Merge branch 'wip-2.6.34' into old-private-masterarchived-private-master

1 files changed, 637 insertions, 111 deletions

diff --git a/drivers/serial/sh-sci.c b/drivers/serial/sh-sci.c
index 6498bd1fb6dd..8eb094c1f61b 100644
--- a/drivers/serial/sh-sci.c
+++ b/drivers/serial/sh-sci.c
@@ -48,9 +48,11 @@
 #include <linux/ctype.h>
 #include <linux/err.h>
 #include <linux/list.h>
+#include <linux/dmaengine.h>
+#include <linux/scatterlist.h>
+#include <linux/slab.h>
 
 #ifdef CONFIG_SUPERH
-#include <asm/clock.h>
 #include <asm/sh_bios.h>
 #endif
 
@@ -79,22 +81,39 @@ struct sci_port {
         struct timer_list       break_timer;
         int                     break_flag;
 
-#ifdef CONFIG_HAVE_CLK
         /* Interface clock */
         struct clk              *iclk;
         /* Data clock */
         struct clk              *dclk;
-#endif
+
         struct list_head        node;
+        struct dma_chan                 *chan_tx;
+        struct dma_chan                 *chan_rx;
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+        struct device                   *dma_dev;
+        enum sh_dmae_slave_chan_id      slave_tx;
+        enum sh_dmae_slave_chan_id      slave_rx;
+        struct dma_async_tx_descriptor  *desc_tx;
+        struct dma_async_tx_descriptor  *desc_rx[2];
+        dma_cookie_t                    cookie_tx;
+        dma_cookie_t                    cookie_rx[2];
+        dma_cookie_t                    active_rx;
+        struct scatterlist              sg_tx;
+        unsigned int                    sg_len_tx;
+        struct scatterlist              sg_rx[2];
+        size_t                          buf_len_rx;
+        struct sh_dmae_slave            param_tx;
+        struct sh_dmae_slave            param_rx;
+        struct work_struct              work_tx;
+        struct work_struct              work_rx;
+        struct timer_list               rx_timer;
+#endif
 };
 
 struct sh_sci_priv {
         spinlock_t lock;
         struct list_head ports;
-
-#ifdef CONFIG_HAVE_CLK
         struct notifier_block clk_nb;
-#endif
 };
 
 /* Function prototypes */
@@ -156,32 +175,6 @@ static void sci_poll_put_char(struct uart_port *port, unsigned char c)
 }
 #endif /* CONFIG_CONSOLE_POLL || CONFIG_SERIAL_SH_SCI_CONSOLE */
 
-#if defined(__H8300S__)
-enum { sci_disable, sci_enable };
-
-static void h8300_sci_config(struct uart_port *port, unsigned int ctrl)
-{
-        volatile unsigned char *mstpcrl = (volatile unsigned char *)MSTPCRL;
-        int ch = (port->mapbase  - SMR0) >> 3;
-        unsigned char mask = 1 << (ch+1);
-
-        if (ctrl == sci_disable)
-                *mstpcrl |= mask;
-        else
-                *mstpcrl &= ~mask;
-}
-
-static void h8300_sci_enable(struct uart_port *port)
-{
-        h8300_sci_config(port, sci_enable);
-}
-
-static void h8300_sci_disable(struct uart_port *port)
-{
-        h8300_sci_config(port, sci_disable);
-}
-#endif
-
 #if defined(__H8300H__) || defined(__H8300S__)
 static void sci_init_pins(struct uart_port *port, unsigned int cflag)
 {
@@ -253,9 +246,9 @@ static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
                    Set SCP6MD1,0 = {01} (output)  */
                 __raw_writew((data & 0x0fcf) | 0x1000, SCPCR);
 
-                data = ctrl_inb(SCPDR);
+                data = __raw_readb(SCPDR);
                 /* Set /RTS2 (bit6) = 0 */
-                ctrl_outb(data & 0xbf, SCPDR);
+                __raw_writeb(data & 0xbf, SCPDR);
         }
 }
 #elif defined(CONFIG_CPU_SUBTYPE_SH7722)
@@ -300,29 +293,44 @@ static inline void sci_init_pins(struct uart_port *port, unsigned int cflag)
     defined(CONFIG_CPU_SUBTYPE_SH7780) || \
     defined(CONFIG_CPU_SUBTYPE_SH7785) || \
     defined(CONFIG_CPU_SUBTYPE_SH7786)
-static inline int scif_txroom(struct uart_port *port)
+static int scif_txfill(struct uart_port *port)
+{
+        return sci_in(port, SCTFDR) & 0xff;
+}
+
+static int scif_txroom(struct uart_port *port)
 {
-        return SCIF_TXROOM_MAX - (sci_in(port, SCTFDR) & 0xff);
+        return SCIF_TXROOM_MAX - scif_txfill(port);
 }
 
