new file mode 100644
@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_H_
+#define USNIC_H_
+
+#define DRV_NAME "usnic_verbs"
+
+#define PCI_DEVICE_ID_CISCO_VIC_USPACE_NIC 0x00cf /* User space NIC */
+
+#define DRV_VERSION "1.0.2"
+#define DRV_RELDATE "September 09, 2013"
+
+#endif /* USNIC_H_ */
new file mode 100644
@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_CMN_PKT_HDR_H
+#define USNIC_CMN_PKT_HDR_H
+
+#define USNIC_ROCE_ETHERTYPE (0x8915)
+#define USNIC_ROCE_GRH_VER (8)
+#define USNIC_PROTO_VER (1)
+#define USNIC_ROCE_GRH_VER_SHIFT (4)
+
+#endif /* USNIC_COMMON_PKT_HDR_H */
new file mode 100644
@@ -0,0 +1,51 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_CMN_UTIL_H
+#define USNIC_CMN_UTIL_H
+
+static inline void
+usnic_mac_to_gid(const char *const mac, char *raw_gid)
+{
+ raw_gid[0] = 0xfe;
+ raw_gid[1] = 0x80;
+ memset(&raw_gid[2], 0, 6);
+ raw_gid[8] = mac[0]^2;
+ raw_gid[9] = mac[1];
+ raw_gid[10] = mac[2];
+ raw_gid[11] = 0xff;
+ raw_gid[12] = 0xfe;
+ raw_gid[13] = mac[3];
+ raw_gid[14] = mac[4];
+ raw_gid[15] = mac[5];
+}
+
+static inline void
+usnic_write_gid_if_id_from_mac(char *mac, char *raw_gid)
+{
+ raw_gid[8] = mac[0]^2;
+ raw_gid[9] = mac[1];
+ raw_gid[10] = mac[2];
+ raw_gid[11] = 0xff;
+ raw_gid[12] = 0xfe;
+ raw_gid[13] = mac[3];
+ raw_gid[14] = mac[4];
+ raw_gid[15] = mac[5];
+}
+
+#endif /* USNIC_COMMON_UTIL_H */
new file mode 100644
@@ -0,0 +1,71 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#include <linux/debugfs.h>
+#include <linux/module.h>
+
+#include "usnic.h"
+#include "usnic_log.h"
+#include "usnic_debugfs.h"
+
+static struct dentry *debugfs_root;
+
+static ssize_t usnic_debugfs_buildinfo_read(struct file *f, char __user *data,
+ size_t count, loff_t *ppos)
+{
+ char buf[500];
+ int res;
+
+ if (*ppos > 0)
+ return 0;
+
+ res = scnprintf(buf, sizeof(buf),
+ "version: %s\n"
+ "build date: %s\n",
+ DRV_VERSION, DRV_RELDATE);
+
+ return simple_read_from_buffer(data, count, ppos, buf, res);
+}
+
+static const struct file_operations usnic_debugfs_buildinfo_ops = {
+ .owner = THIS_MODULE,
+ .open = simple_open,
+ .read = usnic_debugfs_buildinfo_read
+};
+
+void usnic_debugfs_init(void)
+{
+ debugfs_root = debugfs_create_dir(DRV_NAME, NULL);
+ if (IS_ERR(debugfs_root)) {
+ usnic_err("Failed to create debugfs root dir, check if debugfs is enabled in kernel configuration\n");
+ debugfs_root = NULL;
+ return;
+ }
+
+ debugfs_create_file("build-info", S_IRUGO, debugfs_root,
+ NULL, &usnic_debugfs_buildinfo_ops);
+}
+
+void usnic_debugfs_exit(void)
+{
+ if (!debugfs_root)
+ return;
+
+ debugfs_remove_recursive(debugfs_root);
+ debugfs_root = NULL;
+}
new file mode 100644
@@ -0,0 +1,25 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+#ifndef USNIC_DEBUGFS_H_
+#define USNIC_DEBUGFS_H_
+
+void usnic_debugfs_init(void);
+
+void usnic_debugfs_exit(void);
+
+#endif /*!USNIC_DEBUGFS_H_ */
new file mode 100644
@@ -0,0 +1,243 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+#include <linux/netdevice.h>
+#include <linux/pci.h>
+
+#include "enic_api.h"
+#include "usnic_common_pkt_hdr.h"
+#include "usnic_fwd.h"
+#include "usnic_log.h"
+
+struct usnic_fwd_dev *usnic_fwd_dev_alloc(struct pci_dev *pdev)
+{
+ struct usnic_fwd_dev *ufdev;
+
+ ufdev = kzalloc(sizeof(*ufdev), GFP_KERNEL);
+ if (!ufdev)
+ return NULL;
+
+ ufdev->pdev = pdev;
+ ufdev->netdev = pci_get_drvdata(pdev);
+ spin_lock_init(&ufdev->lock);
+
+ return ufdev;
+}
+
+void usnic_fwd_dev_free(struct usnic_fwd_dev *ufdev)
+{
+ kfree(ufdev);
+}
+
+static int usnic_fwd_devcmd(struct usnic_fwd_dev *ufdev, int vnic_idx,
+ enum vnic_devcmd_cmd cmd, u64 *a0, u64 *a1)
+{
+ int status;
+ struct net_device *netdev = ufdev->netdev;
+
+ spin_lock(&ufdev->lock);
+ status = enic_api_devcmd_proxy_by_index(netdev,
+ vnic_idx,
+ cmd,
+ a0, a1,
+ 1000);
+ spin_unlock(&ufdev->lock);
+ if (status) {
+ if (status == ERR_EINVAL && cmd == CMD_DEL_FILTER) {
+ usnic_dbg("Dev %s vnic idx %u cmd %u already deleted",
+ netdev_name(netdev), vnic_idx, cmd);
+ } else {
+ usnic_err("Dev %s vnic idx %u cmd %u failed with status %d\n",
+ netdev_name(netdev), vnic_idx, cmd,
+ status);
+ }
+ } else {
+ usnic_dbg("Dev %s vnic idx %u cmd %u success",
+ netdev_name(netdev), vnic_idx,
+ cmd);
+ }
+
+ return status;
+}
+
+int usnic_fwd_add_usnic_filter(struct usnic_fwd_dev *ufdev, int vnic_idx,
+ int rq_idx, struct usnic_fwd_filter *fwd_filter,
+ struct usnic_fwd_filter_hndl **filter_hndl)
+{
+ struct filter_tlv *tlv, *tlv_va;
+ struct filter *filter;
+ struct filter_action *action;
+ struct pci_dev *pdev;
+ struct usnic_fwd_filter_hndl *usnic_filter_hndl;
+ int status;
+ u64 a0, a1;
+ u64 tlv_size;
+ dma_addr_t tlv_pa;
+
+ pdev = ufdev->pdev;
+ tlv_size = (2*sizeof(struct filter_tlv) +
+ sizeof(struct filter) +
+ sizeof(struct filter_action));
+ tlv = pci_alloc_consistent(pdev, tlv_size, &tlv_pa);
+ if (!tlv) {
+ usnic_err("Failed to allocate memory\n");
+ return -ENOMEM;
+ }
+
+ usnic_filter_hndl = kzalloc(sizeof(*usnic_filter_hndl), GFP_KERNEL);
+ if (!usnic_filter_hndl) {
+ usnic_err("Failed to allocate memory for hndl\n");
+ pci_free_consistent(pdev, tlv_size, tlv, tlv_pa);
+ return -ENOMEM;
+ }
+
+ tlv_va = tlv;
+ a0 = tlv_pa;
+ a1 = tlv_size;
+ memset(tlv, 0, tlv_size);
+ tlv->type = CLSF_TLV_FILTER;
+ tlv->length = sizeof(struct filter);
+ filter = (struct filter *)&tlv->val;
+ filter->type = FILTER_USNIC_ID;
+ filter->u.usnic.ethtype = USNIC_ROCE_ETHERTYPE;
+ filter->u.usnic.flags = FILTER_FIELD_USNIC_ETHTYPE |
+ FILTER_FIELD_USNIC_ID |
+ FILTER_FIELD_USNIC_PROTO;
+ filter->u.usnic.proto_version = (USNIC_ROCE_GRH_VER <<
+ USNIC_ROCE_GRH_VER_SHIFT)
+ | USNIC_PROTO_VER;
+ filter->u.usnic.usnic_id = fwd_filter->port_num;
+ tlv = (struct filter_tlv *)((char *)tlv + sizeof(struct filter_tlv) +
+ sizeof(struct filter));
+ tlv->type = CLSF_TLV_ACTION;
+ tlv->length = sizeof(struct filter_action);
+ action = (struct filter_action *)&tlv->val;
+ action->type = FILTER_ACTION_RQ_STEERING;
+ action->u.rq_idx = rq_idx;
+
+ status = usnic_fwd_devcmd(ufdev, vnic_idx, CMD_ADD_FILTER, &a0, &a1);
+ pci_free_consistent(pdev, tlv_size, tlv_va, tlv_pa);
+ if (status) {
+ usnic_err("VF %s Filter add failed with status:%d",
+ pci_name(pdev),
+ status);
+ kfree(usnic_filter_hndl);
+ return status;
+ } else {
+ usnic_dbg("VF %s FILTER ID:%u",
+ pci_name(pdev),
+ (u32)a0);
+ }
+
+ usnic_filter_hndl->type = FILTER_USNIC_ID;
+ usnic_filter_hndl->id = (u32)a0;
+ usnic_filter_hndl->vnic_idx = vnic_idx;
+ usnic_filter_hndl->ufdev = ufdev;
+ usnic_filter_hndl->filter = fwd_filter;
+ *filter_hndl = usnic_filter_hndl;
+
+ return status;
+}
+
+int usnic_fwd_del_filter(struct usnic_fwd_filter_hndl *filter_hndl)
+{
+ int status;
+ u64 a0, a1;
+ struct net_device *netdev;
+
+ netdev = filter_hndl->ufdev->netdev;
+ a0 = filter_hndl->id;
+
+ status = usnic_fwd_devcmd(filter_hndl->ufdev, filter_hndl->vnic_idx,
+ CMD_DEL_FILTER, &a0, &a1);
+ if (status) {
+ if (status == ERR_EINVAL) {
+ usnic_dbg("Filter %u already deleted for VF Idx %u pf: %s status: %d",
