@@ -1,3 +1,3 @@
# SPDX-License-Identifier: GPL-2.0
obj-$(CONFIG_CRYPTO_DEV_YCC) += ycc.o
-ycc-objs := ycc_drv.o ycc_isr.o
+ycc-objs := ycc_drv.o ycc_isr.o ycc_ring.o
@@ -104,10 +104,16 @@ struct ycc_dev {
struct ycc_bar ycc_bars[4];
struct ycc_dev *assoc_dev;
+ int max_desc;
+ int user_rings;
bool is_polling;
unsigned long status;
struct workqueue_struct *dev_err_q;
char err_irq_name[32];
+
+ struct ycc_ring *rings;
+ unsigned long ring_bitmap;
+
struct work_struct work;
char *msi_name[48];
struct dentry *debug_dir;
@@ -24,11 +24,16 @@
#include <linux/debugfs.h>
#include "ycc_isr.h"
+#include "ycc_ring.h"
static const char ycc_name[] = "ycc";
+static int max_desc = 256;
+static int user_rings;
static bool is_polling = true;
+module_param(max_desc, int, 0644);
module_param(is_polling, bool, 0644);
+module_param(user_rings, int, 0644);
LIST_HEAD(ycc_table);
DEFINE_MUTEX(ycc_mutex);
@@ -41,6 +46,35 @@
#define YCC_MAX_DEVICES (98 * 4) /* Assume 4 sockets */
static DEFINE_IDR(ycc_idr);
+static int ycc_dev_debugfs_statistics_show(struct seq_file *s, void *p)
+{
+ struct ycc_dev *ydev = (struct ycc_dev *)s->private;
+ struct ycc_ring *ring;
+ int i;
+
+ seq_printf(s, "name, type, nr_binds, nr_cmds, nr_resps\n");
+ for (i = 0; i < YCC_RINGPAIR_NUM; i++) {
+ ring = ydev->rings + i;
+ seq_printf(s, "Ring%02d, %d, %llu, %llu, %llu\n", ring->ring_id,
+ ring->type, ring->nr_binds, ring->nr_cmds, ring->nr_resps);
+ }
+
+ return 0;
+}
+
+static int ycc_dev_debugfs_statistics_open(struct inode *inode, struct file *filp)
+{
+ return single_open(filp, ycc_dev_debugfs_statistics_show, inode->i_private);
+}
+
+static const struct file_operations ycc_dev_statistics_fops = {
+ .open = ycc_dev_debugfs_statistics_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+ .owner = THIS_MODULE,
+};
+
static int ycc_device_flr(struct pci_dev *pdev, struct pci_dev *rcec_pdev)
{
int ret;
@@ -136,11 +170,21 @@ static int ycc_resource_setup(struct ycc_dev *ydev)
goto iounmap_queue_bar;
}
+ /* User ring is not supported temporarily */
+ ydev->user_rings = 0;
+ user_rings = 0;
+
+ ret = ycc_dev_rings_init(ydev, max_desc, user_rings);
+ if (ret) {
+ pr_err("Failed to init ycc rings\n");
+ goto iounmap_queue_bar;
+ }
+
ret = ycc_enable_msix(ydev);
if (ret <= 0) {
pr_err("Failed to enable msix, ret: %d\n", ret);
ret = (ret == 0) ? -EINVAL : ret;
- goto iounmap_queue_bar;
+ goto release_rings;
}
ret = ycc_init_global_err(ydev);
@@ -163,12 +207,15 @@ static int ycc_resource_setup(struct ycc_dev *ydev)
ycc_deinit_global_err(ydev);
disable_msix:
ycc_disable_msix(ydev);
+release_rings:
+ ycc_dev_rings_release(ydev, user_rings);
iounmap_queue_bar:
iounmap(abar->vaddr);
iounmap_cfg_bar:
iounmap(cfg_bar->vaddr);
release_mem_regions:
pci_release_regions(pdev);
+
return ret;
}
@@ -177,6 +224,7 @@ static void ycc_resource_free(struct ycc_dev *ydev)
ycc_deinit_global_err(ydev);
ycc_free_irqs(ydev);
ycc_disable_msix(ydev);
+ ycc_dev_rings_release(ydev, ydev->user_rings);
iounmap(ydev->ycc_bars[YCC_SEC_CFG_BAR].vaddr);
iounmap(ydev->ycc_bars[YCC_NSEC_Q_BAR].vaddr);
pci_release_regions(ydev->pdev);
@@ -301,12 +349,15 @@ static void ycc_dev_del(struct ycc_dev *ydev)
static inline int ycc_rciep_init(struct ycc_dev *ydev)
{
struct pci_dev *pdev = ydev->pdev;
+ struct dentry *debugfs;
char name[YCC_MAX_DEBUGFS_NAME + 1];
int idr;
ydev->sec = false;
ydev->dev_name = ycc_name;
ydev->is_polling = is_polling;
+ ydev->user_rings = user_rings;
+ ydev->max_desc = max_desc;
idr = idr_alloc(&ycc_idr, ydev, 0, YCC_MAX_DEVICES, GFP_KERNEL);
if (idr < 0) {
@@ -323,6 +374,11 @@ static inline int ycc_rciep_init(struct ycc_dev *ydev)
if (IS_ERR_OR_NULL(ydev->debug_dir)) {
pr_warn("Failed to create debugfs for RCIEP device\n");
ydev->debug_dir = NULL;
+ } else {
+ debugfs = debugfs_create_file("statistics", 0400, ydev->debug_dir,
+ (void *)ydev, &ycc_dev_statistics_fops);
+ if (IS_ERR_OR_NULL(debugfs))
+ pr_warn("Failed to create statistics entry for RCIEP device\n");
}
return 0;
@@ -351,6 +407,7 @@ static int ycc_drv_probe(struct pci_dev *pdev, const struct pci_device_id *id)
ydev->is_vf = false;
ydev->enable_vf = false;
ydev->node = node;
+ ydev->ring_bitmap = 0;
if (id->device == PCI_DEVICE_ID_RCIEP) {
ydev->type = YCC_RCIEP;
ret = ycc_rciep_init(ydev);
@@ -38,6 +38,13 @@ static irqreturn_t ycc_g_err_isr(int irq, void *data)
return IRQ_HANDLED;
}
+/*
+ * TODO: will implement when ycc ring actually work.
+ */
+void ycc_resp_process(uintptr_t ring_addr)
+{
+}
+
int ycc_enable_msix(struct ycc_dev *ydev)
{
struct pci_dev *rcec_pdev = ydev->assoc_dev->pdev;
new file mode 100644
@@ -0,0 +1,562 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#define pr_fmt(fmt) "YCC: Ring: " fmt
+
+#include <linux/kernel.h>
+#include <linux/spinlock.h>
+#include <linux/rbtree.h>
+#include <linux/slab.h>
+#include <linux/pci.h>
+#include <linux/errno.h>
+#include <linux/dma-mapping.h>
+#include <linux/seq_file.h>
+#include <linux/debugfs.h>
+#include <linux/delay.h>
+
+#include "ycc_dev.h"
+#include "ycc_ring.h"
+
+#define YCC_CMD_DESC_SIZE 64
+#define YCC_RESP_DESC_SIZE 16
+#define YCC_RING_CSR_STRIDE 0x1000
+
+extern struct list_head ycc_table;
+extern void ycc_resp_process(uintptr_t ring_addr);
+
+static struct rb_root ring_rbtree = RB_ROOT;
+static DEFINE_SPINLOCK(ring_rbtree_lock);
+
+/*
+ * Show the status of specified ring's command queue and
+ * response queue.