-static inline int scif_rxroom(struct uart_port *port)
+static int scif_rxfill(struct uart_port *port)
 {
         return sci_in(port, SCRFDR) & 0xff;
 }
 #elif defined(CONFIG_CPU_SUBTYPE_SH7763)
-static inline int scif_txroom(struct uart_port *port)
+static int scif_txfill(struct uart_port *port)
 {
-        if ((port->mapbase == 0xffe00000) ||
-            (port->mapbase == 0xffe08000)) {
+        if (port->mapbase == 0xffe00000 ||
+            port->mapbase == 0xffe08000)
                 /* SCIF0/1*/
-                return SCIF_TXROOM_MAX - (sci_in(port, SCTFDR) & 0xff);
-        } else {
+                return sci_in(port, SCTFDR) & 0xff;
+        else
                 /* SCIF2 */
-                return SCIF2_TXROOM_MAX - (sci_in(port, SCFDR) >> 8);
-        }
+                return sci_in(port, SCFDR) >> 8;
+}
+
+static int scif_txroom(struct uart_port *port)
+{
+        if (port->mapbase == 0xffe00000 ||
+            port->mapbase == 0xffe08000)
+                /* SCIF0/1*/
+                return SCIF_TXROOM_MAX - scif_txfill(port);
+        else
+                /* SCIF2 */
+                return SCIF2_TXROOM_MAX - scif_txfill(port);
 }
 
-static inline int scif_rxroom(struct uart_port *port)
+static int scif_rxfill(struct uart_port *port)
 {
         if ((port->mapbase == 0xffe00000) ||
             (port->mapbase == 0xffe08000)) {
@@ -334,23 +342,33 @@ static inline int scif_rxroom(struct uart_port *port)
         }
 }
 #else
-static inline int scif_txroom(struct uart_port *port)
+static int scif_txfill(struct uart_port *port)
+{
+        return sci_in(port, SCFDR) >> 8;
+}
+
+static int scif_txroom(struct uart_port *port)
 {
-        return SCIF_TXROOM_MAX - (sci_in(port, SCFDR) >> 8);
+        return SCIF_TXROOM_MAX - scif_txfill(port);
 }
 
-static inline int scif_rxroom(struct uart_port *port)
+static int scif_rxfill(struct uart_port *port)
 {
         return sci_in(port, SCFDR) & SCIF_RFDC_MASK;
 }
 #endif
 
-static inline int sci_txroom(struct uart_port *port)
+static int sci_txfill(struct uart_port *port)
 {
-        return (sci_in(port, SCxSR) & SCI_TDRE) != 0;
+        return !(sci_in(port, SCxSR) & SCI_TDRE);
 }
 
-static inline int sci_rxroom(struct uart_port *port)
+static int sci_txroom(struct uart_port *port)
+{
+        return !sci_txfill(port);
+}
+
+static int sci_rxfill(struct uart_port *port)
 {
         return (sci_in(port, SCxSR) & SCxSR_RDxF(port)) != 0;
 }
@@ -437,9 +455,9 @@ static inline void sci_receive_chars(struct uart_port *port)
 
         while (1) {
                 if (port->type == PORT_SCI)
-                        count = sci_rxroom(port);
+                        count = sci_rxfill(port);
                 else
-                        count = scif_rxroom(port);
+                        count = scif_rxfill(port);
 
                 /* Don't copy more bytes than there is room for in the buffer */
                 count = tty_buffer_request_room(tty, count);
@@ -484,10 +502,10 @@ static inline void sci_receive_chars(struct uart_port *port)
                                 }
 
                                 /* Store data and status */
-                                if (status&SCxSR_FER(port)) {
+                                if (status & SCxSR_FER(port)) {
                                         flag = TTY_FRAME;
                                         dev_notice(port->dev, "frame error\n");
-                                } else if (status&SCxSR_PER(port)) {
+                                } else if (status & SCxSR_PER(port)) {
                                         flag = TTY_PARITY;
                                         dev_notice(port->dev, "parity error\n");
                                 } else
@@ -649,13 +667,39 @@ static inline int sci_handle_breaks(struct uart_port *port)
         return copied;
 }
 
-static irqreturn_t sci_rx_interrupt(int irq, void *port)
+static irqreturn_t sci_rx_interrupt(int irq, void *ptr)
 {
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+        struct uart_port *port = ptr;
+        struct sci_port *s = to_sci_port(port);
+
+        if (s->chan_rx) {
+                unsigned long tout;
+                u16 scr = sci_in(port, SCSCR);
+                u16 ssr = sci_in(port, SCxSR);
+
+                /* Disable future Rx interrupts */
+                sci_out(port, SCSCR, scr & ~SCI_CTRL_FLAGS_RIE);
+                /* Clear current interrupt */
+                sci_out(port, SCxSR, ssr & ~(1 | SCxSR_RDxF(port)));
+                /* Calculate delay for 1.5 DMA buffers */
+                tout = (port->timeout - HZ / 50) * s->buf_len_rx * 3 /
+                        port->fifosize / 2;
+                dev_dbg(port->dev, "Rx IRQ: setup timeout in %lu ms\n",
+                        tout * 1000 / HZ);
+                if (tout < 2)
+                        tout = 2;
+                mod_timer(&s->rx_timer, jiffies + tout);
+
+                return IRQ_HANDLED;
+        }
+#endif
+
         /* I think sci_receive_chars has to be called irrespective
          * of whether the I_IXOFF is set, otherwise, how is the interrupt
          * to be disabled?
          */
-        sci_receive_chars(port);
+        sci_receive_chars(ptr);
 