+ filter_hndl->id, filter_hndl->vnic_idx,
+ netdev_name(netdev), status);
+ status = 0;
+ kfree(filter_hndl);
+ } else {
+ usnic_err("PF %s VF Idx %u Filter: %u FILTER DELETE failed with status %d",
+ netdev_name(netdev),
+ filter_hndl->vnic_idx, filter_hndl->id,
+ status);
+ }
+ } else {
+ usnic_dbg("PF %s VF Idx %u Filter: %u FILTER DELETED",
+ netdev_name(netdev), filter_hndl->vnic_idx,
+ filter_hndl->id);
+ kfree(filter_hndl);
+ }
+
+ return status;
+}
+
+int usnic_fwd_enable_rq(struct usnic_fwd_dev *ufdev, int vnic_idx, int rq_idx)
+{
+ int status;
+ struct net_device *pf_netdev;
+ u64 a0, a1;
+
+ pf_netdev = ufdev->netdev;
+ a0 = rq_idx;
+ a1 = CMD_QP_RQWQ;
+
+ status = usnic_fwd_devcmd(ufdev, vnic_idx, CMD_QP_ENABLE, &a0, &a1);
+
+ if (status) {
+ usnic_err("PF %s VNIC Index %u RQ Index: %u ENABLE Failed with status %d",
+ netdev_name(pf_netdev),
+ vnic_idx,
+ rq_idx,
+ status);
+ } else {
+ usnic_dbg("PF %s VNIC Index %u RQ Index: %u ENABLED",
+ netdev_name(pf_netdev),
+ vnic_idx, rq_idx);
+ }
+
+ return status;
+}
+
+int usnic_fwd_disable_rq(struct usnic_fwd_dev *ufdev, int vnic_idx, int rq_idx)
+{
+ int status;
+ u64 a0, a1;
+ struct net_device *pf_netdev;
+
+ pf_netdev = ufdev->netdev;
+ a0 = rq_idx;
+ a1 = CMD_QP_RQWQ;
+
+ status = usnic_fwd_devcmd(ufdev, vnic_idx, CMD_QP_DISABLE, &a0, &a1);
+
+ if (status) {
+ usnic_err("PF %s VNIC Index %u RQ Index: %u DISABLE Failed with status %d",
+ netdev_name(pf_netdev),
+ vnic_idx,
+ rq_idx,
+ status);
+ } else {
+ usnic_dbg("PF %s VNIC Index %u RQ Index: %u DISABLED",
+ netdev_name(pf_netdev),
+ vnic_idx,
+ rq_idx);
+ }
+
+ return status;
+}
new file mode 100644
@@ -0,0 +1,58 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_FWD_H_
+#define USNIC_FWD_H_
+
+#include <linux/if.h>
+#include <linux/pci.h>
+#include <linux/spinlock.h>
+
+#include "usnic_abi.h"
+#include "vnic_devcmd.h"
+
+struct usnic_fwd_dev {
+ struct pci_dev *pdev;
+ struct net_device *netdev;
+ spinlock_t lock;
+};
+
+struct usnic_fwd_filter {
+ enum usnic_transport_type transport;
+ u16 port_num;
+};
+
+struct usnic_fwd_filter_hndl {
+ enum filter_type type;
+ u32 id;
+ u32 vnic_idx;
+ struct usnic_fwd_dev *ufdev;
+ struct list_head link;
+ struct usnic_fwd_filter *filter;
+};
+
+struct usnic_fwd_dev *usnic_fwd_dev_alloc(struct pci_dev *pdev);
+void usnic_fwd_dev_free(struct usnic_fwd_dev *ufdev);
+int usnic_fwd_add_usnic_filter(struct usnic_fwd_dev *ufdev, int vnic_idx,
+ int rq_idx, struct usnic_fwd_filter *filter,
+ struct usnic_fwd_filter_hndl **filter_hndl);
+int usnic_fwd_del_filter(struct usnic_fwd_filter_hndl *filter_hndl);
+int usnic_fwd_enable_rq(struct usnic_fwd_dev *ufdev, int vnic_idx, int rq_idx);
+int usnic_fwd_disable_rq(struct usnic_fwd_dev *ufdev, int vnic_idx, int rq_idx);
+
+#endif /* !USNIC_FWD_H_ */
new file mode 100644
@@ -0,0 +1,58 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_LOG_H_
+#define USNIC_LOG_H_
+
+#include "usnic.h"
+
+extern unsigned int usnic_log_lvl;
+
+#define USNIC_LOG_LVL_NONE (0)
+#define USNIC_LOG_LVL_ERR (1)
+#define USNIC_LOG_LVL_INFO (2)
+#define USNIC_LOG_LVL_DBG (3)
+
+#define usnic_printk(lvl, args...) \
+ do { \
+ printk(lvl "%s:%s:%d: ", DRV_NAME, __func__, \
+ __LINE__); \
+ printk(args); \
+ } while (0)
+
+#define usnic_dbg(args...) \
+ do { \
+ if (unlikely(usnic_log_lvl >= USNIC_LOG_LVL_DBG)) { \
+ usnic_printk(KERN_INFO, args); \
+ } \
+} while (0)
+
+#define usnic_info(args...) \
+do { \
+ if (usnic_log_lvl >= USNIC_LOG_LVL_INFO) { \
+ usnic_printk(KERN_INFO, args); \
+ } \
+} while (0)
+
+#define usnic_err(args...) \
+ do { \
+ if (usnic_log_lvl >= USNIC_LOG_LVL_ERR) { \
+ usnic_printk(KERN_ERR, args); \
+ } \
+ } while (0)
+#endif /* !USNIC_LOG_H_ */
new file mode 100644
@@ -0,0 +1,125 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+#include <linux/bitmap.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#include "usnic_transport.h"
+#include "usnic_log.h"
+
+/* ROCE */
+static unsigned long *roce_bitmap;
+static u16 roce_next_port = 1;
+#define ROCE_BITMAP_SZ ((1 << (8 /*CHAR_BIT*/ * sizeof(u16)))/8 /*CHAR BIT*/)
+DEFINE_SPINLOCK(roce_bitmap_lock);
+
+static const char *transport_to_str(enum usnic_transport_type type)
+{
+ switch (type) {
+ case USNIC_TRANSPORT_UNKNOWN:
+ return "Unknown";
+ case USNIC_TRANSPORT_ROCE_CUSTOM:
+ return "roce custom";
+ case USNIC_TRANSPORT_MAX:
+ return "Max?";
+ default:
+ return "Not known";
+ }
+}
+
+/*
+ * reserve a port number. if "0" specified, we will try to pick one
+ * starting at roce_next_port. roce_next_port will take on the values
+ * 1..4096
+ */
+u16 usnic_transport_rsrv_port(enum usnic_transport_type type, u16 port_num)
+{
+ if (type == USNIC_TRANSPORT_ROCE_CUSTOM) {
+ spin_lock(&roce_bitmap_lock);
+ if (!port_num) {
+ port_num = bitmap_find_next_zero_area(roce_bitmap,
+ ROCE_BITMAP_SZ,
+ roce_next_port /* start */,
+ 1 /* nr */,
+ 0 /* align */);
+ roce_next_port = (port_num & 4095) + 1;
+ } else if (test_bit(port_num, roce_bitmap)) {
+ usnic_err("Failed to allocate port for %s\n",
+ transport_to_str(type));
+ spin_unlock(&roce_bitmap_lock);
+ goto out_fail;
+ }
+ bitmap_set(roce_bitmap, port_num, 1);
+ spin_unlock(&roce_bitmap_lock);
+ } else {
+ usnic_err("Failed to allocate port - transport %s unsupported\n",
+ transport_to_str(type));
+ goto out_fail;
+ }
+
+ usnic_dbg("Allocating port %hu for %s\n", port_num,
+ transport_to_str(type));
+ return port_num;
+
+out_fail:
+ return 0;
+}
+
+void usnic_transport_unrsrv_port(enum usnic_transport_type type, u16 port_num)
+{
+ if (type == USNIC_TRANSPORT_ROCE_CUSTOM) {
+ spin_lock(&roce_bitmap_lock);
+ if (!port_num) {
+ usnic_err("Unreserved unvalid port num 0 for %s\n",
+ transport_to_str(type));
+ goto out_roce_custom;
+ }
+
+ if (!test_bit(port_num, roce_bitmap)) {
+ usnic_err("Unreserving invalid %hu for %s\n",
+ port_num,
+ transport_to_str(type));
+ goto out_roce_custom;
+ }
+ bitmap_clear(roce_bitmap, port_num, 1);
+ usnic_dbg("Freeing port %hu for %s\n", port_num,
+ transport_to_str(type));
+out_roce_custom:
+ spin_unlock(&roce_bitmap_lock);
+ } else {
+ usnic_err("Freeing invalid port %hu for %d\n", port_num, type);
+ }
+}
+
+int usnic_transport_init()
+{
+ roce_bitmap = kzalloc(ROCE_BITMAP_SZ, GFP_KERNEL);
+ if (!roce_bitmap) {
+ usnic_err("Failed to allocate bit map");
+ return -ENOMEM;
+ }
+
+ /* Do not ever allocate bit 0, hence set it here */
+ bitmap_set(roce_bitmap, 0, 1);
+ return 0;
+}
+
+void usnic_transport_fini(void)
+{
+ kfree(roce_bitmap);
+}
new file mode 100644
@@ -0,0 +1,27 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_TRANSPORT_H_
+#define USNIC_TRANSPORT_H_
+
+#include "usnic_abi.h"
+
+u16 usnic_transport_rsrv_port(enum usnic_transport_type type, u16 port_num);
+void usnic_transport_unrsrv_port(enum usnic_transport_type type, u16 port_num);
+int usnic_transport_init(void);
+#endif /* !USNIC_TRANSPORT_H */
new file mode 100644
@@ -0,0 +1,599 @@
+/*
+ * Copyright (c) 2005 Topspin Communications. All rights reserved.
