提交 90664198 编写于 作者: R Ricardo Ribalda Delgado 提交者: Felipe Balbi

usb: gadget: udc: net2280: Remove obsolete module param use_dma_chaining

use_dma_chaining module parameter was designed to avoid creating one irq
per package on a group of packages (with the help of the driver's flag
no_interrupt).

Unfortunately, when this parameter is enabled, the driver fails to work
on both net2280 and 3380 chips.

This patch removes this parameter, which was disabled by default.
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NFelipe Balbi <balbi@ti.com>
上级 48ac1e57
......@@ -12,11 +12,7 @@
* the Mass Storage, Serial, and Ethernet/RNDIS gadget drivers
* as well as Gadget Zero and Gadgetfs.
*
* DMA is enabled by default. Drivers using transfer queues might use
* DMA chaining to remove IRQ latencies between transfers. (Except when
* short OUT transfers happen.) Drivers can use the req->no_interrupt
* hint to completely eliminate some IRQs, if a later IRQ is guaranteed
* and DMA chaining is enabled.
* DMA is enabled by default.
*
* MSI is enabled by default. The legacy IRQ is used if MSI couldn't
* be enabled.
......@@ -85,7 +81,6 @@ static const char *const ep_name[] = {
};
/* use_dma -- general goodness, fewer interrupts, less cpu load (vs PIO)
* use_dma_chaining -- dma descriptor queueing gives even more irq reduction
*
* The net2280 DMA engines are not tightly integrated with their FIFOs;
* not all cases are (yet) handled well in this driver or the silicon.
......@@ -93,12 +88,10 @@ static const char *const ep_name[] = {
* These two parameters let you use PIO or more aggressive DMA.
*/
static bool use_dma = true;
static bool use_dma_chaining;
static bool use_msi = true;
/* "modprobe net2280 use_dma=n" etc */
module_param(use_dma, bool, 0444);
module_param(use_dma_chaining, bool, 0444);
module_param(use_msi, bool, 0444);
/* mode 0 == ep-{a,b,c,d} 1K fifo each
......@@ -202,15 +195,6 @@ net2280_enable(struct usb_ep *_ep, const struct usb_endpoint_descriptor *desc)
/* set speed-dependent max packet; may kick in high bandwidth */
set_max_speed(ep, max);
/* FIFO lines can't go to different packets. PIO is ok, so
* use it instead of troublesome (non-bulk) multi-packet DMA.
*/
if (ep->dma && (max % 4) != 0 && use_dma_chaining) {
ep_dbg(ep->dev, "%s, no dma for maxpacket %d\n",
ep->ep.name, ep->ep.maxpacket);
ep->dma = NULL;
}
/* set type, direction, address; reset fifo counters */
writel(BIT(FIFO_FLUSH), &ep->regs->ep_stat);
tmp = (desc->bmAttributes & USB_ENDPOINT_XFERTYPE_MASK);
......@@ -747,8 +731,7 @@ static void fill_dma_desc(struct net2280_ep *ep,
req->valid = valid;
if (valid)
dmacount |= BIT(VALID_BIT);
if (likely(!req->req.no_interrupt || !use_dma_chaining))
dmacount |= BIT(DMA_DONE_INTERRUPT_ENABLE);
dmacount |= BIT(DMA_DONE_INTERRUPT_ENABLE);
/* td->dmadesc = previously set by caller */
td->dmaaddr = cpu_to_le32 (req->req.dma);
......@@ -862,8 +845,7 @@ static void start_dma(struct net2280_ep *ep, struct net2280_request *req)
req->td->dmadesc = cpu_to_le32 (ep->td_dma);
fill_dma_desc(ep, req, 1);
if (!use_dma_chaining)
req->td->dmacount |= cpu_to_le32(BIT(END_OF_CHAIN));
req->td->dmacount |= cpu_to_le32(BIT(END_OF_CHAIN));
start_queue(ep, tmp, req->td_dma);
}
......@@ -1150,64 +1132,12 @@ static void scan_dma_completions(struct net2280_ep *ep)
static void restart_dma(struct net2280_ep *ep)
{
struct net2280_request *req;
u32 dmactl = dmactl_default;
if (ep->stopped)
return;
req = list_entry(ep->queue.next, struct net2280_request, queue);
if (!use_dma_chaining) {
start_dma(ep, req);
return;
}
/* the 2280 will be processing the queue unless queue hiccups after
* the previous transfer:
* IN: wanted automagic zlp, head doesn't (or vice versa)
* DMA_FIFO_VALIDATE doesn't init from dma descriptors.
* OUT: was "usb-short", we must restart.
*/
if (ep->is_in && !req->valid) {
struct net2280_request *entry, *prev = NULL;
int reqmode, done = 0;
ep_dbg(ep->dev, "%s dma hiccup td %p\n", ep->ep.name, req->td);
ep->in_fifo_validate = likely(req->req.zero ||
(req->req.length % ep->ep.maxpacket) != 0);
if (ep->in_fifo_validate)