         return IRQ_HANDLED;
 }
@@ -711,6 +755,7 @@ static irqreturn_t sci_mpxed_interrupt(int irq, void *ptr)
 {
         unsigned short ssr_status, scr_status, err_enabled;
         struct uart_port *port = ptr;
+        struct sci_port *s = to_sci_port(port);
         irqreturn_t ret = IRQ_NONE;
 
         ssr_status = sci_in(port, SCxSR);
@@ -718,10 +763,15 @@ static irqreturn_t sci_mpxed_interrupt(int irq, void *ptr)
         err_enabled = scr_status & (SCI_CTRL_FLAGS_REIE | SCI_CTRL_FLAGS_RIE);
 
         /* Tx Interrupt */
-        if ((ssr_status & SCxSR_TDxE(port)) && (scr_status & SCI_CTRL_FLAGS_TIE))
+        if ((ssr_status & SCxSR_TDxE(port)) && (scr_status & SCI_CTRL_FLAGS_TIE) &&
+            !s->chan_tx)
                 ret = sci_tx_interrupt(irq, ptr);
-        /* Rx Interrupt */
-        if ((ssr_status & SCxSR_RDxF(port)) && (scr_status & SCI_CTRL_FLAGS_RIE))
+        /*
+         * Rx Interrupt: if we're using DMA, the DMA controller clears RDF /
+         * DR flags
+         */
+        if (((ssr_status & SCxSR_RDxF(port)) || s->chan_rx) &&
+            (scr_status & SCI_CTRL_FLAGS_RIE))
                 ret = sci_rx_interrupt(irq, ptr);
         /* Error Interrupt */
         if ((ssr_status & SCxSR_ERRORS(port)) && err_enabled)
@@ -733,7 +783,6 @@ static irqreturn_t sci_mpxed_interrupt(int irq, void *ptr)
         return ret;
 }
 
-#ifdef CONFIG_HAVE_CLK
 /*
  * Here we define a transistion notifier so that we can update all of our
  * ports' baud rate when the peripheral clock changes.
@@ -751,7 +800,6 @@ static int sci_notifier(struct notifier_block *self,
                 spin_lock_irqsave(&priv->lock, flags);
                 list_for_each_entry(sci_port, &priv->ports, node)
                         sci_port->port.uartclk = clk_get_rate(sci_port->dclk);
-
                 spin_unlock_irqrestore(&priv->lock, flags);
         }
 
@@ -778,7 +826,6 @@ static void sci_clk_disable(struct uart_port *port)
 
         clk_disable(sci_port->dclk);
 }
-#endif
 
 static int sci_request_irq(struct sci_port *port)
 {
@@ -833,8 +880,10 @@ static void sci_free_irq(struct sci_port *port)
 
 static unsigned int sci_tx_empty(struct uart_port *port)
 {
-        /* Can't detect */
-        return TIOCSER_TEMT;
+        unsigned short status = sci_in(port, SCxSR);
+        unsigned short in_tx_fifo = scif_txfill(port);
+
+        return (status & SCxSR_TEND(port)) && !in_tx_fifo ? TIOCSER_TEMT : 0;
 }
 
 static void sci_set_mctrl(struct uart_port *port, unsigned int mctrl)
@@ -846,16 +895,297 @@ static void sci_set_mctrl(struct uart_port *port, unsigned int mctrl)
 
 static unsigned int sci_get_mctrl(struct uart_port *port)
 {
-        /* This routine is used for geting signals of: DTR, DCD, DSR, RI,
+        /* This routine is used for getting signals of: DTR, DCD, DSR, RI,
            and CTS/RTS */
 