+ * Copyright (c) 2005 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2013 Cisco Systems. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses. You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ * Redistribution and use in source and binary forms, with or
+ * without modification, are permitted provided that the following
+ * conditions are met:
+ *
+ * - Redistributions of source code must retain the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer.
+ *
+ * - Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following
+ * disclaimer in the documentation and/or other materials
+ * provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/mm.h>
+#include <linux/dma-mapping.h>
+#include <linux/sched.h>
+#include <linux/hugetlb.h>
+#include <linux/dma-attrs.h>
+#include <linux/iommu.h>
+#include <linux/workqueue.h>
+#include <linux/list.h>
+#include <linux/pci.h>
+
+#include "usnic_log.h"
+#include "usnic_uiom.h"
+#include "usnic_uiom_interval_tree.h"
+
+struct workqueue_struct *usnic_uiom_wq;
+
+#define USNIC_UIOM_PAGE_CHUNK \
+ ((PAGE_SIZE - offsetof(struct usnic_uiom_chunk, page_list)) /\
+ ((void *) &((struct usnic_uiom_chunk *) 0)->page_list[1] - \
+ (void *) &((struct usnic_uiom_chunk *) 0)->page_list[0]))
+
+static void usnic_uiom_reg_account(struct work_struct *work)
+{
+ struct usnic_uiom_reg *umem = container_of(work,
+ struct usnic_uiom_reg, work);
+
+ down_write(&umem->mm->mmap_sem);
+ umem->mm->locked_vm -= umem->diff;
+ up_write(&umem->mm->mmap_sem);
+ mmput(umem->mm);
+ kfree(umem);
+}
+
+static int usnic_uiom_dma_fault(struct iommu_domain *domain,
+ struct device *dev,
+ unsigned long iova, int flags,
+ void *token)
+{
+ usnic_err("Device %s iommu fault domain 0x%pK va 0x%lx flags 0x%x\n",
+ dev_name(dev),
+ domain, iova, flags);
+ return -ENOSYS;
+}
+
+static void usnic_uiom_put_pages(struct list_head *chunk_list, int dirty)
+{
+ struct usnic_uiom_chunk *chunk, *tmp;
+ struct page *page;
+ int i;
+
+ list_for_each_entry_safe(chunk, tmp, chunk_list, list) {
+ for (i = 0; i < chunk->nents; i++) {
+ page = sg_page(&chunk->page_list[i]);
+ if (dirty)
+ set_page_dirty_lock(page);
+ put_page(page);
+ usnic_dbg("pa: 0x%llx\n",
+ sg_phys(&chunk->page_list[i]));
+ }
+ kfree(chunk);
+ }
+}
+
+static int usnic_uiom_get_pages(unsigned long addr, size_t size, int writable,
+ int dmasync, struct list_head *chunk_list)
+{
+ struct page **page_list;
+ struct usnic_uiom_chunk *chunk;
+ unsigned long locked;
+ unsigned long lock_limit;
+ unsigned long cur_base;
+ unsigned long npages;
+ int ret;
+ int off;
+ int i;
+ int flags;
+ DEFINE_DMA_ATTRS(attrs);
+
+ if (dmasync)
+ dma_set_attr(DMA_ATTR_WRITE_BARRIER, &attrs);
+
+ if (!can_do_mlock())
+ return -EPERM;
+
+ INIT_LIST_HEAD(chunk_list);
+
+ page_list = (struct page **) __get_free_page(GFP_KERNEL);
+ if (!page_list)
+ return -ENOMEM;
+
+ npages = PAGE_ALIGN(size + (addr & ~PAGE_MASK)) >> PAGE_SHIFT;
+
+ down_write(¤t->mm->mmap_sem);
+
+ locked = npages + current->mm->locked_vm;
+ lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
+
+ if ((locked > lock_limit) && !capable(CAP_IPC_LOCK)) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ flags = IOMMU_READ | IOMMU_CACHE;
+ flags |= (writable) ? IOMMU_WRITE : 0;
+ cur_base = addr & PAGE_MASK;
+ ret = 0;
+
+ while (npages) {
+ ret = get_user_pages(current, current->mm, cur_base,
+ min_t(unsigned long, npages,
+ PAGE_SIZE / sizeof(struct page *)),
+ 1, !writable, page_list, NULL);
+
+ if (ret < 0)
+ goto out;
+
+ npages -= ret;
+ off = 0;
+
+ while (ret) {
+ chunk = kmalloc(sizeof(*chunk) +
+ sizeof(struct scatterlist) *
+ min_t(int, ret, USNIC_UIOM_PAGE_CHUNK),
+ GFP_KERNEL);
+ if (!chunk) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ chunk->nents = min_t(int, ret, USNIC_UIOM_PAGE_CHUNK);
+ sg_init_table(chunk->page_list, chunk->nents);
+ for (i = 0; i < chunk->nents; ++i) {
+ sg_set_page(&chunk->page_list[i],
+ page_list[i + off],
+ PAGE_SIZE, 0);
+ usnic_dbg("va: 0x%lx pa: 0x%llx\n",
+ cur_base + i*PAGE_SIZE,
+ sg_phys(&chunk->page_list[i]));
+ }
+ cur_base += chunk->nents * PAGE_SIZE;
+ ret -= chunk->nents;
+ off += chunk->nents;
+ list_add_tail(&chunk->list, chunk_list);
+ }
+
+ ret = 0;
+ }
+
+out:
+ if (ret < 0)
+ usnic_uiom_put_pages(chunk_list, 0);
+ else
+ current->mm->locked_vm = locked;
+
+ up_write(¤t->mm->mmap_sem);
+ free_page((unsigned long) page_list);
+ return ret;
+}
+
+static void usnic_uiom_unmap_sorted_intervals(struct list_head *intervals,
+ struct usnic_uiom_pd *pd)
+{
+ struct usnic_uiom_interval_node *interval, *tmp;
+ long unsigned va, size;
+
+ list_for_each_entry_safe(interval, tmp, intervals, link) {
+ va = interval->start << PAGE_SHIFT;
+ size = ((interval->last - interval->start) + 1) << PAGE_SHIFT;
+ while (size > 0) {
+ /* Workaround for RH 970401 */
+ usnic_dbg("va 0x%lx size 0x%lx", va, PAGE_SIZE);
+ iommu_unmap(pd->domain, va, PAGE_SIZE);
+ va += PAGE_SIZE;
+ size -= PAGE_SIZE;
+ }
+ }
+}
+
+static void __usnic_uiom_reg_release(struct usnic_uiom_pd *pd,
+ struct usnic_uiom_reg *uiomr,
+ int dirty)
+{
+ int npages;
+ unsigned long vpn_start, vpn_last;
+ struct usnic_uiom_interval_node *interval, *tmp;
+ int writable = 0;
+ LIST_HEAD(rm_intervals);
+
+ npages = PAGE_ALIGN(uiomr->length + uiomr->offset) >> PAGE_SHIFT;
+ vpn_start = (uiomr->va & PAGE_MASK) >> PAGE_SHIFT;
+ vpn_last = vpn_start + npages - 1;
+
+ spin_lock(&pd->lock);
+ usnic_uiom_remove_interval(&pd->rb_root, vpn_start,
+ vpn_last, &rm_intervals);
+ usnic_uiom_unmap_sorted_intervals(&rm_intervals, pd);
+
