usb: chipidea: udc: manage dynamic amount of tds with a linked list
Instead of having a limited number of usable tds in the udc we use a
linked list to support dynamic amount of needed tds for all special
gadget types. This improves throughput.
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Reviewed-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
diff --git a/drivers/usb/chipidea/udc.c b/drivers/usb/chipidea/udc.c
index 8aed288..960814f 100644
--- a/drivers/usb/chipidea/udc.c
+++ b/drivers/usb/chipidea/udc.c
@@ -368,6 +368,46 @@
/******************************************************************************
* UTIL block
*****************************************************************************/
+
+static void setup_td_bits(struct td_node *tdnode, unsigned length)
+{
+ memset(tdnode->ptr, 0, sizeof(*tdnode->ptr));
+ tdnode->ptr->token = cpu_to_le32(length << __ffs(TD_TOTAL_BYTES));
+ tdnode->ptr->token &= cpu_to_le32(TD_TOTAL_BYTES);
+ tdnode->ptr->token |= cpu_to_le32(TD_STATUS_ACTIVE);
+}
+
+static int add_td_to_list(struct ci13xxx_ep *mEp, struct ci13xxx_req *mReq,
+ unsigned length)
+{
+ struct td_node *lastnode, *node = kzalloc(sizeof(struct td_node),
+ GFP_ATOMIC);
+
+ if (node == NULL)
+ return -ENOMEM;
+
+ node->ptr = dma_pool_alloc(mEp->td_pool, GFP_ATOMIC,
+ &node->dma);
+ if (node->ptr == NULL) {
+ kfree(node);
+ return -ENOMEM;
+ }
+
+ setup_td_bits(node, length);
+
+ if (!list_empty(&mReq->tds)) {
+ /* get the last entry */
+ lastnode = list_entry(mReq->tds.prev,
+ struct td_node, td);
+ lastnode->ptr->next = cpu_to_le32(node->dma);
+ }
+
+ INIT_LIST_HEAD(&node->td);
+ list_add_tail(&node->td, &mReq->tds);
+
+ return 0;
+}
+
/**
* _usb_addr: calculates endpoint address from direction & number
* @ep: endpoint
@@ -390,6 +430,7 @@
unsigned i;
int ret = 0;
unsigned length = mReq->req.length;
+ struct td_node *firstnode, *lastnode;
/* don't queue twice */
if (mReq->req.status == -EALREADY)
@@ -397,58 +438,46 @@
mReq->req.status = -EALREADY;
- if (mReq->req.zero && length && (length % mEp->ep.maxpacket == 0)) {
- mReq->zptr = dma_pool_alloc(mEp->td_pool, GFP_ATOMIC,
- &mReq->zdma);
- if (mReq->zptr == NULL)
- return -ENOMEM;
-
- memset(mReq->zptr, 0, sizeof(*mReq->zptr));
- mReq->zptr->next = cpu_to_le32(TD_TERMINATE);
- mReq->zptr->token = cpu_to_le32(TD_STATUS_ACTIVE);
- if (!mReq->req.no_interrupt)
- mReq->zptr->token |= cpu_to_le32(TD_IOC);
- }
ret = usb_gadget_map_request(&ci->gadget, &mReq->req, mEp->dir);
if (ret)
return ret;
- /*
- * TD configuration
- * TODO - handle requests which spawns into several TDs
- */
- memset(mReq->ptr, 0, sizeof(*mReq->ptr));
- mReq->ptr->token = cpu_to_le32(length << __ffs(TD_TOTAL_BYTES));
- mReq->ptr->token &= cpu_to_le32(TD_TOTAL_BYTES);
- mReq->ptr->token |= cpu_to_le32(TD_STATUS_ACTIVE);
- if (mReq->zptr) {
- mReq->ptr->next = cpu_to_le32(mReq->zdma);
- } else {
- mReq->ptr->next = cpu_to_le32(TD_TERMINATE);
- if (!mReq->req.no_interrupt)
- mReq->ptr->token |= cpu_to_le32(TD_IOC);
- }
- mReq->ptr->page[0] = cpu_to_le32(mReq->req.dma);
+ firstnode = list_first_entry(&mReq->tds,
+ struct td_node, td);
+
+ setup_td_bits(firstnode, length);
+
+ firstnode->ptr->page[0] = cpu_to_le32(mReq->req.dma);
for (i = 1; i < TD_PAGE_COUNT; i++) {
u32 page = mReq->req.dma + i * CI13XXX_PAGE_SIZE;
page &= ~TD_RESERVED_MASK;
- mReq->ptr->page[i] = cpu_to_le32(page);
+ firstnode->ptr->page[i] = cpu_to_le32(page);
}
+ if (mReq->req.zero && length && (length % mEp->ep.maxpacket == 0))
+ add_td_to_list(mEp, mReq, 0);
+
+ lastnode = list_entry(mReq->tds.prev,
+ struct td_node, td);
+
+ lastnode->ptr->next = cpu_to_le32(TD_TERMINATE);
+ if (!mReq->req.no_interrupt)
+ lastnode->ptr->token |= cpu_to_le32(TD_IOC);
wmb();
if (!list_empty(&mEp->qh.queue)) {
struct ci13xxx_req *mReqPrev;
int n = hw_ep_bit(mEp->num, mEp->dir);
int tmp_stat;