dmactl |= BIT(DMA_FIFO_VALIDATE);
list_for_each_entry(entry, &ep->queue, queue) {
__le32 dmacount;
if (entry == req)
continue;
dmacount = entry->td->dmacount;
if (!done) {
reqmode = likely(entry->req.zero ||
(entry->req.length % ep->ep.maxpacket));
if (reqmode == ep->in_fifo_validate) {
entry->valid = 1;
dmacount |= valid_bit;
entry->td->dmacount = dmacount;
prev = entry;
continue;
} else {
/* force a hiccup */
prev->td->dmacount |= dma_done_ie;
done = 1;
}
}
/* walk the rest of the queue so unlinks behave */
entry->valid = 0;
dmacount &= ~valid_bit;
entry->td->dmacount = dmacount;
prev = entry;
}
}
writel(0, &ep->dma->dmactl);
start_queue(ep, dmactl, req->td_dma);
start_dma(ep, req);
}
static void abort_dma_228x(struct net2280_ep *ep)
......@@ -1306,25 +1236,6 @@ static int net2280_dequeue(struct usb_ep *_ep, struct usb_request *_req)
done(ep, req, -ECONNRESET);
}
req = NULL;
/* patch up hardware chaining data */
} else if (ep->dma && use_dma_chaining) {
if (req->queue.prev == ep->queue.next) {
writel(le32_to_cpu(req->td->dmadesc),
&ep->dma->dmadesc);
if (req->td->dmacount & dma_done_ie)
writel(readl(&ep->dma->dmacount) |
le32_to_cpu(dma_done_ie),
&ep->dma->dmacount);
} else {
struct net2280_request *prev;
prev = list_entry(req->queue.prev,
struct net2280_request, queue);
prev->td->dmadesc = req->td->dmadesc;
if (req->td->dmacount & dma_done_ie)
prev->td->dmacount |= dma_done_ie;
}
}
if (req)
......@@ -1609,9 +1520,7 @@ static ssize_t registers_show(struct device *_dev,
"pci irqenb0 %02x irqenb1 %08x "
"irqstat0 %04x irqstat1 %08x\n",
driver_name, dev->chiprev,
use_dma
? (use_dma_chaining ? "chaining" : "enabled")
: "disabled",
use_dma ? "enabled" : "disabled",
readl(&dev->regs->devinit),
readl(&dev->regs->fifoctl),
s,
......@@ -3423,17 +3332,12 @@ static void handle_stat1_irqs(struct net2280 *dev, u32 stat)
continue;
}
/* chaining should stop on abort, short OUT from fifo,
* or (stat0 codepath) short OUT transfer.
*/
if (!use_dma_chaining) {
if (!(tmp & BIT(DMA_TRANSACTION_DONE_INTERRUPT))) {
ep_dbg(ep->dev, "%s no xact done? %08x\n",
ep->ep.name, tmp);
continue;
}
stop_dma(ep->dma);
if (!(tmp & BIT(DMA_TRANSACTION_DONE_INTERRUPT))) {
ep_dbg(ep->dev, "%s no xact done? %08x\n",
ep->ep.name, tmp);
continue;
}
stop_dma(ep->dma);
/* OUT transfers terminate when the data from the
* host is in our memory. Process whatever's done.
......@@ -3448,30 +3352,9 @@ static void handle_stat1_irqs(struct net2280 *dev, u32 stat)
scan_dma_completions(ep);
/* disable dma on inactive queues; else maybe restart */
if (list_empty(&ep->queue)) {
if (use_dma_chaining)
stop_dma(ep->dma);
} else {
if (!list_empty(&ep->queue)) {
tmp = readl(&dma->dmactl);
if (!use_dma_chaining || (tmp & BIT(DMA_ENABLE)) == 0)
restart_dma(ep);
else if (ep->is_in && use_dma_chaining) {
struct net2280_request *req;
__le32 dmacount;
/* the descriptor at the head of the chain
* may still have VALID_BIT clear; that's
* used to trigger changing DMA_FIFO_VALIDATE
* (affects automagic zlp writes).
*/
req = list_entry(ep->queue.next,
struct net2280_request, queue);
dmacount = req->td->dmacount;
dmacount &= cpu_to_le32(BIT(VALID_BIT) |
DMA_BYTE_COUNT_MASK);
if (dmacount && (dmacount & valid_bit) == 0)
restart_dma(ep);
}
restart_dma(ep);
}
ep->irqs++;
}
......@@ -3581,9 +3464,6 @@ static int net2280_probe(struct pci_dev *pdev, const struct pci_device_id *id)
void __iomem *base = NULL;
int retval, i;
if (!use_dma)
use_dma_chaining = 0;
/* alloc, and start init */
dev = kzalloc(sizeof(*dev), GFP_KERNEL);
if (dev == NULL) {
......@@ -3742,8 +3622,7 @@ static int net2280_probe(struct pci_dev *pdev, const struct pci_device_id *id)
ep_info(dev, "irq %d, pci mem %p, chip rev %04x\n",
pdev->irq, base, dev->chiprev);
ep_info(dev, "version: " DRIVER_VERSION "; dma %s %s\n",
use_dma ? (use_dma_chaining ? "chaining" : "enabled")
: "disabled",
use_dma ? "enabled" : "disabled",
dev->enhanced_mode ? "enhanced mode" : "legacy mode");
retval = device_create_file(&pdev->dev, &dev_attr_registers);
if (retval)
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册