         return TIOCM_DTR | TIOCM_RTS | TIOCM_DSR;
 }
 
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+static void sci_dma_tx_complete(void *arg)
+{
+        struct sci_port *s = arg;
+        struct uart_port *port = &s->port;
+        struct circ_buf *xmit = &port->state->xmit;
+        unsigned long flags;
+
+        dev_dbg(port->dev, "%s(%d)\n", __func__, port->line);
+
+        spin_lock_irqsave(&port->lock, flags);
+
+        xmit->tail += s->sg_tx.length;
+        xmit->tail &= UART_XMIT_SIZE - 1;
+
+        port->icount.tx += s->sg_tx.length;
+
+        async_tx_ack(s->desc_tx);
+        s->cookie_tx = -EINVAL;
+        s->desc_tx = NULL;
+
+        spin_unlock_irqrestore(&port->lock, flags);
+
+        if (uart_circ_chars_pending(xmit) < WAKEUP_CHARS)
+                uart_write_wakeup(port);
+
+        if (uart_circ_chars_pending(xmit))
+                schedule_work(&s->work_tx);
+}
+
+/* Locking: called with port lock held */
+static int sci_dma_rx_push(struct sci_port *s, struct tty_struct *tty,
+                           size_t count)
+{
+        struct uart_port *port = &s->port;
+        int i, active, room;
+
+        room = tty_buffer_request_room(tty, count);
+
+        if (s->active_rx == s->cookie_rx[0]) {
+                active = 0;
+        } else if (s->active_rx == s->cookie_rx[1]) {
+                active = 1;
+        } else {
+                dev_err(port->dev, "cookie %d not found!\n", s->active_rx);
+                return 0;
+        }
+
+        if (room < count)
+                dev_warn(port->dev, "Rx overrun: dropping %u bytes\n",
+                         count - room);
+        if (!room)
+                return room;
+
+        for (i = 0; i < room; i++)
+                tty_insert_flip_char(tty, ((u8 *)sg_virt(&s->sg_rx[active]))[i],
+                                     TTY_NORMAL);
+
+        port->icount.rx += room;
+
+        return room;
+}
+
+static void sci_dma_rx_complete(void *arg)
+{
+        struct sci_port *s = arg;
+        struct uart_port *port = &s->port;
+        struct tty_struct *tty = port->state->port.tty;
+        unsigned long flags;
+        int count;
+
+        dev_dbg(port->dev, "%s(%d)\n", __func__, port->line);
+
+        spin_lock_irqsave(&port->lock, flags);
+
+        count = sci_dma_rx_push(s, tty, s->buf_len_rx);
+
+        mod_timer(&s->rx_timer, jiffies + msecs_to_jiffies(5));
+
+        spin_unlock_irqrestore(&port->lock, flags);
+
+        if (count)
+                tty_flip_buffer_push(tty);
+
+        schedule_work(&s->work_rx);
+}
+
+static void sci_start_rx(struct uart_port *port);
+static void sci_start_tx(struct uart_port *port);
+
+static void sci_rx_dma_release(struct sci_port *s, bool enable_pio)
+{
+        struct dma_chan *chan = s->chan_rx;
+        struct uart_port *port = &s->port;
+
+        s->chan_rx = NULL;
+        s->cookie_rx[0] = s->cookie_rx[1] = -EINVAL;
+        dma_release_channel(chan);
+        dma_free_coherent(port->dev, s->buf_len_rx * 2,
+                          sg_virt(&s->sg_rx[0]), sg_dma_address(&s->sg_rx[0]));
+        if (enable_pio)
+                sci_start_rx(port);
+}
+
+static void sci_tx_dma_release(struct sci_port *s, bool enable_pio)
+{
+        struct dma_chan *chan = s->chan_tx;
+        struct uart_port *port = &s->port;
+
+        s->chan_tx = NULL;
+        s->cookie_tx = -EINVAL;
+        dma_release_channel(chan);
+        if (enable_pio)
+                sci_start_tx(port);
+}
+
+static void sci_submit_rx(struct sci_port *s)
+{
+        struct dma_chan *chan = s->chan_rx;
+        int i;
+
+        for (i = 0; i < 2; i++) {
+                struct scatterlist *sg = &s->sg_rx[i];
+                struct dma_async_tx_descriptor *desc;
+
+                desc = chan->device->device_prep_slave_sg(chan,
+                        sg, 1, DMA_FROM_DEVICE, DMA_PREP_INTERRUPT);
+
+                if (desc) {
+                        s->desc_rx[i] = desc;
+                        desc->callback = sci_dma_rx_complete;
+                        desc->callback_param = s;
+                        s->cookie_rx[i] = desc->tx_submit(desc);
+                }
+
+                if (!desc || s->cookie_rx[i] < 0) {
+                        if (i) {
+                                async_tx_ack(s->desc_rx[0]);
+                                s->cookie_rx[0] = -EINVAL;
+                        }
+                        if (desc) {
+                                async_tx_ack(desc);
+                                s->cookie_rx[i] = -EINVAL;
+                        }
+                        dev_warn(s->port.dev,
+                                 "failed to re-start DMA, using PIO\n");
+                        sci_rx_dma_release(s, true);
+                        return;
+                }
+        }
+
+        s->active_rx = s->cookie_rx[0];
+
+        dma_async_issue_pending(chan);
+}
+
+static void work_fn_rx(struct work_struct *work)
+{
+        struct sci_port *s = container_of(work, struct sci_port, work_rx);
+        struct uart_port *port = &s->port;
+        struct dma_async_tx_descriptor *desc;
+        int new;
+
+        if (s->active_rx == s->cookie_rx[0]) {
+                new = 0;
+        } else if (s->active_rx == s->cookie_rx[1]) {
+                new = 1;
+        } else {
+                dev_err(port->dev, "cookie %d not found!\n", s->active_rx);
+                return;
+        }
+        desc = s->desc_rx[new];
+
+        if (dma_async_is_tx_complete(s->chan_rx, s->active_rx, NULL, NULL) !=
+            DMA_SUCCESS) {
+                /* Handle incomplete DMA receive */
+                struct tty_struct *tty = port->state->port.tty;
+                struct dma_chan *chan = s->chan_rx;
+                struct sh_desc *sh_desc = container_of(desc, struct sh_desc,
+                                                       async_tx);
+                unsigned long flags;
+                int count;
+
+                chan->device->device_terminate_all(chan);
+                dev_dbg(port->dev, "Read %u bytes with cookie %d\n",
+                        sh_desc->partial, sh_desc->cookie);
+
+                spin_lock_irqsave(&port->lock, flags);
+                count = sci_dma_rx_push(s, tty, sh_desc->partial);
+                spin_unlock_irqrestore(&port->lock, flags);
+
+                if (count)
+                        tty_flip_buffer_push(tty);
+
+                sci_submit_rx(s);
+
+                return;
+        }
+
+        s->cookie_rx[new] = desc->tx_submit(desc);
+        if (s->cookie_rx[new] < 0) {
+                dev_warn(port->dev, "Failed submitting Rx DMA descriptor\n");
+                sci_rx_dma_release(s, true);
+                return;
+        }
+
+        dev_dbg(port->dev, "%s: cookie %d #%d\n", __func__,
+                s->cookie_rx[new], new);
+
+        s->active_rx = s->cookie_rx[!new];
+}
+
+static void work_fn_tx(struct work_struct *work)
+{
+        struct sci_port *s = container_of(work, struct sci_port, work_tx);
+        struct dma_async_tx_descriptor *desc;
+        struct dma_chan *chan = s->chan_tx;
+        struct uart_port *port = &s->port;
+        struct circ_buf *xmit = &port->state->xmit;
+        struct scatterlist *sg = &s->sg_tx;
+
+        /*
+         * DMA is idle now.
+         * Port xmit buffer is already mapped, and it is one page... Just adjust
+         * offsets and lengths. Since it is a circular buffer, we have to
+         * transmit till the end, and then the rest. Take the port lock to get a
+         * consistent xmit buffer state.
+         */
+        spin_lock_irq(&port->lock);
+        sg->offset = xmit->tail & (UART_XMIT_SIZE - 1);
+        sg->dma_address = (sg_dma_address(sg) & ~(UART_XMIT_SIZE - 1)) +
+                sg->offset;
+        sg->length = min((int)CIRC_CNT(xmit->head, xmit->tail, UART_XMIT_SIZE),
+                CIRC_CNT_TO_END(xmit->head, xmit->tail, UART_XMIT_SIZE));
+        sg->dma_length = sg->length;
+        spin_unlock_irq(&port->lock);
+
+        BUG_ON(!sg->length);
+
+        desc = chan->device->device_prep_slave_sg(chan,
+                        sg, s->sg_len_tx, DMA_TO_DEVICE,
+                        DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+        if (!desc) {
+                /* switch to PIO */
+                sci_tx_dma_release(s, true);
+                return;
+        }
+
+        dma_sync_sg_for_device(port->dev, sg, 1, DMA_TO_DEVICE);
+
+        spin_lock_irq(&port->lock);
+        s->desc_tx = desc;
+        desc->callback = sci_dma_tx_complete;
+        desc->callback_param = s;
+        spin_unlock_irq(&port->lock);
+        s->cookie_tx = desc->tx_submit(desc);
+        if (s->cookie_tx < 0) {
+                dev_warn(port->dev, "Failed submitting Tx DMA descriptor\n");
+                /* switch to PIO */
+                sci_tx_dma_release(s, true);
+                return;
+        }
+
+        dev_dbg(port->dev, "%s: %p: %d...%d, cookie %d\n", __func__,
+                xmit->buf, xmit->tail, xmit->head, s->cookie_tx);
+
+        dma_async_issue_pending(chan);
+}
+#endif
+
 static void sci_start_tx(struct uart_port *port)
 {
         unsigned short ctrl;
 