+ */
+static int ycc_ring_debugfs_status_show(struct seq_file *s, void *p)
+{
+ struct ycc_ring *ring = (struct ycc_ring *)s->private;
+
+ seq_printf(s, "Ring ID: %d\n", ring->ring_id);
+ seq_printf(s, "Desscriptor Entry Size: %d, CMD Descriptor Size: %d, RESP Descriptor Size :%d\n",
+ ring->max_desc, YCC_CMD_DESC_SIZE, YCC_RESP_DESC_SIZE);
+ seq_printf(s, "CMD base addr:%llx, RESP base addr:%llx\n",
+ ring->cmd_base_paddr, ring->resp_base_paddr);
+ seq_printf(s, "CMD wr ptr:%d, CMD rd ptr: %d\n",
+ YCC_CSR_RD(ring->csr_vaddr, REG_RING_CMD_WR_PTR),
+ YCC_CSR_RD(ring->csr_vaddr, REG_RING_CMD_RD_PTR));
+ seq_printf(s, "RESP rd ptr:%d, RESP wr ptr: %d\n",
+ YCC_CSR_RD(ring->csr_vaddr, REG_RING_RSP_RD_PTR),
+ YCC_CSR_RD(ring->csr_vaddr, REG_RING_RSP_WR_PTR));
+
+ return 0;
+}
+
+static int ycc_ring_debugfs_status_open(struct inode *inode, struct file *filp)
+{
+ return single_open(filp, ycc_ring_debugfs_status_show, inode->i_private);
+}
+
+static const struct file_operations ycc_ring_status_fops = {
+ .open = ycc_ring_debugfs_status_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+ .owner = THIS_MODULE,
+};
+
+/*
+ * Dump the raw content of specified ring's command queue and
+ * response queue.
+ */
+static int ycc_ring_debugfs_dump_show(struct seq_file *s, void *p)
+{
+ struct ycc_ring *ring = (struct ycc_ring *)s->private;
+
+ seq_printf(s, "Ring ID: %d\n", ring->ring_id);
+ seq_puts(s, "-------- Ring CMD Descriptors --------\n");
+ seq_hex_dump(s, "", DUMP_PREFIX_ADDRESS, 32, 4, ring->cmd_base_vaddr,
+ YCC_CMD_DESC_SIZE * ring->max_desc, false);
+ seq_puts(s, "-------- Ring RESP Descriptors --------\n");
+ seq_hex_dump(s, "", DUMP_PREFIX_ADDRESS, 32, 4, ring->resp_base_vaddr,
+ YCC_RESP_DESC_SIZE * ring->max_desc, false);
+
+ return 0;
+}
+
+static int ycc_ring_debugfs_dump_open(struct inode *inode, struct file *filp)
+{
+ return single_open(filp, ycc_ring_debugfs_dump_show, inode->i_private);
+}
+
+static const struct file_operations ycc_ring_dump_fops = {
+ .open = ycc_ring_debugfs_dump_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = single_release,
+ .owner = THIS_MODULE,
+};
+
+/*
+ * Create debugfs for rings, only for KERN_RING
+ * "/sys/kernel/debugfs/ycc_b:d.f/ring${x}"
+ */
+static int ycc_create_ring_debugfs(struct ycc_ring *ring)
+{
+ struct dentry *debugfs;
+ char name[8];
+
+ if (!ring || !ring->ydev || !ring->ydev->debug_dir)
+ return -EINVAL;
+
+ snprintf(name, sizeof(name), "ring%02d", ring->ring_id);
+ debugfs = debugfs_create_dir(name, ring->ydev->debug_dir);
+ if (IS_ERR_OR_NULL(debugfs))
+ goto out;
+
+ ring->debug_dir = debugfs;
+
+ debugfs = debugfs_create_file("status", 0400, ring->debug_dir,
+ (void *)ring, &ycc_ring_status_fops);
+ if (IS_ERR_OR_NULL(debugfs))
+ goto remove_debugfs;
+
+ debugfs = debugfs_create_file("dump", 0400, ring->debug_dir,
+ (void *)ring, &ycc_ring_dump_fops);
+ if (IS_ERR_OR_NULL(debugfs))
+ goto remove_debugfs;
+
+ return 0;
+
+remove_debugfs:
+ debugfs_remove_recursive(ring->debug_dir);
+out:
+ ring->debug_dir = NULL;
+ return PTR_ERR(debugfs);
+}
+
+static void ycc_remove_ring_debugfs(struct ycc_ring *ring)
+{
+ debugfs_remove_recursive(ring->debug_dir);
+}
+
+/*
+ * Allocate memory for rings and initiate basic fields
+ */
+static int ycc_alloc_rings(struct ycc_dev *ydev)
+{
+ int num = YCC_RINGPAIR_NUM;
+ struct ycc_bar *abar;
+ u32 i;
+
+ if (ydev->rings)
+ return 0;
+
+ if (ydev->is_vf) {
+ num = 1;
+ abar = &ydev->ycc_bars[0];
+ } else if (ydev->sec) {
+ abar = &ydev->ycc_bars[YCC_SEC_Q_BAR];
+ } else {
+ abar = &ydev->ycc_bars[YCC_NSEC_Q_BAR];
+ }
+
+ ydev->rings = kzalloc_node(num * sizeof(struct ycc_ring),
+ GFP_KERNEL, ydev->node);
+ if (!ydev->rings)
+ return -ENOMEM;
+
+ for (i = 0; i < num; i++) {
+ ydev->rings[i].ring_id = i;
+ ydev->rings[i].ydev = ydev;
+ ydev->rings[i].csr_vaddr = abar->vaddr + i * YCC_RING_CSR_STRIDE;
+ ydev->rings[i].csr_paddr = abar->paddr + i * YCC_RING_CSR_STRIDE;
+ }
+
+ return 0;
+}
+
+/*
+ * Free memory for rings
+ */
+static void ycc_free_rings(struct ycc_dev *ydev)
+{
+ kfree(ydev->rings);
+ ydev->rings = NULL;
+}
+
+/*
+ * Initiate ring and create command queue and response queue.
+ */
+static int ycc_init_ring(struct ycc_ring *ring, u32 max_desc)
+{
+ struct ycc_dev *ydev = ring->ydev;
+ u32 cmd_ring_size, resp_ring_size;
+
+ ring->type = KERN_RING;
+ ring->max_desc = max_desc;
+
+ cmd_ring_size = ring->max_desc * YCC_CMD_DESC_SIZE;
+ resp_ring_size = ring->max_desc * YCC_RESP_DESC_SIZE;
+
+ ring->cmd_base_vaddr = dma_alloc_coherent(&ydev->pdev->dev,
+ cmd_ring_size,
+ &ring->cmd_base_paddr,
+ GFP_KERNEL);
+ if (!ring->cmd_base_vaddr) {
+ pr_err("Failed to alloc cmd dma memory\n");
+ return -ENOMEM;
+ }
+ memset(ring->cmd_base_vaddr, CMD_INVALID_CONTENT_U8, cmd_ring_size);
+
+ ring->resp_base_vaddr = dma_alloc_coherent(&ydev->pdev->dev,
+ resp_ring_size,
+ &ring->resp_base_paddr,
+ GFP_KERNEL);
+ if (!ring->resp_base_vaddr) {
+ pr_err("Failed to alloc resp dma memory\n");
+ dma_free_coherent(&ydev->pdev->dev, cmd_ring_size,
+ ring->cmd_base_vaddr, ring->cmd_base_paddr);
+ return -ENOMEM;
+ }
+ memset(ring->resp_base_vaddr, CMD_INVALID_CONTENT_U8, resp_ring_size);
+
+ YCC_CSR_WR(ring->csr_vaddr, REG_RING_RSP_AFULL_TH, 0);
+ YCC_CSR_WR(ring->csr_vaddr, REG_RING_CMD_BASE_ADDR_LO,
+ (u32)ring->cmd_base_paddr & 0xffffffff);
+ YCC_CSR_WR(ring->csr_vaddr, REG_RING_CMD_BASE_ADDR_HI,
+ ((u64)ring->cmd_base_paddr >> 32) & 0xffffffff);
+ YCC_CSR_WR(ring->csr_vaddr, REG_RING_RSP_BASE_ADDR_LO,
+ (u32)ring->resp_base_paddr & 0xffffffff);
+ YCC_CSR_WR(ring->csr_vaddr, REG_RING_RSP_BASE_ADDR_HI,
+ ((u64)ring->resp_base_paddr >> 32) & 0xffffffff);
+
+ if (ycc_create_ring_debugfs(ring))
+ pr_warn("Failed to create debugfs entry for ring: %d\n", ring->ring_id);
+
+ atomic_set(&ring->ref_cnt, 0);
+ spin_lock_init(&ring->lock);
+ return 0;
+}
+
+/*
+ * Release dma memory for command queue and response queue.