+ list_for_each_entry_safe(interval, tmp, &rm_intervals, link) {
+ if (interval->flags & IOMMU_WRITE)
+ writable = 1;
+ list_del(&interval->link);
+ kfree(interval);
+ }
+
+ usnic_uiom_put_pages(&uiomr->chunk_list, dirty & writable);
+ spin_unlock(&pd->lock);
+}
+
+static int usnic_uiom_map_sorted_intervals(struct list_head *intervals,
+ struct usnic_uiom_reg *uiomr)
+{
+ int i, err;
+ size_t size;
+ struct usnic_uiom_chunk *chunk;
+ struct usnic_uiom_interval_node *interval_node;
+ long int pa, pa_start, pa_end;
+ long int va_start = -EINVAL;
+ struct usnic_uiom_pd *pd = uiomr->pd;
+ long int va = uiomr->va & PAGE_MASK;
+ int flags = IOMMU_READ | IOMMU_CACHE;
+
+ flags |= (uiomr->writable) ? IOMMU_WRITE : 0;
+ chunk = list_first_entry(&uiomr->chunk_list, struct usnic_uiom_chunk,
+ list);
+ list_for_each_entry(interval_node, intervals, link) {
+iter_chunk:
+ for (i = 0; i < chunk->nents; i++, va += PAGE_SIZE) {
+ pa = sg_phys(&chunk->page_list[i]);
+ if ((va >> PAGE_SHIFT) < interval_node->start)
+ continue;
+
+ if ((va >> PAGE_SHIFT) == interval_node->start) {
+ /* First page of the interval */
+ va_start = va;
+ pa_start = pa;
+ pa_end = pa;
+ }
+
+ BUG_ON(va_start == -EINVAL);
+
+ if ((pa_end + PAGE_SIZE != pa) &&
+ (pa != pa_start)) {
+ /* PAs are not contiguous */
+ size = pa_end - pa_start + PAGE_SIZE;
+ usnic_dbg("va 0x%lx pa 0x%lx size 0x%lx flags 0x%x",
+ va_start, pa_start, size, flags);
+ err = iommu_map(pd->domain, va_start, pa_start,
+ size, flags);
+ if (err) {
+ usnic_err("Failed to map va 0x%lx pa 0x%lx size 0x%zx with err %d\n",
+ va_start, pa_start, size, err);
+ goto err_out;
+ }
+ va_start = va;
+ pa_start = pa;
+ pa_end = pa;
+ }
+
+ if ((va >> PAGE_SHIFT) == interval_node->last) {
+ /* Last page of the interval */
+ size = pa - pa_start + PAGE_SIZE;
+ usnic_dbg("va 0x%lx pa 0x%lx size 0x%lx flags 0x%x\n",
+ va_start, pa_start, size, flags);
+ err = iommu_map(pd->domain, va_start, pa_start,
+ size, flags);
+ if (err) {
+ usnic_err("Failed to map va 0x%lx pa 0x%lx size 0x%zx with err %d\n",
+ va_start, pa_start, size, err);
+ goto err_out;
+ }
+ break;
+ }
+
+ if (pa != pa_start)
+ pa_end += PAGE_SIZE;
+ }
+
+ if (i == chunk->nents) {
+ /*
+ * Hit last entry of the chunk,
+ * hence advance to next chunk
+ */
+ chunk = list_first_entry(&chunk->list,
+ struct usnic_uiom_chunk,
+ list);
+ goto iter_chunk;
+ }
+ }
+
+ return 0;
+
+err_out:
+ usnic_uiom_unmap_sorted_intervals(intervals, pd);
+ return err;
+}
+
+struct usnic_uiom_reg *usnic_uiom_reg_get(struct usnic_uiom_pd *pd,
+ unsigned long addr, size_t size,
+ int writable, int dmasync)
+{
+ struct usnic_uiom_reg *uiomr;
+ unsigned long va_base, vpn_start, vpn_last;
+ unsigned long npages;
+ int offset, err;
+ LIST_HEAD(sorted_diff_intervals);
+
+ /*
+ * Intel IOMMU map throws an error if a translation entry is
+ * changed from read to write. This module may not unmap
+ * and then remap the entry after fixing the permission
+ * b/c this open up a small windows where hw DMA may page fault
+ * Hence, make all entries to be writable.
+ */
+ writable = 1;
+
+ va_base = addr & PAGE_MASK;
+ offset = addr & ~PAGE_MASK;
+ npages = PAGE_ALIGN(size + offset) >> PAGE_SHIFT;
+ vpn_start = (addr & PAGE_MASK) >> PAGE_SHIFT;
+ vpn_last = vpn_start + npages - 1;
+
+ uiomr = kmalloc(sizeof(*uiomr), GFP_KERNEL);
+ if (!uiomr)
+ return ERR_PTR(-ENOMEM);
+
+ uiomr->va = va_base;
+ uiomr->offset = offset;
+ uiomr->length = size;
+ uiomr->writable = writable;
+ uiomr->pd = pd;
+
+ err = usnic_uiom_get_pages(addr, size, writable, dmasync,
+ &uiomr->chunk_list);
+ if (err) {
+ usnic_err("Failed get_pages vpn [0x%lx,0x%lx] err %d\n",
+ vpn_start, vpn_last, err);
+ goto out_free_uiomr;
+ }
+
+ spin_lock(&pd->lock);
+ err = usnic_uiom_get_intervals_diff(vpn_start, vpn_last,
+ (writable) ? IOMMU_WRITE : 0,
+ IOMMU_WRITE,
+ &pd->rb_root,
+ &sorted_diff_intervals);
+ if (err) {
+ usnic_err("Failed disjoint interval vpn [0x%lx,0x%lx] err %d\n",
+ vpn_start, vpn_last, err);
+ goto out_put_pages;
+ }
+
+ err = usnic_uiom_map_sorted_intervals(&sorted_diff_intervals, uiomr);
+ if (err) {
+ usnic_err("Failed map interval vpn [0x%lx,0x%lx] err %d\n",
+ vpn_start, vpn_last, err);
+ goto out_put_intervals;
+
+ }
+
+ err = usnic_uiom_insert_interval(&pd->rb_root, vpn_start, vpn_last,
+ (writable) ? IOMMU_WRITE : 0);
+ if (err) {
+ usnic_err("Failed insert interval vpn [0x%lx,0x%lx] err %d\n",
+ vpn_start, vpn_last, err);
+ goto out_unmap_intervals;
+ }
+
+ usnic_uiom_put_interval_set(&sorted_diff_intervals);
+ spin_unlock(&pd->lock);
+
+ return uiomr;
+
+out_unmap_intervals:
+ usnic_uiom_unmap_sorted_intervals(&sorted_diff_intervals, pd);
+out_put_intervals:
+ usnic_uiom_put_interval_set(&sorted_diff_intervals);
+out_put_pages:
+ usnic_uiom_put_pages(&uiomr->chunk_list, 0);
+ spin_unlock(&pd->lock);
+out_free_uiomr:
+ kfree(uiomr);
+ return ERR_PTR(err);
+}
+
+void usnic_uiom_reg_release(struct usnic_uiom_reg *uiomr, int closing)
+{
+ struct mm_struct *mm;
+ unsigned long diff;
+
+ __usnic_uiom_reg_release(uiomr->pd, uiomr, 1);
+
+ mm = get_task_mm(current);
+ if (!mm) {
+ kfree(uiomr);
+ return;
+ }
+
+ diff = PAGE_ALIGN(uiomr->length + uiomr->offset) >> PAGE_SHIFT;
+
+ /*
+ * We may be called with the mm's mmap_sem already held. This
+ * can happen when a userspace munmap() is the call that drops
+ * the last reference to our file and calls our release
+ * method. If there are memory regions to destroy, we'll end
+ * up here and not be able to take the mmap_sem. In that case
+ * we defer the vm_locked accounting to the system workqueue.