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+        struct sci_port *s = to_sci_port(port);
+
+        if (s->chan_tx) {
+                if (!uart_circ_empty(&s->port.state->xmit) && s->cookie_tx < 0)
+                        schedule_work(&s->work_tx);
+
+                return;
+        }
+#endif
+
         /* Set TIE (Transmit Interrupt Enable) bit in SCSCR */
         ctrl = sci_in(port, SCSCR);
         ctrl |= SCI_CTRL_FLAGS_TIE;
@@ -872,13 +1202,12 @@ static void sci_stop_tx(struct uart_port *port)
         sci_out(port, SCSCR, ctrl);
 }
 
-static void sci_start_rx(struct uart_port *port, unsigned int tty_start)
+static void sci_start_rx(struct uart_port *port)
 {
-        unsigned short ctrl;
+        unsigned short ctrl = SCI_CTRL_FLAGS_RIE | SCI_CTRL_FLAGS_REIE;
 
         /* Set RIE (Receive Interrupt Enable) bit in SCSCR */
-        ctrl = sci_in(port, SCSCR);
-        ctrl |= SCI_CTRL_FLAGS_RIE | SCI_CTRL_FLAGS_REIE;
+        ctrl |= sci_in(port, SCSCR);
         sci_out(port, SCSCR, ctrl);
 }
 
@@ -902,16 +1231,154 @@ static void sci_break_ctl(struct uart_port *port, int break_state)
         /* Nothing here yet .. */
 }
 