+ */
+static void ycc_release_ring(struct ycc_ring *ring)
+{
+ u32 ring_size;
+
+ /* This ring should not be in use here */
+ WARN_ON(atomic_read(&ring->ref_cnt));
+
+ if (ring->cmd_base_vaddr) {
+ ring_size = ring->max_desc * YCC_CMD_DESC_SIZE;
+ dma_free_coherent(&ring->ydev->pdev->dev, ring_size,
+ ring->cmd_base_vaddr,
+ ring->cmd_base_paddr);
+ ring->cmd_base_vaddr = NULL;
+ }
+ if (ring->resp_base_vaddr) {
+ ring_size = ring->max_desc * YCC_RESP_DESC_SIZE;
+ dma_free_coherent(&ring->ydev->pdev->dev, ring_size,
+ ring->resp_base_vaddr,
+ ring->resp_base_paddr);
+ ring->resp_base_vaddr = NULL;
+ }
+
+ ycc_remove_ring_debugfs(ring);
+ ring->type = FREE_RING;
+}
+
+int ycc_dev_rings_init(struct ycc_dev *ydev, u32 max_desc, int user_rings)
+{
+ int kern_rings = YCC_RINGPAIR_NUM - user_rings;
+ struct ycc_ring *ring;
+ int ret, i;
+
+ ret = ycc_alloc_rings(ydev);
+ if (ret) {
+ pr_err("Failed to allocate memory for rings\n");
+ return ret;
+ }
+
+ for (i = 0; i < kern_rings; i++) {
+ ring = &ydev->rings[i];
+ ret = ycc_init_ring(ring, max_desc);
+ if (ret)
+ goto free_kern_rings;
+
+ tasklet_init(&ring->resp_handler, ycc_resp_process, (uintptr_t)ring);
+ }
+
+ return 0;
+
+free_kern_rings:
+ while (i--) {
+ ring = &ydev->rings[i];
+ ycc_release_ring(ring);
+ }
+
+ ycc_free_rings(ydev);
+ return ret;
+}
+
+void ycc_dev_rings_release(struct ycc_dev *ydev, int user_rings)
+{
+ int kern_rings = YCC_RINGPAIR_NUM - user_rings;
+ struct ycc_ring *ring;
+ int i;
+
+ for (i = 0; i < kern_rings; i++) {
+ ring = &ydev->rings[i];
+
+ tasklet_kill(&ring->resp_handler);
+ ycc_release_ring(ring);
+ }
+
+ ycc_free_rings(ydev);
+}
+
+/*
+ * Check if the command queue is full.
