7 files changed, 163 insertions, 78 deletions

diff --git a/drivers/usb/musb/Kconfig b/drivers/usb/musb/Kconfig
index 9985db08e7db..b66e8544d8b9 100644
--- a/drivers/usb/musb/Kconfig
+++ b/drivers/usb/musb/Kconfig
@@ -20,8 +20,8 @@ config USB_MUSB_HDRC
           it's being used with, including the USB peripheral role,
           or the USB host role, or both.
 
-          Texas Instruments parts using this IP include DaVinci 644x,
-          OMAP 243x, OMAP 343x, and TUSB 6010.
+          Texas Instruments familiies using this IP include DaVinci
+          (35x, 644x ...), OMAP 243x, OMAP 3, and TUSB 6010.
 
           Analog Devices parts using this IP include Blackfin BF54x,
           BF525 and BF527.
@@ -40,7 +40,7 @@ config USB_MUSB_SOC
         default y if (BF54x && !BF544)
         default y if (BF52x && !BF522 && !BF523)
 
-comment "DaVinci 644x USB support"
+comment "DaVinci 35x and 644x USB support"
         depends on USB_MUSB_HDRC && ARCH_DAVINCI
 
 comment "OMAP 243x high speed USB support"
diff --git a/drivers/usb/musb/davinci.c b/drivers/usb/musb/davinci.c
index 2dc7606f319c..10d11ab113ab 100644
--- a/drivers/usb/musb/davinci.c
+++ b/drivers/usb/musb/davinci.c
@@ -48,6 +48,9 @@
 #include "cppi_dma.h"
 
 
+#define USB_PHY_CTRL    IO_ADDRESS(USBPHY_CTL_PADDR)
+#define DM355_DEEPSLEEP IO_ADDRESS(DM355_DEEPSLEEP_PADDR)
+
 /* REVISIT (PM) we should be able to keep the PHY in low power mode most
  * of the time (24 MHZ oscillator and PLL off, etc) by setting POWER.D0
  * and, when in host mode, autosuspending idle root ports... PHYPLLON
@@ -56,20 +59,26 @@
 
 static inline void phy_on(void)
 {
-        /* start the on-chip PHY and its PLL */
-        __raw_writel(USBPHY_SESNDEN | USBPHY_VBDTCTEN | USBPHY_PHYPLLON,
-                        (void __force __iomem *) IO_ADDRESS(USBPHY_CTL_PADDR));
-        while ((__raw_readl((void __force __iomem *)
-                                IO_ADDRESS(USBPHY_CTL_PADDR))
-                        & USBPHY_PHYCLKGD) == 0)
+        u32     phy_ctrl = __raw_readl(USB_PHY_CTRL);
+
+        /* power everything up; start the on-chip PHY and its PLL */
+        phy_ctrl &= ~(USBPHY_OSCPDWN | USBPHY_OTGPDWN | USBPHY_PHYPDWN);
+        phy_ctrl |= USBPHY_SESNDEN | USBPHY_VBDTCTEN | USBPHY_PHYPLLON;
+        __raw_writel(phy_ctrl, USB_PHY_CTRL);
+
+        /* wait for PLL to lock before proceeding */
+        while ((__raw_readl(USB_PHY_CTRL) & USBPHY_PHYCLKGD) == 0)
                 cpu_relax();
 }
 
 static inline void phy_off(void)
 {
-        /* powerdown the on-chip PHY and its oscillator */
-        __raw_writel(USBPHY_OSCPDWN | USBPHY_PHYPDWN, (void __force __iomem *)
-                        IO_ADDRESS(USBPHY_CTL_PADDR));
+        u32     phy_ctrl = __raw_readl(USB_PHY_CTRL);
+
+        /* powerdown the on-chip PHY, its PLL, and the OTG block */
+        phy_ctrl &= ~(USBPHY_SESNDEN | USBPHY_VBDTCTEN | USBPHY_PHYPLLON);
+        phy_ctrl |= USBPHY_OSCPDWN | USBPHY_OTGPDWN | USBPHY_PHYPDWN;
+        __raw_writel(phy_ctrl, USB_PHY_CTRL);
 }
 
 static int dma_off = 1;
@@ -126,10 +135,6 @@ void musb_platform_disable(struct musb *musb)
 }
 