+ */
+ if (closing) {
+ if (!down_write_trylock(&mm->mmap_sem)) {
+ INIT_WORK(&uiomr->work, usnic_uiom_reg_account);
+ uiomr->mm = mm;
+ uiomr->diff = diff;
+
+ queue_work(usnic_uiom_wq, &uiomr->work);
+ return;
+ }
+ } else
+ down_write(&mm->mmap_sem);
+
+ current->mm->locked_vm -= diff;
+ up_write(&mm->mmap_sem);
+ mmput(mm);
+ kfree(uiomr);
+}
+
+struct usnic_uiom_pd *usnic_uiom_alloc_pd(void)
+{
+ struct usnic_uiom_pd *pd;
+ void *domain;
+
+ pd = kzalloc(sizeof(*pd), GFP_KERNEL);
+ if (!pd)
+ return ERR_PTR(-ENOMEM);
+
+ pd->domain = domain = iommu_domain_alloc(&pci_bus_type);
+ if (IS_ERR_OR_NULL(domain)) {
+ usnic_err("Failed to allocate IOMMU domain with err %ld\n",
+ PTR_ERR(pd->domain));
+ kfree(pd);
+ return ERR_PTR(domain ? PTR_ERR(domain) : -ENOMEM);
+ }
+
+ iommu_set_fault_handler(pd->domain, usnic_uiom_dma_fault, NULL);
+
+ spin_lock_init(&pd->lock);
+ INIT_LIST_HEAD(&pd->devs);
+
+ return pd;
+}
+
+void usnic_uiom_dealloc_pd(struct usnic_uiom_pd *pd)
+{
+ iommu_domain_free(pd->domain);
+ kfree(pd);
+}
+
+int usnic_uiom_attach_dev_to_pd(struct usnic_uiom_pd *pd, struct device *dev)
+{
+ struct usnic_uiom_dev *uiom_dev;
+ int err;
+
+ uiom_dev = kzalloc(sizeof(*uiom_dev), GFP_KERNEL);
+ if (!uiom_dev)
+ return -ENOMEM;
+ uiom_dev->dev = dev;
+
+ err = iommu_attach_device(pd->domain, dev);
+ if (err)
+ goto out_free_dev;
+
+ if (!iommu_domain_has_cap(pd->domain, IOMMU_CAP_CACHE_COHERENCY)) {
+ usnic_err("IOMMU of %s does not support cache coherency\n",
+ dev_name(dev));
+ err = -EINVAL;
+ goto out_detach_device;
+ }
+
+ spin_lock(&pd->lock);
+ list_add_tail(&uiom_dev->link, &pd->devs);
+ pd->dev_cnt++;
+ spin_unlock(&pd->lock);
+
+ return 0;
+
+out_detach_device:
+ iommu_detach_device(pd->domain, dev);
+out_free_dev:
+ kfree(uiom_dev);
+ return err;
+}
+
+void usnic_uiom_detach_dev_from_pd(struct usnic_uiom_pd *pd, struct device *dev)
+{
+ struct usnic_uiom_dev *uiom_dev;
+ int found = 0;
+
+ spin_lock(&pd->lock);
+ list_for_each_entry(uiom_dev, &pd->devs, link) {
+ if (uiom_dev->dev == dev) {
+ found = 1;
+ break;
+ }
+ }
+
+ if (!found) {
+ usnic_err("Unable to free dev %s - not found\n",
+ dev_name(dev));
+ spin_unlock(&pd->lock);
+ return;
+ }
+
+ list_del(&uiom_dev->link);
+ pd->dev_cnt--;
+ spin_unlock(&pd->lock);
+
+ return iommu_detach_device(pd->domain, dev);
+}
+
+struct device **usnic_uiom_get_dev_list(struct usnic_uiom_pd *pd)
+{
+ struct usnic_uiom_dev *uiom_dev;
+ struct device **devs;
+ int i = 0;
+
+ spin_lock(&pd->lock);
+ devs = kzalloc(sizeof(*devs)*(pd->dev_cnt + 1), GFP_KERNEL);
+ if (!devs) {
+ devs = ERR_PTR(-ENOMEM);
+ goto out;
+ }
+
+ list_for_each_entry(uiom_dev, &pd->devs, link) {
+ devs[i++] = uiom_dev->dev;
+ }
+out:
+ spin_unlock(&pd->lock);
+ return devs;
+}
+
+void usnic_uiom_free_dev_list(struct device **devs)
+{
+ kfree(devs);
+}
+
+int usnic_uiom_init(char *drv_name)
+{
+ if (!iommu_present(&pci_bus_type)) {
+ usnic_err("IOMMU required but not present or enabled. USNIC QPs will not function w/o enabling IOMMU\n");
+ return -EPERM;
+ }
+
+ usnic_uiom_wq = create_workqueue(drv_name);
+ if (!usnic_uiom_wq) {
+ usnic_err("Unable to alloc wq for drv %s\n", drv_name);
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+void usnic_uiom_fini(void)
+{
+ flush_workqueue(usnic_uiom_wq);
+ destroy_workqueue(usnic_uiom_wq);
+}
new file mode 100644
@@ -0,0 +1,80 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_UIOM_H_
+#define USNIC_UIOM_H_
+
+#include <linux/list.h>
+#include <linux/scatterlist.h>
+
+#include "usnic_uiom_interval_tree.h"
+
+#define USNIC_UIOM_READ (1)
+#define USNIC_UIOM_WRITE (2)
+
+#define USNIC_UIOM_MAX_PD_CNT (1000)
+#define USNIC_UIOM_MAX_MR_CNT (1000000)
+#define USNIC_UIOM_MAX_MR_SIZE (~0UL)
+#define USNIC_UIOM_PAGE_SIZE (PAGE_SIZE)
+
+struct usnic_uiom_dev {
+ struct device *dev;
+ struct list_head link;
+};
+
+struct usnic_uiom_pd {
+ struct iommu_domain *domain;
+ spinlock_t lock;
+ struct rb_root rb_root;
+ struct list_head devs;
+ int dev_cnt;
+};
+
+struct usnic_uiom_reg {
+ struct usnic_uiom_pd *pd;
+ unsigned long va;
+ size_t length;
+ int offset;
+ int page_size;
+ int writable;
+ struct list_head chunk_list;
+ struct work_struct work;
+ struct mm_struct *mm;
+ unsigned long diff;
+};
+
+struct usnic_uiom_chunk {
+ struct list_head list;
+ int nents;
+ struct scatterlist page_list[0];
+};
+
+struct usnic_uiom_pd *usnic_uiom_alloc_pd(void);
+void usnic_uiom_dealloc_pd(struct usnic_uiom_pd *pd);
+int usnic_uiom_attach_dev_to_pd(struct usnic_uiom_pd *pd, struct device *dev);
+void usnic_uiom_detach_dev_from_pd(struct usnic_uiom_pd *pd,
+ struct device *dev);
+struct device **usnic_uiom_get_dev_list(struct usnic_uiom_pd *pd);
+void usnic_uiom_free_dev_list(struct device **devs);
+struct usnic_uiom_reg *usnic_uiom_reg_get(struct usnic_uiom_pd *pd,
+ unsigned long addr, size_t size,
+ int access, int dmasync);
+void usnic_uiom_reg_release(struct usnic_uiom_reg *uiomr, int closing);
+int usnic_uiom_init(char *drv_name);
+void usnic_uiom_fini(void);
+#endif /* USNIC_UIOM_H_ */
new file mode 100644
@@ -0,0 +1,237 @@
+#include <linux/init.h>
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/list_sort.h>
+#include <linux/version.h>
+
+#include <linux/interval_tree_generic.h>
+#include "usnic_uiom_interval_tree.h"
+
+#define START(node) ((node)->start)
+#define LAST(node) ((node)->last)
+
+#define MAKE_NODE(node, start, end, ref_cnt, flags, err, err_out) \
+ do { \
+ node = usnic_uiom_interval_node_alloc(start, \
+ end, ref_cnt, flags); \
+ if (!node) { \
+ err = -ENOMEM; \
+ goto err_out; \
+ } \
+ } while (0)
+
+#define MARK_FOR_ADD(node, list) (list_add_tail(&node->link, list))
+
+#define MAKE_NODE_AND_APPEND(node, start, end, ref_cnt, flags, err, \
+ err_out, list) \
+ do { \
+ MAKE_NODE(node, start, end, \
+ ref_cnt, flags, err, \
+ err_out); \
+ MARK_FOR_ADD(node, list); \
+ } while (0)
+
+#define FLAGS_EQUAL(flags1, flags2, mask) \
+ (((flags1) & (mask)) == ((flags2) & (mask)))
+
+static struct usnic_uiom_interval_node*
+usnic_uiom_interval_node_alloc(long int start, long int last, int ref_cnt,
+ int flags)
+{
+ struct usnic_uiom_interval_node *interval = kzalloc(sizeof(*interval),
+ GFP_KERNEL);
+ if (!interval)
+ return NULL;
+
+ interval->start = start;
+ interval->last = last;
+ interval->flags = flags;
+ interval->ref_cnt = ref_cnt;
+
+ return interval;
+}
+
+static int interval_cmp(void *priv, struct list_head *a, struct list_head *b)
+{
+ struct usnic_uiom_interval_node *node_a, *node_b;
+
+ node_a = list_entry(a, struct usnic_uiom_interval_node, link);
+ node_b = list_entry(b, struct usnic_uiom_interval_node, link);
+
+ /* long to int */
+ if (node_a->start < node_b->start)
+ return -1;
+ else if (node_a->start > node_b->start)
+ return 1;
+
+ return 0;
+}
+
+static void
+find_intervals_intersection_sorted(struct rb_root *root, unsigned long start,
+ unsigned long last,
+ struct list_head *list)
+{
+ struct usnic_uiom_interval_node *node;
+
+ INIT_LIST_HEAD(list);
+
+ for (node = usnic_uiom_interval_tree_iter_first(root, start, last);
+ node;
+ node = usnic_uiom_interval_tree_iter_next(node, start, last))
+ list_add_tail(&node->link, list);
+
+ list_sort(NULL, list, interval_cmp);
+}
+
+int usnic_uiom_get_intervals_diff(unsigned long start, unsigned long last,
+ int flags, int flag_mask,
+ struct rb_root *root,
+ struct list_head *diff_set)
+{
+ struct usnic_uiom_interval_node *interval, *tmp;
+ int err = 0;
+ long int pivot = start;
+ LIST_HEAD(intersection_set);
+
+ INIT_LIST_HEAD(diff_set);
+
+ find_intervals_intersection_sorted(root, start, last,
+ &intersection_set);
+
+ list_for_each_entry(interval, &intersection_set, link) {
+ if (pivot < interval->start) {
+ MAKE_NODE_AND_APPEND(tmp, pivot, interval->start - 1,
+ 1, flags, err, err_out,
+ diff_set);
+ pivot = interval->start;
+ }
+
+ /*
+ * Invariant: Set [start, pivot] is either in diff_set or root,
+ * but not in both.