+ */
+static inline bool ycc_ring_full(struct ycc_ring *ring)
+{
+ return ring->cmd_rd_ptr == (ring->cmd_wr_ptr + 1) % ring->max_desc;
+}
+
+/*
+ * Check if the response queue is empty
+ */
+static inline bool ycc_ring_empty(struct ycc_ring *ring)
+{
+ return ring->resp_rd_ptr == ring->resp_wr_ptr;
+}
+
+#define __rb_node_to_type(a) rb_entry(a, struct ycc_ring, node)
+
+static inline bool ycc_ring_less(struct rb_node *a, const struct rb_node *b)
+{
+ return (atomic_read(&__rb_node_to_type(a)->ref_cnt)
+ < atomic_read(&__rb_node_to_type(b)->ref_cnt));
+}
+
+static struct ycc_ring *ycc_select_ring(void)
+{
+ struct ycc_ring *found = NULL;
+ struct rb_node *rnode;
+ struct list_head *itr;
+ struct ycc_dev *ydev;
+ int idx;
+
+ if (list_empty(&ycc_table))
+ return NULL;
+
+ /*
+ * No need to protect the list through lock here. The external
+ * ycc_table list only insert/remove entry when probing/removing
+ * the driver. Before calling this function, the reference count
+ * of THIS_MODULE has increased in the crypto layer. So removing
+ * driver will be blocked at present. Meanwhile, inserting new entry
+ * to the list has no impact on 'reading' behavior.
+ */
+ list_for_each(itr, &ycc_table) {
+ ydev = list_entry(itr, struct ycc_dev, list);
+
+ /* RCEC has no rings */
+ if (ydev->type != YCC_RCIEP)
+ continue;
+
+ /* RCIEP is not ready */
+ if (!test_bit(YDEV_STATUS_READY, &ydev->status))
+ continue;
+
+ do {
+ idx = find_first_zero_bit(&ydev->ring_bitmap, YCC_RINGPAIR_NUM);
+ if (idx == YCC_RINGPAIR_NUM)
+ break;
+
+ found = ydev->rings + idx;
+ if (found->type != KERN_RING) {
+ /* Found ring is not for kernel, mark it and continue */
+ set_bit(idx, &ydev->ring_bitmap);
+ continue;
+ }
+ } while (test_and_set_bit(idx, &ydev->ring_bitmap));
+
+ if (idx < YCC_RINGPAIR_NUM && found)
+ goto out;
+ }
+
+ /*
+ * We didn't find the exact ring which means each ring
+ * has been occupied. Fallback to slow path.
+ */
+ spin_lock(&ring_rbtree_lock);
+ rnode = rb_first(&ring_rbtree);
+ rb_erase(rnode, &ring_rbtree);
+ spin_unlock(&ring_rbtree_lock);
+
+ found = __rb_node_to_type(rnode);
+
+out:
+ ycc_ring_get(found);
+ spin_lock(&ring_rbtree_lock);
+ rb_add(&found->node, &ring_rbtree, ycc_ring_less);
+ spin_unlock(&ring_rbtree_lock);
+ return found;
+}
+
+/*
+ * Bind the ring to crypto
+ */
+struct ycc_ring *ycc_crypto_get_ring(void)
+{
+ struct ycc_ring *ring;
+
+ ring = ycc_select_ring();
+ if (ring) {
+ ycc_dev_get(ring->ydev);
+ ring->nr_binds++;
+ if (ring->ydev->is_polling && atomic_read(&ring->ref_cnt) == 1)
+ tasklet_hi_schedule(&ring->resp_handler);
+ }
+
+ return ring;
+}
+
+void ycc_crypto_free_ring(struct ycc_ring *ring)
+{
+ struct rb_node *rnode = &ring->node;
+
+ spin_lock(&ring_rbtree_lock);
+ rb_erase(rnode, &ring_rbtree);
+ if (atomic_dec_and_test(&ring->ref_cnt)) {
+ clear_bit(ring->ring_id, &ring->ydev->ring_bitmap);
+ tasklet_kill(&ring->resp_handler);
+ } else {
+ rb_add(rnode, &ring_rbtree, ycc_ring_less);
+ }
+
+ spin_unlock(&ring_rbtree_lock);
+
+ ycc_dev_put(ring->ydev);
+}
+
+/*
+ * Submit command to ring's command queue.