 
-/* REVISIT it's not clear whether DaVinci can support full OTG.  */
-
-static int vbus_state = -1;
-
 #ifdef CONFIG_USB_MUSB_HDRC_HCD
 #define portstate(stmt)         stmt
 #else
@@ -137,10 +142,19 @@ static int vbus_state = -1;
 #endif
 
 
-/* VBUS SWITCHING IS BOARD-SPECIFIC */
+/*
+ * VBUS SWITCHING IS BOARD-SPECIFIC ... at least for the DM6446 EVM,
+ * which doesn't wire DRVVBUS to the FET that switches it.  Unclear
+ * if that's a problem with the DM6446 chip or just with that board.
+ *
+ * In either case, the DM355 EVM automates DRVVBUS the normal way,
+ * when J10 is out, and TI documents it as handling OTG.
+ */
 
 #ifdef CONFIG_MACH_DAVINCI_EVM
 
+static int vbus_state = -1;
+
 /* I2C operations are always synchronous, and require a task context.
  * With unloaded systems, using the shared workqueue seems to suffice
  * to satisfy the 100msec A_WAIT_VRISE timeout...
@@ -150,12 +164,12 @@ static void evm_deferred_drvvbus(struct work_struct *ignored)
         gpio_set_value_cansleep(GPIO_nVBUS_DRV, vbus_state);
         vbus_state = !vbus_state;
 }
-static DECLARE_WORK(evm_vbus_work, evm_deferred_drvvbus);
 
 #endif  /* EVM */
 
 static void davinci_source_power(struct musb *musb, int is_on, int immediate)
 {
+#ifdef CONFIG_MACH_DAVINCI_EVM
         if (is_on)
                 is_on = 1;
 
@@ -163,16 +177,17 @@ static void davinci_source_power(struct musb *musb, int is_on, int immediate)
                 return;
         vbus_state = !is_on;            /* 0/1 vs "-1 == unknown/init" */
 
-#ifdef CONFIG_MACH_DAVINCI_EVM
         if (machine_is_davinci_evm()) {
+                static DECLARE_WORK(evm_vbus_work, evm_deferred_drvvbus);
+
                 if (immediate)
                         gpio_set_value_cansleep(GPIO_nVBUS_DRV, vbus_state);
                 else
                         schedule_work(&evm_vbus_work);
         }
-#endif
         if (immediate)
                 vbus_state = is_on;
+#endif
 }
 
 static void davinci_set_vbus(struct musb *musb, int is_on)
@@ -391,6 +406,17 @@ int __init musb_platform_init(struct musb *musb)
         musb->board_set_vbus = davinci_set_vbus;
         davinci_source_power(musb, 0, 1);
 
+        /* dm355 EVM swaps D+/D- for signal integrity, and
+         * is clocked from the main 24 MHz crystal.
+         */
+        if (machine_is_davinci_dm355_evm()) {
+                u32     phy_ctrl = __raw_readl(USB_PHY_CTRL);
+
+                phy_ctrl &= ~(3 << 9);
+                phy_ctrl |= USBPHY_DATAPOL;
+                __raw_writel(phy_ctrl, USB_PHY_CTRL);
+        }
+
         /* reset the controller */
         musb_writel(tibase, DAVINCI_USB_CTRL_REG, 0x1);
 
@@ -401,8 +427,7 @@ int __init musb_platform_init(struct musb *musb)
 
         /* NOTE:  irqs are in mixed mode, not bypass to pure-musb */
         pr_debug("DaVinci OTG revision %08x phy %03x control %02x\n",
-                revision, __raw_readl((void __force __iomem *)
-                                IO_ADDRESS(USBPHY_CTL_PADDR)),
+                revision, __raw_readl(USB_PHY_CTRL),
                 musb_readb(tibase, DAVINCI_USB_CTRL_REG));
 
         musb->isr = davinci_interrupt;
diff --git a/drivers/usb/musb/davinci.h b/drivers/usb/musb/davinci.h
index 7fb6238e270f..046c84433cad 100644
--- a/drivers/usb/musb/davinci.h
+++ b/drivers/usb/musb/davinci.h
@@ -15,14 +15,21 @@
  */
 