- u32 next = mReq->dma & TD_ADDR_MASK;
+ struct td_node *prevlastnode;
+ u32 next = firstnode->dma & TD_ADDR_MASK;
mReqPrev = list_entry(mEp->qh.queue.prev,
struct ci13xxx_req, queue);
- if (mReqPrev->zptr)
- mReqPrev->zptr->next = cpu_to_le32(next);
- else
- mReqPrev->ptr->next = cpu_to_le32(next);
+ prevlastnode = list_entry(mReqPrev->tds.prev,
+ struct td_node, td);
+
+ prevlastnode->ptr->next = cpu_to_le32(next);
wmb();
if (hw_read(ci, OP_ENDPTPRIME, BIT(n)))
goto done;
@@ -462,7 +491,7 @@
}
/* QH configuration */
- mEp->qh.ptr->td.next = cpu_to_le32(mReq->dma); /* TERMINATE = 0 */
+ mEp->qh.ptr->td.next = cpu_to_le32(firstnode->dma);
mEp->qh.ptr->td.token &=
cpu_to_le32(~(TD_STATUS_HALTED|TD_STATUS_ACTIVE));
@@ -491,19 +520,25 @@
*/
static int _hardware_dequeue(struct ci13xxx_ep *mEp, struct ci13xxx_req *mReq)
{
- u32 tmptoken = le32_to_cpu(mReq->ptr->token);
+ u32 tmptoken;
+ struct td_node *node, *tmpnode, *firstnode;
if (mReq->req.status != -EALREADY)
return -EINVAL;
- if ((TD_STATUS_ACTIVE & tmptoken) != 0)
- return -EBUSY;
+ firstnode = list_first_entry(&mReq->tds,
+ struct td_node, td);
- if (mReq->zptr) {
- if ((cpu_to_le32(TD_STATUS_ACTIVE) & mReq->zptr->token) != 0)
+ list_for_each_entry_safe(node, tmpnode, &mReq->tds, td) {
+ tmptoken = le32_to_cpu(node->ptr->token);
+ if ((TD_STATUS_ACTIVE & tmptoken) != 0)
return -EBUSY;
- dma_pool_free(mEp->td_pool, mReq->zptr, mReq->zdma);
- mReq->zptr = NULL;
+ if (node != firstnode) {
+ dma_pool_free(mEp->td_pool, node->ptr, node->dma);
+ list_del_init(&node->td);
+ node->ptr = NULL;
+ kfree(node);
+ }
}
mReq->req.status = 0;
@@ -537,6 +572,7 @@
__releases(mEp->lock)
__acquires(mEp->lock)
{
+ struct td_node *node, *tmpnode, *firstnode;
if (mEp == NULL)
return -EINVAL;
@@ -549,9 +585,17 @@
list_entry(mEp->qh.queue.next,
struct ci13xxx_req, queue);
- if (mReq->zptr) {
- dma_pool_free(mEp->td_pool, mReq->zptr, mReq->zdma);
- mReq->zptr = NULL;
+ firstnode = list_first_entry(&mReq->tds,
+ struct td_node, td);
+
+ list_for_each_entry_safe(node, tmpnode, &mReq->tds, td) {
+ if (node != firstnode) {
+ dma_pool_free(mEp->td_pool, node->ptr,
+ node->dma);
+ list_del_init(&node->td);
+ node->ptr = NULL;
+ kfree(node);
+ }
}
list_del_init(&mReq->queue);
@@ -838,9 +882,13 @@
struct ci13xxx_req *mReq, *mReqTemp;
struct ci13xxx_ep *mEpTemp = mEp;
int retval = 0;
+ struct td_node *firstnode;
list_for_each_entry_safe(mReq, mReqTemp, &mEp->qh.queue,
queue) {
+ firstnode = list_first_entry(&mReq->tds,
+ struct td_node, td);
+
retval = _hardware_dequeue(mEp, mReq);
if (retval < 0)
break;
@@ -1143,19 +1191,26 @@
{
struct ci13xxx_ep *mEp = container_of(ep, struct ci13xxx_ep, ep);
struct ci13xxx_req *mReq = NULL;
+ struct td_node *node;
if (ep == NULL)
return NULL;
mReq = kzalloc(sizeof(struct ci13xxx_req), gfp_flags);
- if (mReq != NULL) {
+ node = kzalloc(sizeof(struct td_node), gfp_flags);
+ if (mReq != NULL && node != NULL) {
INIT_LIST_HEAD(&mReq->queue);
+ INIT_LIST_HEAD(&mReq->tds);
+ INIT_LIST_HEAD(&node->td);
- mReq->ptr = dma_pool_alloc(mEp->td_pool, gfp_flags,
- &mReq->dma);
- if (mReq->ptr == NULL) {
+ node->ptr = dma_pool_alloc(mEp->td_pool, gfp_flags,
+ &node->dma);
+ if (node->ptr == NULL) {
+ kfree(node);
kfree(mReq);
mReq = NULL;
+ } else {
+ list_add_tail(&node->td, &mReq->tds);
}
}
@@ -1171,6 +1226,7 @@
{
struct ci13xxx_ep *mEp = container_of(ep, struct ci13xxx_ep, ep);
struct ci13xxx_req *mReq = container_of(req, struct ci13xxx_req, req);
+ struct td_node *firstnode;
unsigned long flags;
if (ep == NULL || req == NULL) {
@@ -1182,8 +1238,11 @@
spin_lock_irqsave(mEp->lock, flags);
- if (mReq->ptr)
- dma_pool_free(mEp->td_pool, mReq->ptr, mReq->dma);
+ firstnode = list_first_entry(&mReq->tds,
+ struct td_node, td);
+
+ if (firstnode->ptr)
+ dma_pool_free(mEp->td_pool, firstnode->ptr, firstnode->dma);
kfree(mReq);
spin_unlock_irqrestore(mEp->lock, flags);