+ */
+
+ if (pivot > interval->last) {
+ continue;
+ } else if (pivot <= interval->last &&
+ FLAGS_EQUAL(interval->flags, flags,
+ flag_mask)) {
+ pivot = interval->last + 1;
+ }
+ }
+
+ if (pivot <= last)
+ MAKE_NODE_AND_APPEND(tmp, pivot, last, 1, flags, err, err_out,
+ diff_set);
+
+ return 0;
+
+err_out:
+ list_for_each_entry_safe(interval, tmp, diff_set, link) {
+ list_del(&interval->link);
+ kfree(interval);
+ }
+
+ return err;
+}
+
+void usnic_uiom_put_interval_set(struct list_head *intervals)
+{
+ struct usnic_uiom_interval_node *interval, *tmp;
+ list_for_each_entry_safe(interval, tmp, intervals, link)
+ kfree(interval);
+}
+
+int usnic_uiom_insert_interval(struct rb_root *root, unsigned long start,
+ unsigned long last, int flags)
+{
+ struct usnic_uiom_interval_node *interval, *tmp;
+ unsigned long istart, ilast;
+ int iref_cnt, iflags;
+ unsigned long lpivot = start;
+ int err = 0;
+ LIST_HEAD(to_add);
+ LIST_HEAD(intersection_set);
+
+ find_intervals_intersection_sorted(root, start, last,
+ &intersection_set);
+
+ list_for_each_entry(interval, &intersection_set, link) {
+ /*
+ * Invariant - lpivot is the left edge of next interval to be
+ * inserted
+ */
+ istart = interval->start;
+ ilast = interval->last;
+ iref_cnt = interval->ref_cnt;
+ iflags = interval->flags;
+
+ if (istart < lpivot) {
+ MAKE_NODE_AND_APPEND(tmp, istart, lpivot - 1, iref_cnt,
+ iflags, err, err_out, &to_add);
+ } else if (istart > lpivot) {
+ MAKE_NODE_AND_APPEND(tmp, lpivot, istart - 1, 1, flags,
+ err, err_out, &to_add);
+ lpivot = istart;
+ } else {
+ lpivot = istart;
+ }
+
+ if (ilast > last) {
+ MAKE_NODE_AND_APPEND(tmp, lpivot, last, iref_cnt + 1,
+ iflags | flags, err, err_out,
+ &to_add);
+ MAKE_NODE_AND_APPEND(tmp, last + 1, ilast, iref_cnt,
+ iflags, err, err_out, &to_add);
+ } else {
+ MAKE_NODE_AND_APPEND(tmp, lpivot, ilast, iref_cnt + 1,
+ iflags | flags, err, err_out,
+ &to_add);
+ }
+
+ lpivot = ilast + 1;
+ }
+
+ if (lpivot <= last)
+ MAKE_NODE_AND_APPEND(tmp, lpivot, last, 1, flags, err, err_out,
+ &to_add);
+
+ list_for_each_entry_safe(interval, tmp, &intersection_set, link) {
+ usnic_uiom_interval_tree_remove(interval, root);
+ kfree(interval);
+ }
+
+ list_for_each_entry(interval, &to_add, link)
+ usnic_uiom_interval_tree_insert(interval, root);
+
+ return 0;
+
+err_out:
+ list_for_each_entry_safe(interval, tmp, &to_add, link)
+ kfree(interval);
+
+ return err;
+}
+
+void usnic_uiom_remove_interval(struct rb_root *root, unsigned long start,
+ unsigned long last, struct list_head *removed)
+{
+ struct usnic_uiom_interval_node *interval;
+
+ for (interval = usnic_uiom_interval_tree_iter_first(root, start, last);
+ interval;
+ interval = usnic_uiom_interval_tree_iter_next(interval,
+ start,
+ last)) {
+ if (--interval->ref_cnt == 0)
+ list_add_tail(&interval->link, removed);
+ }
+
+ list_for_each_entry(interval, removed, link)
+ usnic_uiom_interval_tree_remove(interval, root);
+}
+
+INTERVAL_TREE_DEFINE(struct usnic_uiom_interval_node, rb,
+ unsigned long, __subtree_last,
+ START, LAST, , usnic_uiom_interval_tree)
new file mode 100644
@@ -0,0 +1,74 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_UIOM_INTERVAL_TREE_H_
+#define USNIC_UIOM_INTERVAL_TREE_H_
+
+#include <linux/version.h>
+#include <linux/rbtree.h>
+
+struct usnic_uiom_interval_node {
+ struct rb_node rb;
+ struct list_head link;
+ unsigned long start;
+ unsigned long last;
+ unsigned long __subtree_last;
+ unsigned int ref_cnt;
+ int flags;
+};
+
+extern void
+usnic_uiom_interval_tree_insert(struct usnic_uiom_interval_node *node,
+ struct rb_root *root);
+extern void
+usnic_uiom_interval_tree_remove(struct usnic_uiom_interval_node *node,
+ struct rb_root *root);
+extern struct usnic_uiom_interval_node *
+usnic_uiom_interval_tree_iter_first(struct rb_root *root,
+ unsigned long start,
+ unsigned long last);
+extern struct usnic_uiom_interval_node *
+usnic_uiom_interval_tree_iter_next(struct usnic_uiom_interval_node *node,
+ unsigned long start, unsigned long last);
+/*
+ * Inserts {start...last} into {root}. If there are overlaps,
+ * nodes will be broken up and merged
+ */
+int usnic_uiom_insert_interval(struct rb_root *root,
+ unsigned long start, unsigned long last,
+ int flags);
+/*
+ * Removed {start...last} from {root}. The nodes removed are returned in
+ * 'removed.' The caller is responsibile for freeing memory of nodes in
+ * 'removed.'
+ */
+void usnic_uiom_remove_interval(struct rb_root *root,
+ unsigned long start, unsigned long last,
+ struct list_head *removed);
+/*
+ * Returns {start...last} - {root} (relative complement of {start...last} in
+ * {root}) in diff_set sorted ascendingly
+ */
+int usnic_uiom_get_intervals_diff(unsigned long start,
+ unsigned long last, int flags,
+ int flag_mask,
+ struct rb_root *root,
+ struct list_head *diff_set);
+/* Call this to free diff_set returned by usnic_uiom_get_intervals_diff */
+void usnic_uiom_put_interval_set(struct list_head *intervals);
+#endif /* USNIC_UIOM_INTERVAL_TREE_H_ */
new file mode 100644
@@ -0,0 +1,473 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+#include <linux/errno.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+
+#include "usnic_ib.h"
+#include "vnic_resource.h"
+#include "usnic_log.h"
+#include "usnic_vnic.h"
+
+struct usnic_vnic {
+ struct vnic_dev *vdev;
+ struct vnic_dev_bar bar[PCI_NUM_RESOURCES];
+ struct usnic_vnic_res_chunk chunks[USNIC_VNIC_RES_TYPE_MAX];
+ spinlock_t res_lock;
+};
+
+static enum vnic_res_type _to_vnic_res_type(enum usnic_vnic_res_type res_type)
+{
+#define DEFINE_USNIC_VNIC_RES_AT(usnic_vnic_res_t, vnic_res_type, desc, val) \
+ vnic_res_type,
+#define DEFINE_USNIC_VNIC_RES(usnic_vnic_res_t, vnic_res_type, desc) \
+ vnic_res_type,
+ static enum vnic_res_type usnic_vnic_type_2_vnic_type[] = {
+ USNIC_VNIC_RES_TYPES};
+#undef DEFINE_USNIC_VNIC_RES
+#undef DEFINE_USNIC_VNIC_RES_AT
+
+ if (res_type >= USNIC_VNIC_RES_TYPE_MAX)
+ return RES_TYPE_MAX;
+
+ return usnic_vnic_type_2_vnic_type[res_type];
+}
+
+const char *usnic_vnic_res_type_to_str(enum usnic_vnic_res_type res_type)
+{
+#define DEFINE_USNIC_VNIC_RES_AT(usnic_vnic_res_t, vnic_res_type, desc, val) \
+ desc,
+#define DEFINE_USNIC_VNIC_RES(usnic_vnic_res_t, vnic_res_type, desc) \
+ desc,
+ static const char * const usnic_vnic_res_type_desc[] = {
+ USNIC_VNIC_RES_TYPES};
+#undef DEFINE_USNIC_VNIC_RES