 /* Integrated highspeed/otg PHY */
-#define USBPHY_CTL_PADDR        (DAVINCI_SYSTEM_MODULE_BASE + 0x34)
-#define USBPHY_PHYCLKGD         (1 << 8)
-#define USBPHY_SESNDEN          (1 << 7)        /* v(sess_end) comparator */
-#define USBPHY_VBDTCTEN         (1 << 6)        /* v(bus) comparator */
-#define USBPHY_PHYPLLON         (1 << 4)        /* override pll suspend */
-#define USBPHY_CLKO1SEL         (1 << 3)
-#define USBPHY_OSCPDWN          (1 << 2)
-#define USBPHY_PHYPDWN          (1 << 0)
+#define USBPHY_CTL_PADDR        (DAVINCI_SYSTEM_MODULE_BASE + 0x34)
+#define USBPHY_DATAPOL          BIT(11) /* (dm355) switch D+/D- */
+#define USBPHY_PHYCLKGD         BIT(8)
+#define USBPHY_SESNDEN          BIT(7)  /* v(sess_end) comparator */
+#define USBPHY_VBDTCTEN         BIT(6)  /* v(bus) comparator */
+#define USBPHY_VBUSSENS         BIT(5)  /* (dm355,ro) is vbus > 0.5V */
+#define USBPHY_PHYPLLON         BIT(4)  /* override pll suspend */
+#define USBPHY_CLKO1SEL         BIT(3)
+#define USBPHY_OSCPDWN          BIT(2)
+#define USBPHY_OTGPDWN          BIT(1)
+#define USBPHY_PHYPDWN          BIT(0)
+
+#define DM355_DEEPSLEEP_PADDR   (DAVINCI_SYSTEM_MODULE_BASE + 0x48)
+#define DRVVBUS_FORCE           BIT(2)
+#define DRVVBUS_OVERRIDE        BIT(1)
 
 /* For now include usb OTG module registers here */
 #define DAVINCI_USB_VERSION_REG         0x00
diff --git a/drivers/usb/musb/musb_core.c b/drivers/usb/musb/musb_core.c
index af77e4659006..338cd1611ab3 100644
--- a/drivers/usb/musb/musb_core.c
+++ b/drivers/usb/musb/musb_core.c
@@ -769,7 +769,7 @@ static irqreturn_t musb_stage2_irq(struct musb *musb, u8 int_usb,
                 case OTG_STATE_A_SUSPEND:
                         usb_hcd_resume_root_hub(musb_to_hcd(musb));
                         musb_root_disconnect(musb);
-                        if (musb->a_wait_bcon != 0)
+                        if (musb->a_wait_bcon != 0 && is_otg_enabled(musb))
                                 musb_platform_try_idle(musb, jiffies
                                         + msecs_to_jiffies(musb->a_wait_bcon));
                         break;
diff --git a/drivers/usb/musb/musb_core.h b/drivers/usb/musb/musb_core.h
index 630946a2d9fc..efb39b5e55b5 100644
--- a/drivers/usb/musb/musb_core.h
+++ b/drivers/usb/musb/musb_core.h
@@ -331,7 +331,6 @@ struct musb {
         struct list_head        control;        /* of musb_qh */
         struct list_head        in_bulk;        /* of musb_qh */
         struct list_head        out_bulk;       /* of musb_qh */
-        struct musb_qh          *periodic[32];  /* tree of interrupt+iso */
 #endif
 
         /* called with IRQs blocked; ON/nonzero implies starting a session,
@@ -479,10 +478,11 @@ static inline void musb_configure_ep0(struct musb *musb)
 static inline int musb_read_fifosize(struct musb *musb,
                 struct musb_hw_ep *hw_ep, u8 epnum)
 {
+        void *mbase = musb->mregs;
         u8 reg = 0;
 