+ */
+int ycc_enqueue(struct ycc_ring *ring, void *cmd)
+{
+ int ret = 0;
+
+ if (!ring || !ring->ydev || !cmd)
+ return -EINVAL;
+
+ spin_lock_bh(&ring->lock);
+ if (!test_bit(YDEV_STATUS_READY, &ring->ydev->status) || ycc_ring_stopped(ring)) {
+ pr_debug("Enqueue error, device status: %ld, ring stopped: %d\n",
+ ring->ydev->status, ycc_ring_stopped(ring));
+
+ /* Fallback to software */
+ ret = -EAGAIN;
+ goto out;
+ }
+
+ ring->cmd_rd_ptr = YCC_CSR_RD(ring->csr_vaddr, REG_RING_CMD_RD_PTR);
+ if (ycc_ring_full(ring)) {
+ pr_debug("Enqueue error, ring %d is full\n", ring->ring_id);
+ ret = -EAGAIN;
+ goto out;
+ }
+
+ memcpy(ring->cmd_base_vaddr + ring->cmd_wr_ptr * YCC_CMD_DESC_SIZE, cmd,
+ YCC_CMD_DESC_SIZE);
+
+ /* Ensure that cmd_wr_ptr update after memcpy */
+ dma_wmb();
+ if (++ring->cmd_wr_ptr == ring->max_desc)
+ ring->cmd_wr_ptr = 0;
+
+ ring->nr_cmds++;
+ YCC_CSR_WR(ring->csr_vaddr, REG_RING_CMD_WR_PTR, ring->cmd_wr_ptr);
+
+out:
+ spin_unlock_bh(&ring->lock);
+ return ret;
+}
+
+static inline void ycc_check_cmd_state(u16 state)
+{
+ switch (state) {
+ case CMD_SUCCESS:
+ break;
+ case CMD_ILLEGAL:
+ pr_debug("Illegal command\n");
+ break;
+ case CMD_UNDERATTACK:
+ pr_debug("Attack is detected\n");
+ break;
+ case CMD_INVALID:
+ pr_debug("Invalid command\n");
+ break;
+ case CMD_ERROR:
+ pr_debug("Command error\n");
+ break;
+ case CMD_EXCESS:
+ pr_debug("Excess permission\n");
+ break;
+ case CMD_KEY_ERROR:
+ pr_debug("Invalid internal key\n");
+ break;
+ case CMD_VERIFY_ERROR:
+ pr_debug("Mac/tag verify failed\n");
+ break;
+ default:
+ pr_debug("Unknown error\n");
+ break;
+ }
+}
+
+void ycc_handle_resp(struct ycc_ring *ring, struct ycc_resp_desc *desc)
+{
+ struct ycc_flags *aflag;
+
+ dma_rmb();
+
+ aflag = (struct ycc_flags *)desc->private_ptr;
+ if (!aflag || (u64)aflag == CMD_INVALID_CONTENT_U64) {
+ pr_debug("Invalid command aflag\n");
+ return;
+ }
+
+ ycc_check_cmd_state(desc->state);
+ aflag->ycc_done_callback(aflag->ptr, desc->state);
+
+ memset(desc, CMD_INVALID_CONTENT_U8, sizeof(*desc));
+ kfree(aflag);
+}
+
+/*
+ * dequeue, read response descriptor
+ */
+void ycc_dequeue(struct ycc_ring *ring)
+{
+ struct ycc_resp_desc *resp;
+ int cnt = 0;
+
+ if (!test_bit(YDEV_STATUS_READY, &ring->ydev->status) || ycc_ring_stopped(ring))
+ return;
+
+ ring->resp_wr_ptr = YCC_CSR_RD(ring->csr_vaddr, REG_RING_RSP_WR_PTR);
+ while (!ycc_ring_empty(ring)) {
+ resp = (struct ycc_resp_desc *)ring->resp_base_vaddr +
+ ring->resp_rd_ptr;
+ ycc_handle_resp(ring, resp);
+
+ cnt++;
+ ring->nr_resps++;
+ if (++ring->resp_rd_ptr == ring->max_desc)
+ ring->resp_rd_ptr = 0;
+ }
+
+ if (cnt)