+#undef DEFINE_USNIC_VNIC_RES_AT
+
+ if (res_type >= USNIC_VNIC_RES_TYPE_MAX)
+ return "unknown";
+
+ return usnic_vnic_res_type_desc[res_type];
+
+}
+
+const char *usnic_vnic_pci_name(struct usnic_vnic *vnic)
+{
+ return pci_name(usnic_vnic_get_pdev(vnic));
+}
+
+int usnic_vnic_dump(struct usnic_vnic *vnic, char *buf,
+ int buf_sz,
+ void *hdr_obj,
+ int (*printtitle)(void *, char*, int),
+ int (*printcols)(char *, int),
+ int (*printrow)(void *, char *, int))
+{
+ struct usnic_vnic_res_chunk *chunk;
+ struct usnic_vnic_res *res;
+ struct vnic_dev_bar *bar0;
+ int i, j, offset;
+
+ offset = 0;
+ bar0 = usnic_vnic_get_bar(vnic, 0);
+ offset += scnprintf(buf + offset, buf_sz - offset,
+ "VF:%hu BAR0 bus_addr=0x%llx vaddr=0x%p size=%ld ",
+ usnic_vnic_get_index(vnic),
+ bar0->bus_addr,
+ bar0->vaddr, bar0->len);
+ if (printtitle)
+ offset += printtitle(hdr_obj, buf + offset, buf_sz - offset);
+ offset += scnprintf(buf + offset, buf_sz - offset, "\n");
+ offset += scnprintf(buf + offset, buf_sz - offset,
+ "|RES\t|CTRL_PIN\t\t|IN_USE\t");
+ if (printcols)
+ offset += printcols(buf + offset, buf_sz - offset);
+ offset += scnprintf(buf + offset, buf_sz - offset, "\n");
+
+ spin_lock(&vnic->res_lock);
+ for (i = 0; i < ARRAY_SIZE(vnic->chunks); i++) {
+ chunk = &vnic->chunks[i];
+ if (chunk->cnt == 0)
+ continue;
+ for (j = 0; j < chunk->cnt; j++) {
+ res = chunk->res[j];
+ offset += scnprintf(buf + offset, buf_sz - offset,
+ "|%s[%u]\t|0x%p\t|%u\t",
+ usnic_vnic_res_type_to_str(res->type),
+ res->vnic_idx, res->ctrl, !!res->owner);
+ if (printrow) {
+ offset += printrow(res->owner, buf + offset,
+ buf_sz - offset);
+ }
+ offset += scnprintf(buf + offset, buf_sz - offset,
+ "\n");
+ }
+ }
+ spin_unlock(&vnic->res_lock);
+ return offset;
+}
+
+void usnic_vnic_res_spec_update(struct usnic_vnic_res_spec *spec,
+ enum usnic_vnic_res_type trgt_type,
+ u16 cnt)
+{
+ int i;
+
+ for (i = 0; i < USNIC_VNIC_RES_TYPE_MAX; i++) {
+ if (spec->resources[i].type == trgt_type) {
+ spec->resources[i].cnt = cnt;
+ return;
+ }
+ }
+
+ WARN_ON(1);
+}
+
+int usnic_vnic_res_spec_satisfied(const struct usnic_vnic_res_spec
+ const *min_spec,
+ const struct usnic_vnic_res_spec
+ const *res_spec)
+{
+ int found, i, j;
+
+ for (i = 0; i < USNIC_VNIC_RES_TYPE_MAX; i++) {
+ found = 0;
+
+ for (j = 0; j < USNIC_VNIC_RES_TYPE_MAX; j++) {
+ if (res_spec->resources[i].type !=
+ min_spec->resources[i].type)
+ continue;
+ found = 1;
+ if (min_spec->resources[i].cnt >
+ res_spec->resources[i].cnt)
+ return -EINVAL;
+ break;
+ }
+
+ if (!found)
+ return -EINVAL;
+ }
+ return 0;
+}
+
+int usnic_vnic_spec_dump(char *buf, int buf_sz, const struct usnic_vnic_res_spec
+ const *res_spec)
+{
+ enum usnic_vnic_res_type res_type;
+ int res_cnt;
+ int i;
+ int offset = 0;
+
+ for (i = 0; i < USNIC_VNIC_RES_TYPE_MAX; i++) {
+ res_type = res_spec->resources[i].type;
+ res_cnt = res_spec->resources[i].cnt;
+ offset += scnprintf(buf + offset, buf_sz - offset,
+ "Res: %s Cnt: %d ",
+ usnic_vnic_res_type_to_str(res_type),
+ res_cnt);
+ }
+
+ offset += scnprintf(buf + offset, buf_sz - offset,
+ "\n");
+ return offset;
+}
+
+int usnic_vnic_check_room(struct usnic_vnic *vnic,
+ const struct usnic_vnic_res_spec const *res_spec)
+{
+ int i;
+ enum usnic_vnic_res_type res_type;
+ int res_cnt;
+
+ for (i = 0; i < USNIC_VNIC_RES_TYPE_MAX; i++) {
+ res_type = res_spec->resources[i].type;
+ res_cnt = res_spec->resources[i].cnt;
+
+ if (res_type == USNIC_VNIC_RES_TYPE_EOL)
+ break;
+
+ if (res_cnt > usnic_vnic_res_free_cnt(vnic, res_type))
+ return -EBUSY;
+ }
+
+ return 0;
+}
+
+int usnic_vnic_res_cnt(struct usnic_vnic *vnic,
+ enum usnic_vnic_res_type type)
+{
+ return vnic->chunks[type].cnt;
+}
+
+int usnic_vnic_res_free_cnt(struct usnic_vnic *vnic,
+ enum usnic_vnic_res_type type)
+{
+ return vnic->chunks[type].free_cnt;
+}
+
+struct usnic_vnic_res_chunk *
+usnic_vnic_get_resources(struct usnic_vnic *vnic, enum usnic_vnic_res_type type,
+ int cnt, void *owner)
+{
+ struct usnic_vnic_res_chunk *src, *ret;
+ struct usnic_vnic_res *res;
+ int i;
+
+ if (usnic_vnic_res_free_cnt(vnic, type) < cnt || cnt < 1 || !owner)
+ return ERR_PTR(-EINVAL);
+
+ ret = kzalloc(sizeof(*ret), GFP_KERNEL);
+ if (!ret) {
+ usnic_err("Failed to allocate chunk for %s - Out of memory\n",
+ usnic_vnic_pci_name(vnic));
+ return ERR_PTR(-ENOMEM);
+ }
+
+ ret->res = kzalloc(sizeof(*(ret->res))*cnt, GFP_KERNEL);
+ if (!ret->res) {
+ usnic_err("Failed to allocate resources for %s. Out of memory\n",
+ usnic_vnic_pci_name(vnic));
+ kfree(ret);
+ return ERR_PTR(-ENOMEM);
+ }
+
+ spin_lock(&vnic->res_lock);
+ src = &vnic->chunks[type];
+ for (i = 0; i < src->cnt && ret->cnt < cnt; i++) {
+ res = src->res[i];
+ if (!res->owner) {
+ src->free_cnt--;
+ res->owner = owner;
+ ret->res[ret->cnt++] = res;
+ }
+ }
+
+ spin_unlock(&vnic->res_lock);
+ ret->type = type;
+ ret->vnic = vnic;
+ WARN_ON(ret->cnt != cnt);
+
+ return ret;
+}
+
+void usnic_vnic_put_resources(struct usnic_vnic_res_chunk *chunk)
+{
+
+ struct usnic_vnic_res *res;
+ int i;
+ struct usnic_vnic *vnic = chunk->vnic;
+
+ spin_lock(&vnic->res_lock);
+ while ((i = --chunk->cnt) >= 0) {
+ res = chunk->res[i];
+ chunk->res[i] = NULL;
+ res->owner = NULL;
+ vnic->chunks[res->type].free_cnt++;
+ }
+ spin_unlock(&vnic->res_lock);
+
+ kfree(chunk->res);
+ kfree(chunk);
+}
+
+u16 usnic_vnic_get_index(struct usnic_vnic *vnic)
+{
+ return usnic_vnic_get_pdev(vnic)->devfn - 1;
+}
+
+static int usnic_vnic_alloc_res_chunk(struct usnic_vnic *vnic,
+ enum usnic_vnic_res_type type,
+ struct usnic_vnic_res_chunk *chunk)
+{
+ int cnt, err, i;
+ struct usnic_vnic_res *res;
+
+ cnt = vnic_dev_get_res_count(vnic->vdev, _to_vnic_res_type(type));
+ if (cnt < 1)
+ return -EINVAL;
+
+ chunk->cnt = chunk->free_cnt = cnt;
+ chunk->res = kzalloc(sizeof(*(chunk->res))*cnt, GFP_KERNEL);
+ if (!chunk->res)
+ return -ENOMEM;
+
+ for (i = 0; i < cnt; i++) {
+ res = kzalloc(sizeof(*res), GFP_KERNEL);
+ if (!res) {
+ err = -ENOMEM;
+ goto fail;
+ }
+ res->type = type;
+ res->vnic_idx = i;
+ res->vnic = vnic;
+ res->ctrl = vnic_dev_get_res(vnic->vdev,
+ _to_vnic_res_type(type), i);
+ chunk->res[i] = res;
+ }
+
+ chunk->vnic = vnic;
+ return 0;
+fail:
+ for (i--; i >= 0; i--)
+ kfree(chunk->res[i]);
+ kfree(chunk->res);
+ return err;
+}
+
+static void usnic_vnic_free_res_chunk(struct usnic_vnic_res_chunk *chunk)
+{
+ int i;