         /* read from core using indexed model */
-        reg = musb_readb(hw_ep->regs, 0x10 + MUSB_FIFOSIZE);
+        reg = musb_readb(mbase, MUSB_EP_OFFSET(epnum, MUSB_FIFOSIZE));
         /* 0's returned when no more endpoints */
         if (!reg)
                 return -ENODEV;
@@ -509,6 +509,7 @@ static inline void musb_configure_ep0(struct musb *musb)
 {
         musb->endpoints[0].max_packet_sz_tx = MUSB_EP0_FIFOSIZE;
         musb->endpoints[0].max_packet_sz_rx = MUSB_EP0_FIFOSIZE;
+        musb->endpoints[0].is_shared_fifo = true;
 }
 #endif /* CONFIG_BLACKFIN */
 
diff --git a/drivers/usb/musb/musb_host.c b/drivers/usb/musb/musb_host.c
index 6dbbd0786a6a..499c431a6d62 100644
--- a/drivers/usb/musb/musb_host.c
+++ b/drivers/usb/musb/musb_host.c
@@ -64,11 +64,8 @@
  *
  * - DMA (Mentor/OMAP) ...has at least toggle update problems
  *
- * - Still no traffic scheduling code to make NAKing for bulk or control
- *   transfers unable to starve other requests; or to make efficient use
- *   of hardware with periodic transfers.  (Note that network drivers
- *   commonly post bulk reads that stay pending for a long time; these
- *   would make very visible trouble.)
+ * - [23-feb-2009] minimal traffic scheduling to avoid bulk RX packet
+ *   starvation ... nothing yet for TX, interrupt, or bulk.
  *
  * - Not tested with HNP, but some SRP paths seem to behave.
  *
@@ -88,11 +85,8 @@
  *
  * CONTROL transfers all go through ep0.  BULK ones go through dedicated IN
  * and OUT endpoints ... hardware is dedicated for those "async" queue(s).
- *
  * (Yes, bulk _could_ use more of the endpoints than that, and would even
- * benefit from it ... one remote device may easily be NAKing while others
- * need to perform transfers in that same direction.  The same thing could
- * be done in software though, assuming dma cooperates.)
+ * benefit from it.)
  *
  * INTERUPPT and ISOCHRONOUS transfers are scheduled to the other endpoints.
  * So far that scheduling is both dumb and optimistic:  the endpoint will be
@@ -201,8 +195,9 @@ musb_start_urb(struct musb *musb, int is_in, struct musb_qh *qh)
                 len = urb->iso_frame_desc[0].length;
                 break;
         default:                /* bulk, interrupt */
-                buf = urb->transfer_buffer;
-                len = urb->transfer_buffer_length;
+                /* actual_length may be nonzero on retry paths */
+                buf = urb->transfer_buffer + urb->actual_length;
+                len = urb->transfer_buffer_length - urb->actual_length;
         }
 
         DBG(4, "qh %p urb %p dev%d ep%d%s%s, hw_ep %d, %p/%d\n",
@@ -395,7 +390,6 @@ musb_giveback(struct musb_qh *qh, struct urb *urb, int status)
                          * de-allocated if it's tracked and allocated;
                          * and where we'd update the schedule tree...
                          */
-                        musb->periodic[ep->epnum] = NULL;
                         kfree(qh);
                         qh = NULL;
                         break;
@@ -1045,7 +1039,8 @@ irqreturn_t musb_h_ep0_irq(struct musb *musb)
 
                 /* NOTE:  this code path would be a good place to PAUSE a
                  * control transfer, if another one is queued, so that
-                 * ep0 is more likely to stay busy.
+                 * ep0 is more likely to stay busy.  That's already done
+                 * for bulk RX transfers.
                  *
                  * if (qh->ring.next != &musb->control), then
                  * we have a candidate... NAKing is *NOT* an error
@@ -1197,6 +1192,7 @@ void musb_host_tx(struct musb *musb, u8 epnum)
                 /* NOTE:  this code path would be a good place to PAUSE a
                  * transfer, if there's some other (nonperiodic) tx urb
                  * that could use this fifo.  (dma complicates it...)
+                 * That's already done for bulk RX transfers.
                  *
                  * if (bulk && qh->ring.next != &musb->out_bulk), then
                  * we have a candidate... NAKing is *NOT* an error
@@ -1358,6 +1354,50 @@ finish:
 