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+static bool filter(struct dma_chan *chan, void *slave)
+{
+        struct sh_dmae_slave *param = slave;
+
+        dev_dbg(chan->device->dev, "%s: slave ID %d\n", __func__,
+                param->slave_id);
+
+        if (param->dma_dev == chan->device->dev) {
+                chan->private = param;
+                return true;
+        } else {
+                return false;
+        }
+}
+
+static void rx_timer_fn(unsigned long arg)
+{
+        struct sci_port *s = (struct sci_port *)arg;
+        struct uart_port *port = &s->port;
+
+        u16 scr = sci_in(port, SCSCR);
+        sci_out(port, SCSCR, scr | SCI_CTRL_FLAGS_RIE);
+        dev_dbg(port->dev, "DMA Rx timed out\n");
+        schedule_work(&s->work_rx);
+}
+
+static void sci_request_dma(struct uart_port *port)
+{
+        struct sci_port *s = to_sci_port(port);
+        struct sh_dmae_slave *param;
+        struct dma_chan *chan;
+        dma_cap_mask_t mask;
+        int nent;
+
+        dev_dbg(port->dev, "%s: port %d DMA %p\n", __func__,
+                port->line, s->dma_dev);
+
+        if (!s->dma_dev)
+                return;
+
+        dma_cap_zero(mask);
+        dma_cap_set(DMA_SLAVE, mask);
+
+        param = &s->param_tx;
+
+        /* Slave ID, e.g., SHDMA_SLAVE_SCIF0_TX */
+        param->slave_id = s->slave_tx;
+        param->dma_dev = s->dma_dev;
+
+        s->cookie_tx = -EINVAL;
+        chan = dma_request_channel(mask, filter, param);
+        dev_dbg(port->dev, "%s: TX: got channel %p\n", __func__, chan);
+        if (chan) {
+                s->chan_tx = chan;
+                sg_init_table(&s->sg_tx, 1);
+                /* UART circular tx buffer is an aligned page. */
+                BUG_ON((int)port->state->xmit.buf & ~PAGE_MASK);
+                sg_set_page(&s->sg_tx, virt_to_page(port->state->xmit.buf),
+                            UART_XMIT_SIZE, (int)port->state->xmit.buf & ~PAGE_MASK);
+                nent = dma_map_sg(port->dev, &s->sg_tx, 1, DMA_TO_DEVICE);
+                if (!nent)
+                        sci_tx_dma_release(s, false);
+                else
+                        dev_dbg(port->dev, "%s: mapped %d@%p to %x\n", __func__,
+                                sg_dma_len(&s->sg_tx),
+                                port->state->xmit.buf, sg_dma_address(&s->sg_tx));
+
+                s->sg_len_tx = nent;
+
+                INIT_WORK(&s->work_tx, work_fn_tx);
+        }
+
+        param = &s->param_rx;
+
+        /* Slave ID, e.g., SHDMA_SLAVE_SCIF0_RX */
+        param->slave_id = s->slave_rx;
+        param->dma_dev = s->dma_dev;
+
+        chan = dma_request_channel(mask, filter, param);
+        dev_dbg(port->dev, "%s: RX: got channel %p\n", __func__, chan);
+        if (chan) {
+                dma_addr_t dma[2];
+                void *buf[2];
+                int i;
+
+                s->chan_rx = chan;
+
+                s->buf_len_rx = 2 * max(16, (int)port->fifosize);
+                buf[0] = dma_alloc_coherent(port->dev, s->buf_len_rx * 2,
+                                            &dma[0], GFP_KERNEL);
+
+                if (!buf[0]) {
+                        dev_warn(port->dev,
+                                 "failed to allocate dma buffer, using PIO\n");
+                        sci_rx_dma_release(s, true);
+                        return;
+                }
+
+                buf[1] = buf[0] + s->buf_len_rx;
+                dma[1] = dma[0] + s->buf_len_rx;
+
+                for (i = 0; i < 2; i++) {
+                        struct scatterlist *sg = &s->sg_rx[i];
+
+                        sg_init_table(sg, 1);
+                        sg_set_page(sg, virt_to_page(buf[i]), s->buf_len_rx,
+                                    (int)buf[i] & ~PAGE_MASK);
+                        sg->dma_address = dma[i];
+                        sg->dma_length = sg->length;
+                }
+
+                INIT_WORK(&s->work_rx, work_fn_rx);
+                setup_timer(&s->rx_timer, rx_timer_fn, (unsigned long)s);
+
+                sci_submit_rx(s);
+        }
+}
+
+static void sci_free_dma(struct uart_port *port)
+{
+        struct sci_port *s = to_sci_port(port);
+
+        if (!s->dma_dev)
+                return;
+
+        if (s->chan_tx)
+                sci_tx_dma_release(s, false);
+        if (s->chan_rx)
+                sci_rx_dma_release(s, false);
+}
+#endif
+
 static int sci_startup(struct uart_port *port)
 {
         struct sci_port *s = to_sci_port(port);
 
+        dev_dbg(port->dev, "%s(%d)\n", __func__, port->line);
+
         if (s->enable)
                 s->enable(port);
 
         sci_request_irq(s);
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+        sci_request_dma(port);
+#endif
         sci_start_tx(port);
-        sci_start_rx(port, 1);
+        sci_start_rx(port);
 
         return 0;
 }
@@ -920,8 +1387,13 @@ static void sci_shutdown(struct uart_port *port)
 {
         struct sci_port *s = to_sci_port(port);
 