+ for (i = 0; i < chunk->cnt; i++)
+ kfree(chunk->res[i]);
+ kfree(chunk->res);
+}
+
+static int usnic_vnic_discover_resources(struct pci_dev *pdev,
+ struct usnic_vnic *vnic)
+{
+ enum usnic_vnic_res_type res_type;
+ int i;
+ int err = 0;
+
+ for (i = 0; i < ARRAY_SIZE(vnic->bar); i++) {
+ if (!(pci_resource_flags(pdev, i) & IORESOURCE_MEM))
+ continue;
+ vnic->bar[i].len = pci_resource_len(pdev, i);
+ vnic->bar[i].vaddr = pci_iomap(pdev, i, vnic->bar[i].len);
+ if (!vnic->bar[i].vaddr) {
+ usnic_err("Cannot memory-map BAR %d, aborting\n",
+ i);
+ err = -ENODEV;
+ goto out_clean_bar;
+ }
+ vnic->bar[i].bus_addr = pci_resource_start(pdev, i);
+ }
+
+ vnic->vdev = vnic_dev_register(NULL, pdev, pdev, vnic->bar,
+ ARRAY_SIZE(vnic->bar));
+ if (!vnic->vdev) {
+ usnic_err("Failed to register device %s\n",
+ pci_name(pdev));
+ err = -EINVAL;
+ goto out_clean_bar;
+ }
+
+ for (res_type = USNIC_VNIC_RES_TYPE_EOL + 1;
+ res_type < USNIC_VNIC_RES_TYPE_MAX; res_type++) {
+ err = usnic_vnic_alloc_res_chunk(vnic, res_type,
+ &vnic->chunks[res_type]);
+ if (err) {
+ usnic_err("Failed to alloc res %s with err %d\n",
+ usnic_vnic_res_type_to_str(res_type),
+ err);
+ goto out_clean_chunks;
+ }
+ }
+
+ return 0;
+
+out_clean_chunks:
+ for (res_type--; res_type > USNIC_VNIC_RES_TYPE_EOL; res_type--)
+ usnic_vnic_free_res_chunk(&vnic->chunks[res_type]);
+ vnic_dev_unregister(vnic->vdev);
+out_clean_bar:
+ for (i = 0; i < ARRAY_SIZE(vnic->bar); i++) {
+ if (!(pci_resource_flags(pdev, i) & IORESOURCE_MEM))
+ continue;
+ if (!vnic->bar[i].vaddr)
+ break;
+
+ iounmap(vnic->bar[i].vaddr);
+ }
+
+ return err;
+}
+
+struct pci_dev *usnic_vnic_get_pdev(struct usnic_vnic *vnic)
+{
+ return vnic_dev_get_pdev(vnic->vdev);
+}
+
+struct vnic_dev_bar *usnic_vnic_get_bar(struct usnic_vnic *vnic,
+ int bar_num)
+{
+ return (bar_num < ARRAY_SIZE(vnic->bar)) ? &vnic->bar[bar_num] : NULL;
+}
+
+static void usnic_vnic_release_resources(struct usnic_vnic *vnic)
+{
+ int i;
+ struct pci_dev *pdev;
+ enum usnic_vnic_res_type res_type;
+
+ pdev = usnic_vnic_get_pdev(vnic);
+
+ for (res_type = USNIC_VNIC_RES_TYPE_EOL + 1;
+ res_type < USNIC_VNIC_RES_TYPE_MAX; res_type++)
+ usnic_vnic_free_res_chunk(&vnic->chunks[res_type]);
+
+ vnic_dev_unregister(vnic->vdev);
+
+ for (i = 0; i < ARRAY_SIZE(vnic->bar); i++) {
+ if (!(pci_resource_flags(pdev, i) & IORESOURCE_MEM))
+ continue;
+ iounmap(vnic->bar[i].vaddr);
+ }
+}
+
+struct usnic_vnic *usnic_vnic_alloc(struct pci_dev *pdev)
+{
+ struct usnic_vnic *vnic;
+ int err = 0;
+
+ if (!pci_is_enabled(pdev)) {
+ usnic_err("PCI dev %s is disabled\n", pci_name(pdev));
+ return ERR_PTR(-EINVAL);
+ }
+
+ vnic = kzalloc(sizeof(*vnic), GFP_KERNEL);
+ if (!vnic) {
+ usnic_err("Failed to alloc vnic for %s - out of memory\n",
+ pci_name(pdev));
+ return ERR_PTR(-ENOMEM);
+ }
+
+ spin_lock_init(&vnic->res_lock);
+
+ err = usnic_vnic_discover_resources(pdev, vnic);
+ if (err) {
+ usnic_err("Failed to discover %s resources with err %d\n",
+ pci_name(pdev), err);
+ goto out_free_vnic;
+ }
+
+ usnic_dbg("Allocated vnic for %s\n", usnic_vnic_pci_name(vnic));
+
+ return vnic;
+
+out_free_vnic:
+ kfree(vnic);
+
+ return ERR_PTR(err);
+}
+
+void usnic_vnic_free(struct usnic_vnic *vnic)
+{
+ usnic_vnic_release_resources(vnic);
+ kfree(vnic);
+}
new file mode 100644
@@ -0,0 +1,105 @@
+/*
+ * Copyright (c) 2013, Cisco Systems, Inc. All rights reserved.
+ *
+ * This program is free software; you may redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+
+#ifndef USNIC_VNIC_H_
+#define USNIC_VNIC_H_
+
+#include <linux/pci.h>
+
+#include "vnic_dev.h"
+
+/* =USNIC_VNIC_RES_TYPE= =VNIC_RES= =DESC= */
+#define USNIC_VNIC_RES_TYPES \
+ DEFINE_USNIC_VNIC_RES_AT(EOL, RES_TYPE_EOL, "EOL", 0) \
+ DEFINE_USNIC_VNIC_RES(WQ, RES_TYPE_WQ, "WQ") \
+ DEFINE_USNIC_VNIC_RES(RQ, RES_TYPE_RQ, "RQ") \
+ DEFINE_USNIC_VNIC_RES(CQ, RES_TYPE_CQ, "CQ") \
+ DEFINE_USNIC_VNIC_RES(INTR, RES_TYPE_INTR_CTRL, "INT") \
+ DEFINE_USNIC_VNIC_RES(MAX, RES_TYPE_MAX, "MAX")\
+
+#define DEFINE_USNIC_VNIC_RES_AT(usnic_vnic_res_t, vnic_res_type, desc, val) \
+ USNIC_VNIC_RES_TYPE_##usnic_vnic_res_t = val,
+#define DEFINE_USNIC_VNIC_RES(usnic_vnic_res_t, vnic_res_type, desc) \
+ USNIC_VNIC_RES_TYPE_##usnic_vnic_res_t,
+enum usnic_vnic_res_type {
+ USNIC_VNIC_RES_TYPES
+};
+#undef DEFINE_USNIC_VNIC_RES
+#undef DEFINE_USNIC_VNIC_RES_AT
+
+struct usnic_vnic_res {
+ enum usnic_vnic_res_type type;
+ unsigned int vnic_idx;
+ struct usnic_vnic *vnic;
+ void __iomem *ctrl;
+ void *owner;
+};
+
+struct usnic_vnic_res_chunk {
+ enum usnic_vnic_res_type type;
+ int cnt;
+ int free_cnt;
+ struct usnic_vnic_res **res;
+ struct usnic_vnic *vnic;
+};
+
+struct usnic_vnic_res_desc {
+ enum usnic_vnic_res_type type;
+ uint16_t cnt;
+};
+
+struct usnic_vnic_res_spec {
+ struct usnic_vnic_res_desc resources[USNIC_VNIC_RES_TYPE_MAX];
+};
+
+const char *usnic_vnic_res_type_to_str(enum usnic_vnic_res_type res_type);
+const char *usnic_vnic_pci_name(struct usnic_vnic *vnic);
+int usnic_vnic_dump(struct usnic_vnic *vnic, char *buf, int buf_sz,
+ void *hdr_obj,
+ int (*printtitle)(void *, char*, int),
+ int (*printcols)(char *, int),
+ int (*printrow)(void *, char *, int));
+void usnic_vnic_res_spec_update(struct usnic_vnic_res_spec *spec,
+ enum usnic_vnic_res_type trgt_type,
+ u16 cnt);
+int usnic_vnic_res_spec_satisfied(const struct usnic_vnic_res_spec
+ const *min_spec,
+ const struct usnic_vnic_res_spec
+ const *res_spec);
+int usnic_vnic_spec_dump(char *buf, int buf_sz, const struct usnic_vnic_res_spec
+ const *res_spec);
+int usnic_vnic_check_room(struct usnic_vnic *vnic,
+ const struct usnic_vnic_res_spec const *res_spec);
+int usnic_vnic_res_cnt(struct usnic_vnic *vnic,
+ enum usnic_vnic_res_type type);
+int usnic_vnic_res_free_cnt(struct usnic_vnic *vnic,
+ enum usnic_vnic_res_type type);
+struct usnic_vnic_res_chunk *
+usnic_vnic_get_resources(struct usnic_vnic *vnic,
+ enum usnic_vnic_res_type type,
+ int cnt,
+ void *owner);
+void usnic_vnic_put_resources(struct usnic_vnic_res_chunk *chunk);
+struct pci_dev *usnic_vnic_get_pdev(struct usnic_vnic *vnic);
+struct vnic_dev_bar *usnic_vnic_get_bar(struct usnic_vnic *vnic,
+ int bar_num);
+struct usnic_vnic *usnic_vnic_alloc(struct pci_dev *pdev);
+void usnic_vnic_free(struct usnic_vnic *vnic);
+u16 usnic_vnic_get_index(struct usnic_vnic *vnic);
+
+#endif /*!USNIC_VNIC_H_*/