 #endif
 
+/* Schedule next QH from musb->in_bulk and move the current qh to
+ * the end; avoids starvation for other endpoints.
+ */
+static void musb_bulk_rx_nak_timeout(struct musb *musb, struct musb_hw_ep *ep)
+{
+        struct dma_channel      *dma;
+        struct urb              *urb;
+        void __iomem            *mbase = musb->mregs;
+        void __iomem            *epio = ep->regs;
+        struct musb_qh          *cur_qh, *next_qh;
+        u16                     rx_csr;
+
+        musb_ep_select(mbase, ep->epnum);
+        dma = is_dma_capable() ? ep->rx_channel : NULL;
+
+        /* clear nak timeout bit */
+        rx_csr = musb_readw(epio, MUSB_RXCSR);
+        rx_csr |= MUSB_RXCSR_H_WZC_BITS;
+        rx_csr &= ~MUSB_RXCSR_DATAERROR;
+        musb_writew(epio, MUSB_RXCSR, rx_csr);
+
+        cur_qh = first_qh(&musb->in_bulk);
+        if (cur_qh) {
+                urb = next_urb(cur_qh);
+                if (dma_channel_status(dma) == MUSB_DMA_STATUS_BUSY) {
+                        dma->status = MUSB_DMA_STATUS_CORE_ABORT;
+                        musb->dma_controller->channel_abort(dma);
+                        urb->actual_length += dma->actual_len;
+                        dma->actual_len = 0L;
+                }
+                musb_save_toggle(ep, 1, urb);
+
+                /* move cur_qh to end of queue */
+                list_move_tail(&cur_qh->ring, &musb->in_bulk);
+
+                /* get the next qh from musb->in_bulk */
+                next_qh = first_qh(&musb->in_bulk);
+
+                /* set rx_reinit and schedule the next qh */
+                ep->rx_reinit = 1;
+                musb_start_urb(musb, 1, next_qh);
+        }
+}
+
 /*
  * Service an RX interrupt for the given IN endpoint; docs cover bulk, iso,
  * and high-bandwidth IN transfer cases.
@@ -1421,18 +1461,26 @@ void musb_host_rx(struct musb *musb, u8 epnum)
         } else if (rx_csr & MUSB_RXCSR_DATAERROR) {
 
                 if (USB_ENDPOINT_XFER_ISOC != qh->type) {
-                        /* NOTE this code path would be a good place to PAUSE a
-                         * transfer, if there's some other (nonperiodic) rx urb
-                         * that could use this fifo.  (dma complicates it...)
+                        DBG(6, "RX end %d NAK timeout\n", epnum);
+
+                        /* NOTE: NAKing is *NOT* an error, so we want to
+                         * continue.  Except ... if there's a request for
+                         * another QH, use that instead of starving it.
                          *
-                         * if (bulk && qh->ring.next != &musb->in_bulk), then
-                         * we have a candidate... NAKing is *NOT* an error
+                         * Devices like Ethernet and serial adapters keep
+                         * reads posted at all times, which will starve
+                         * other devices without this logic.
                          */
-                        DBG(6, "RX end %d NAK timeout\n", epnum);
+                        if (usb_pipebulk(urb->pipe)
+                                        && qh->mux == 1
+                                        && !list_is_singular(&musb->in_bulk)) {
+                                musb_bulk_rx_nak_timeout(musb, hw_ep);
+                                return;
+                        }
                         musb_ep_select(mbase, epnum);
-                        musb_writew(epio, MUSB_RXCSR,
-                                        MUSB_RXCSR_H_WZC_BITS
-                                        | MUSB_RXCSR_H_REQPKT);
+                        rx_csr |= MUSB_RXCSR_H_WZC_BITS;
+                        rx_csr &= ~MUSB_RXCSR_DATAERROR;
+                        musb_writew(epio, MUSB_RXCSR, rx_csr);
 
                         goto finish;
                 } else {
@@ -1711,31 +1759,27 @@ static int musb_schedule(
 