+        dev_dbg(port->dev, "%s(%d)\n", __func__, port->line);
+
         sci_stop_rx(port);
         sci_stop_tx(port);
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+        sci_free_dma(port);
+#endif
         sci_free_irq(s);
 
         if (s->disable)
@@ -931,11 +1403,21 @@ static void sci_shutdown(struct uart_port *port)
 static void sci_set_termios(struct uart_port *port, struct ktermios *termios,
                             struct ktermios *old)
 {
-        unsigned int status, baud, smr_val;
+        unsigned int status, baud, smr_val, max_baud;
         int t = -1;
 
-        baud = uart_get_baud_rate(port, termios, old, 0, port->uartclk/16);
-        if (likely(baud))
+        /*
+         * earlyprintk comes here early on with port->uartclk set to zero.
+         * the clock framework is not up and running at this point so here
+         * we assume that 115200 is the maximum baud rate. please note that
+         * the baud rate is not programmed during earlyprintk - it is assumed
+         * that the previous boot loader has enabled required clocks and
+         * setup the baud rate generator hardware for us already.
+         */
+        max_baud = port->uartclk ? port->uartclk / 16 : 115200;
+
+        baud = uart_get_baud_rate(port, termios, old, 0, max_baud);
+        if (likely(baud && port->uartclk))
                 t = SCBRR_VALUE(baud, port->uartclk);
 
         do {
@@ -961,6 +1443,9 @@ static void sci_set_termios(struct uart_port *port, struct ktermios *termios,
 
         sci_out(port, SCSMR, smr_val);
 
+        dev_dbg(port->dev, "%s: SMR %x, t %x, SCSCR %x\n", __func__, smr_val, t,
+                SCSCR_INIT(port));
+
         if (t > 0) {
                 if (t >= 256) {
                         sci_out(port, SCSMR, (sci_in(port, SCSMR) & ~3) | 1);
@@ -978,7 +1463,7 @@ static void sci_set_termios(struct uart_port *port, struct ktermios *termios,
         sci_out(port, SCSCR, SCSCR_INIT(port));
 
         if ((termios->c_cflag & CREAD) != 0)
-                sci_start_rx(port, 0);
+                sci_start_rx(port);
 }
 
 static const char *sci_type(struct uart_port *port)
@@ -1073,40 +1558,53 @@ static void __devinit sci_init_single(struct platform_device *dev,
                                       unsigned int index,
                                       struct plat_sci_port *p)
 {
-        sci_port->port.ops      = &sci_uart_ops;
-        sci_port->port.iotype   = UPIO_MEM;
-        sci_port->port.line     = index;
-        sci_port->port.fifosize = 1;
+        struct uart_port *port = &sci_port->port;
 
-#if defined(__H8300H__) || defined(__H8300S__)
-#ifdef __H8300S__
-        sci_port->enable        = h8300_sci_enable;
-        sci_port->disable       = h8300_sci_disable;
-#endif
-        sci_port->port.uartclk  = CONFIG_CPU_CLOCK;
-#elif defined(CONFIG_HAVE_CLK)
-        sci_port->iclk          = p->clk ? clk_get(&dev->dev, p->clk) : NULL;
-        sci_port->dclk          = clk_get(&dev->dev, "peripheral_clk");
-        sci_port->enable        = sci_clk_enable;
-        sci_port->disable       = sci_clk_disable;
-#else
-#error "Need a valid uartclk"
-#endif
+        port->ops       = &sci_uart_ops;
+        port->iotype    = UPIO_MEM;
+        port->line      = index;
+
+        switch (p->type) {
+        case PORT_SCIFA:
+                port->fifosize = 64;
+                break;
+        case PORT_SCIF:
+                port->fifosize = 16;
+                break;
+        default:
+                port->fifosize = 1;
+                break;
+        }
+
+        if (dev) {
+                sci_port->iclk = p->clk ? clk_get(&dev->dev, p->clk) : NULL;
+                sci_port->dclk = clk_get(&dev->dev, "peripheral_clk");
+                sci_port->enable = sci_clk_enable;
+                sci_port->disable = sci_clk_disable;
+                port->dev = &dev->dev;
+        }
 
         sci_port->break_timer.data = (unsigned long)sci_port;
         sci_port->break_timer.function = sci_break_timer;
         init_timer(&sci_port->break_timer);
 
-        sci_port->port.mapbase  = p->mapbase;
-        sci_port->port.membase  = p->membase;
+        port->mapbase   = p->mapbase;
+        port->membase   = p->membase;
 
-        sci_port->port.irq      = p->irqs[SCIx_TXI_IRQ];
-        sci_port->port.flags    = p->flags;
-        sci_port->port.dev      = &dev->dev;
-        sci_port->type          = sci_port->port.type = p->type;
+        port->irq       = p->irqs[SCIx_TXI_IRQ];
+        port->flags     = p->flags;
+        sci_port->type  = port->type = p->type;
 
-        memcpy(&sci_port->irqs, &p->irqs, sizeof(p->irqs));
+#ifdef CONFIG_SERIAL_SH_SCI_DMA
+        sci_port->dma_dev       = p->dma_dev;
+        sci_port->slave_tx      = p->dma_slave_tx;
+        sci_port->slave_rx      = p->dma_slave_rx;
 
