qp->client_ready = false; - rc = ntb_read_local_spad(qp->ndev, QP_LINKS, &val); - if (rc) { - dev_err(&pdev->dev, "Error reading spad %d\n", QP_LINKS); - return; - } + val = ntb_spad_read(qp->ndev, QP_LINKS); - rc = ntb_write_remote_spad(qp->ndev, QP_LINKS, - val & ~(1 << qp->qp_num)); - if (rc) - dev_err(&pdev->dev, "Error writing %x to remote spad %d\n", - val & ~(1 << qp->qp_num), QP_LINKS); + ntb_peer_spad_write(qp->ndev, QP_LINKS, + val & ~BIT(qp->qp_num)); - if (qp->qp_link == NTB_LINK_UP) + if (qp->link_is_up) ntb_send_link_down(qp); else cancel_delayed_work_sync(&qp->link_work); @@ -1728,7 +1875,7 @@ bool ntb_transport_link_query(struct ntb_transport_qp *qp) if (!qp) return false; - return qp->qp_link == NTB_LINK_UP; + return qp->link_is_up; } EXPORT_SYMBOL_GPL(ntb_transport_link_query); @@ -1774,3 +1921,71 @@ unsigned int ntb_transport_max_size(struct ntb_transport_qp *qp) return max; } EXPORT_SYMBOL_GPL(ntb_transport_max_size); + +static void ntb_transport_doorbell_callback(void *data, int vector) +{ + struct ntb_transport_ctx *nt = data; + struct ntb_transport_qp *qp; + u64 db_bits; + unsigned int qp_num; + + db_bits = (nt->qp_bitmap & ~nt->qp_bitmap_free & + ntb_db_vector_mask(nt->ndev, vector)); + + while (db_bits) { + qp_num = __ffs(db_bits); + qp = &nt->qp_vec[qp_num]; + + tasklet_schedule(&qp->rxc_db_work); + + db_bits &= ~BIT_ULL(qp_num); + } +} + +static const struct ntb_ctx_ops ntb_transport_ops = { + .link_event = ntb_transport_event_callback, + .db_event = ntb_transport_doorbell_callback, +}; + +static struct ntb_client ntb_transport_client = { + .ops = { + .probe = ntb_transport_probe, + .remove = ntb_transport_free, + }, +}; + +static int __init ntb_transport_init(void) +{ + int rc; + + pr_info("%s, version %s\n", NTB_TRANSPORT_DESC, NTB_TRANSPORT_VER); + + if (debugfs_initialized()) + nt_debugfs_dir = debugfs_create_dir(KBUILD_MODNAME, NULL); + + rc = bus_register(&ntb_transport_bus); + if (rc) + goto err_bus; + + rc = ntb_register_client(&ntb_transport_client); + if (rc) + goto err_client; + + return 0; + +err_client: + bus_unregister(&ntb_transport_bus); +err_bus: + debugfs_remove_recursive(nt_debugfs_dir); + return rc; +} +module_init(ntb_transport_init); + +static void __exit ntb_transport_exit(void) +{ + debugfs_remove_recursive(nt_debugfs_dir); + + ntb_unregister_client(&ntb_transport_client); + bus_unregister(&ntb_transport_bus); +} +module_exit(ntb_transport_exit); |