         /* else, periodic transfers get muxed to other endpoints */
 
-        /* FIXME this doesn't consider direction, so it can only
-         * work for one half of the endpoint hardware, and assumes
-         * the previous cases handled all non-shared endpoints...
-         */
-
-        /* we know this qh hasn't been scheduled, so all we need to do
+        /*
+         * We know this qh hasn't been scheduled, so all we need to do
          * is choose which hardware endpoint to put it on ...
          *
          * REVISIT what we really want here is a regular schedule tree
-         * like e.g. OHCI uses, but for now musb->periodic is just an
-         * array of the _single_ logical endpoint associated with a
-         * given physical one (identity mapping logical->physical).
-         *
-         * that simplistic approach makes TT scheduling a lot simpler;
-         * there is none, and thus none of its complexity...
+         * like e.g. OHCI uses.
          */
         best_diff = 4096;
         best_end = -1;
 
-        for (epnum = 1; epnum < musb->nr_endpoints; epnum++) {
+        for (epnum = 1, hw_ep = musb->endpoints + 1;
+                        epnum < musb->nr_endpoints;
+                        epnum++, hw_ep++) {
                 int     diff;
 
-                if (musb->periodic[epnum])
+                if (is_in || hw_ep->is_shared_fifo) {
+                        if (hw_ep->in_qh  != NULL)
+                                continue;
+                } else  if (hw_ep->out_qh != NULL)
                         continue;
-                hw_ep = &musb->endpoints[epnum];
+
                 if (hw_ep == musb->bulk_ep)
                         continue;
 
@@ -1756,6 +1800,17 @@ static int musb_schedule(
                         head = &musb->in_bulk;
                 else
                         head = &musb->out_bulk;
+
+                /* Enable bulk RX NAK timeout scheme when bulk requests are
+                 * multiplexed.  This scheme doen't work in high speed to full
+                 * speed scenario as NAK interrupts are not coming from a
+                 * full speed device connected to a high speed device.
+                 * NAK timeout interval is 8 (128 uframe or 16ms) for HS and
+                 * 4 (8 frame or 8ms) for FS device.
+                 */
+                if (is_in && qh->dev)
+                        qh->intv_reg =
+                                (USB_SPEED_HIGH == qh->dev->speed) ? 8 : 4;
                 goto success;
         } else if (best_end < 0) {
                 return -ENOSPC;
@@ -1764,7 +1819,6 @@ static int musb_schedule(
         idle = 1;
         qh->mux = 0;
         hw_ep = musb->endpoints + best_end;
-        musb->periodic[best_end] = qh;
         DBG(4, "qh %p periodic slot %d\n", qh, best_end);
 success:
         if (head) {
@@ -1888,13 +1942,11 @@ static int musb_urb_enqueue(
                  *
                  * The downside of disabling this is that transfer scheduling
                  * gets VERY unfair for nonperiodic transfers; a misbehaving
-                 * peripheral could make that hurt.  Or for reads, one that's
-                 * perfectly normal:  network and other drivers keep reads
-                 * posted at all times, having one pending for a week should
-                 * be perfectly safe.
+                 * peripheral could make that hurt.  That's perfectly normal
+                 * for reads from network or serial adapters ... so we have
+                 * partial NAKlimit support for bulk RX.
                  *
-                 * The upside of disabling it is avoidng transfer scheduling
-                 * code to put this aside for while.
+                 * The upside of disabling it is simpler transfer scheduling.
                  */
                 interval = 0;
         }
diff --git a/drivers/usb/musb/musb_virthub.c b/drivers/usb/musb/musb_virthub.c
index e0e9ce584175..bf677acc83db 100644
--- a/drivers/usb/musb/musb_virthub.c
+++ b/drivers/usb/musb/musb_virthub.c
@@ -285,7 +285,7 @@ int musb_hub_control(
                 desc->bDescLength = 9;
                 desc->bDescriptorType = 0x29;
                 desc->bNbrPorts = 1;
-                desc->wHubCharacteristics = __constant_cpu_to_le16(
+                desc->wHubCharacteristics = cpu_to_le16(
                                   0x0001        /* per-port power switching */
                                 | 0x0010        /* no overcurrent reporting */
                                 );