+        dev_dbg(port->dev, "%s: DMA device %p, tx %d, rx %d\n", __func__,
+                p->dma_dev, p->dma_slave_tx, p->dma_slave_rx);
+#endif
+
+        memcpy(&sci_port->irqs, &p->irqs, sizeof(p->irqs));
 }
 
 #ifdef CONFIG_SERIAL_SH_SCI_CONSOLE
@@ -1147,7 +1645,7 @@ static void serial_console_write(struct console *co, const char *s,
                 sci_port->disable(port);
 }
 
-static int __init serial_console_setup(struct console *co, char *options)
+static int __devinit serial_console_setup(struct console *co, char *options)
 {
         struct sci_port *sci_port;
         struct uart_port *port;
@@ -1165,9 +1663,14 @@ static int __init serial_console_setup(struct console *co, char *options)
         if (co->index >= SCI_NPORTS)
                 co->index = 0;
 
-        sci_port = &sci_ports[co->index];
-        port = &sci_port->port;
-        co->data = port;
+        if (co->data) {
+                port = co->data;
+                sci_port = to_sci_port(port);
+        } else {
+                sci_port = &sci_ports[co->index];
+                port = &sci_port->port;
+                co->data = port;
+        }
 
         /*
          * Also need to check port->type, we don't actually have any
@@ -1211,6 +1714,15 @@ static int __init sci_console_init(void)
         return 0;
 }
 console_initcall(sci_console_init);
+
+static struct sci_port early_serial_port;
+static struct console early_serial_console = {
+        .name           = "early_ttySC",
+        .write          = serial_console_write,
+        .flags          = CON_PRINTBUFFER,
+};
+static char early_serial_buf[32];
+
 #endif /* CONFIG_SERIAL_SH_SCI_CONSOLE */
 
 #if defined(CONFIG_SERIAL_SH_SCI_CONSOLE)
@@ -1239,14 +1751,11 @@ static int sci_remove(struct platform_device *dev)
         struct sci_port *p;
         unsigned long flags;
 
-#ifdef CONFIG_HAVE_CLK
         cpufreq_unregister_notifier(&priv->clk_nb, CPUFREQ_TRANSITION_NOTIFIER);
-#endif
 
         spin_lock_irqsave(&priv->lock, flags);
         list_for_each_entry(p, &priv->ports, node)
                 uart_remove_one_port(&sci_uart_driver, &p->port);
-
         spin_unlock_irqrestore(&priv->lock, flags);
 
         kfree(priv);
@@ -1299,6 +1808,21 @@ static int __devinit sci_probe(struct platform_device *dev)
         struct sh_sci_priv *priv;
         int i, ret = -EINVAL;
 
+#ifdef CONFIG_SERIAL_SH_SCI_CONSOLE
+        if (is_early_platform_device(dev)) {
+                if (dev->id == -1)
+                        return -ENOTSUPP;
+                early_serial_console.index = dev->id;
+                early_serial_console.data = &early_serial_port.port;
+                sci_init_single(NULL, &early_serial_port, dev->id, p);
+                serial_console_setup(&early_serial_console, early_serial_buf);
+                if (!strstr(early_serial_buf, "keep"))
+                        early_serial_console.flags |= CON_BOOT;
+                register_console(&early_serial_console);
+                return 0;
+        }
+#endif
+
         priv = kzalloc(sizeof(*priv), GFP_KERNEL);
         if (!priv)
                 return -ENOMEM;
@@ -1307,10 +1831,8 @@ static int __devinit sci_probe(struct platform_device *dev)
         spin_lock_init(&priv->lock);
         platform_set_drvdata(dev, priv);
 
-#ifdef CONFIG_HAVE_CLK
         priv->clk_nb.notifier_call = sci_notifier;
         cpufreq_register_notifier(&priv->clk_nb, CPUFREQ_TRANSITION_NOTIFIER);
-#endif
 
         if (dev->id != -1) {
                 ret = sci_probe_single(dev, dev->id, p, &sci_ports[dev->id]);
@@ -1363,14 +1885,14 @@ static int sci_resume(struct device *dev)
         return 0;
 }
 
-static struct dev_pm_ops sci_dev_pm_ops = {
+static const struct dev_pm_ops sci_dev_pm_ops = {
         .suspend        = sci_suspend,
         .resume         = sci_resume,
 };
 
 static struct platform_driver sci_driver = {
         .probe          = sci_probe,
-        .remove         = __devexit_p(sci_remove),
+        .remove         = sci_remove,
         .driver         = {
                 .name   = "sh-sci",
                 .owner  = THIS_MODULE,
@@ -1400,6 +1922,10 @@ static void __exit sci_exit(void)
         uart_unregister_driver(&sci_uart_driver);
 }
 
+#ifdef CONFIG_SERIAL_SH_SCI_CONSOLE
+early_platform_init_buffer("earlyprintk", &sci_driver,
+                           early_serial_buf, ARRAY_SIZE(early_serial_buf));
+#endif
 module_init(sci_init);
 module_exit(sci_exit);