+ YCC_CSR_WR(ring->csr_vaddr, REG_RING_RSP_RD_PTR, ring->resp_rd_ptr);
+}
new file mode 100644
@@ -0,0 +1,111 @@
+// SPDX-License-Identifier: GPL-2.0
+#ifndef __YCC_RING_H
+#define __YCC_RING_H
+
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+
+#include "ycc_dev.h"
+
+#define CMD_ILLEGAL 0x15
+#define CMD_UNDERATTACK 0x25
+#define CMD_INVALID 0x35
+#define CMD_ERROR 0x45
+#define CMD_EXCESS 0x55
+#define CMD_KEY_ERROR 0x65
+#define CMD_VERIFY_ERROR 0x85
+#define CMD_SUCCESS 0xa5
+#define CMD_CANCELLED 0xff
+
+#define CMD_INVALID_CONTENT_U8 0x7f
+#define CMD_INVALID_CONTENT_U64 0x7f7f7f7f7f7f7f7fULL
+
+enum ring_type {
+ FREE_RING,
+ USER_RING,
+ KERN_RING,
+ INVAL_RING,
+};
+
+struct ycc_ring {
+ u16 ring_id;
+ u32 status;
+
+ struct rb_node node;
+ atomic_t ref_cnt;
+
+ void __iomem *csr_vaddr; /* config register address */
+ resource_size_t csr_paddr;
+ struct ycc_dev *ydev; /* belongs to which ydev */
+ struct dentry *debug_dir;
+
+ u32 max_desc; /* max desc entry numbers */
+ u32 irq_th;
+ spinlock_t lock; /* used to send cmd, protect write ptr */
+ enum ring_type type;
+
+ void *cmd_base_vaddr; /* base addr of cmd ring */
+ dma_addr_t cmd_base_paddr;
+ u16 cmd_wr_ptr; /* current cmd write pointer */
+ u16 cmd_rd_ptr; /* current cmd read pointer */
+ void *resp_base_vaddr; /* base addr of resp ring */
+ dma_addr_t resp_base_paddr;
+ u16 resp_wr_ptr; /* current resp write pointer */
+ u16 resp_rd_ptr; /* current resp read pointer */
+
+ struct tasklet_struct resp_handler;
+
+ /* for statistics */
+ u64 nr_binds;
+ u64 nr_cmds;
+ u64 nr_resps;
+};
+
+struct ycc_flags {
+ void *ptr;
+ int (*ycc_done_callback)(void *ptr, u16 state);
+};
+
+struct ycc_resp_desc {
+ u64 private_ptr;
+ u16 state;
+ u8 reserved[6];
+};
+
+union ycc_real_cmd {
+ /*
+ * TODO: Real command will implement when
+ * corresponding algorithm is ready
+ */
+ u8 padding[32];
+};
+
+struct ycc_cmd_desc {
+ union ycc_real_cmd cmd;
+ u64 private_ptr;
+ u8 reserved0[16];
+ u8 reserved1[8];
+} __packed;
+
+static inline void ycc_ring_get(struct ycc_ring *ring)
+{
+ atomic_inc(&ring->ref_cnt);
+}
+
+static inline void ycc_ring_put(struct ycc_ring *ring)
+{
+ atomic_dec(&ring->ref_cnt);
+}
+
+static inline bool ycc_ring_stopped(struct ycc_ring *ring)
+{
+ return !!(YCC_CSR_RD(ring->csr_vaddr, REG_RING_CFG) & RING_STOP_BIT);
+}
+
+int ycc_enqueue(struct ycc_ring *ring, void *cmd);
+void ycc_dequeue(struct ycc_ring *ring);
+struct ycc_ring *ycc_crypto_get_ring(void);
+void ycc_crypto_free_ring(struct ycc_ring *ring);
+int ycc_dev_rings_init(struct ycc_dev *ydev, u32 max_desc, int user_rings);
+void ycc_dev_rings_release(struct ycc_dev *ydev, int user_rings);
+#endif