xtensa: s6000 dma engine support

There are four slightly different dma engines on the s6000 family.
One for memory-memory transfers, the other three for memory-device.

This patch implements a platform-specific kernel-API to control these
engines.  It is needed for the network, video, audio peripherals on
s6000.

Signed-off-by: Oskar Schirmer <os@emlix.com>
Signed-off-by: Daniel Glockner <dg@emlix.com>
Signed-off-by: Fabian Godehardt <fg@emlix.com>
Cc: Daniel Glockner <dg@emlix.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Chris Zankel <chris@zankel.net>

arch/xtensa/variants/s6000/Makefile

@@ -1,4 +1,4 @@
 # s6000 Makefile
 
-obj-y		+= irq.o gpio.o
+obj-y		+= irq.o gpio.o dmac.o
 obj-$(CONFIG_XTENSA_CALIBRATE_CCOUNT)	+= delay.o

arch/xtensa/variants/s6000/dmac.c

@@ -0,0 +1,173 @@
+/*
+ * Authors:	Oskar Schirmer <os@emlix.com>
+ *		Daniel Gloeckner <dg@emlix.com>
+ * (c) 2008 emlix GmbH http://www.emlix.com
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ */
+
+#include <linux/kernel.h>
+#include <linux/io.h>
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/spinlock.h>
+#include <asm/cacheflush.h>
+#include <variant/dmac.h>
+
+/* DMA engine lookup */
+
+struct s6dmac_ctrl s6dmac_ctrl[S6_DMAC_NB];
+
+
+/* DMA control, per engine */
+
+void s6dmac_put_fifo_cache(u32 dmac, int chan, u32 src, u32 dst, u32 size)
+{
+	if (xtensa_need_flush_dma_source(src)) {
+		u32 base = src;
+		u32 span = size;
+		u32 chunk = readl(DMA_CHNL(dmac, chan) + S6_DMA_CMONCHUNK);
+		if (chunk && (size > chunk)) {
+			s32 skip =
+				readl(DMA_CHNL(dmac, chan) + S6_DMA_SRCSKIP);
+			u32 gaps = (size+chunk-1)/chunk - 1;
+			if (skip >= 0) {
+				span += gaps * skip;
+			} else if (-skip > chunk) {
+				s32 decr = gaps * (chunk + skip);
+				base += decr;
+				span = chunk - decr;
+			} else {
+				span = max(span + gaps * skip,
+					(chunk + skip) * gaps - skip);
+			}
+		}
+		flush_dcache_unaligned(base, span);
+	}
+	if (xtensa_need_invalidate_dma_destination(dst)) {
+		u32 base = dst;
+		u32 span = size;
+		u32 chunk = readl(DMA_CHNL(dmac, chan) + S6_DMA_CMONCHUNK);
+		if (chunk && (size > chunk)) {
+			s32 skip =
+				readl(DMA_CHNL(dmac, chan) + S6_DMA_DSTSKIP);
+			u32 gaps = (size+chunk-1)/chunk - 1;
+			if (skip >= 0) {
+				span += gaps * skip;
+			} else if (-skip > chunk) {
+				s32 decr = gaps * (chunk + skip);
+				base += decr;
+				span = chunk - decr;
+			} else {
+				span = max(span + gaps * skip,
+					(chunk + skip) * gaps - skip);
+			}
+		}
+		invalidate_dcache_unaligned(base, span);
+	}
+	s6dmac_put_fifo(dmac, chan, src, dst, size);
+}
+
+void s6dmac_disable_error_irqs(u32 dmac, u32 mask)
+{
+	unsigned long flags;
+	spinlock_t *spinl = &s6dmac_ctrl[_dmac_addr_index(dmac)].lock;
+	spin_lock_irqsave(spinl, flags);
+	_s6dmac_disable_error_irqs(dmac, mask);
+	spin_unlock_irqrestore(spinl, flags);
+}
+
+u32 s6dmac_int_sources(u32 dmac, u32 channel)
+{
+	u32 mask, ret, tmp;
+	mask = 1 << channel;
+
+	tmp = readl(dmac + S6_DMA_TERMCNTIRQSTAT);
+	tmp &= mask;
+	writel(tmp, dmac + S6_DMA_TERMCNTIRQCLR);
+	ret = tmp >> channel;
+
+	tmp = readl(dmac + S6_DMA_PENDCNTIRQSTAT);
+	tmp &= mask;
+	writel(tmp, dmac + S6_DMA_PENDCNTIRQCLR);
+	ret |= (tmp >> channel) << 1;
+
+	tmp = readl(dmac + S6_DMA_LOWWMRKIRQSTAT);
+	tmp &= mask;
+	writel(tmp, dmac + S6_DMA_LOWWMRKIRQCLR);
+	ret |= (tmp >> channel) << 2;
+
+	tmp = readl(dmac + S6_DMA_INTRAW0);
+	tmp &= (mask << S6_DMA_INT0_OVER) | (mask << S6_DMA_INT0_UNDER);
+	writel(tmp, dmac + S6_DMA_INTCLEAR0);
+
+	if (tmp & (mask << S6_DMA_INT0_UNDER))
+		ret |= 1 << 3;
+	if (tmp & (mask << S6_DMA_INT0_OVER))
+		ret |= 1 << 4;
+
+	tmp = readl(dmac + S6_DMA_MASTERERRINFO);
+	mask <<= S6_DMA_INT1_CHANNEL;
+	if (((tmp >> S6_DMA_MASTERERR_CHAN(0)) & S6_DMA_MASTERERR_CHAN_MASK)
+			== channel)
+		mask |= 1 << S6_DMA_INT1_MASTER;
+	if (((tmp >> S6_DMA_MASTERERR_CHAN(1)) & S6_DMA_MASTERERR_CHAN_MASK)
+			== channel)
+		mask |= 1 << (S6_DMA_INT1_MASTER + 1);
+	if (((tmp >> S6_DMA_MASTERERR_CHAN(2)) & S6_DMA_MASTERERR_CHAN_MASK)
+			== channel)
+		mask |= 1 << (S6_DMA_INT1_MASTER + 2);
+
+	tmp = readl(dmac + S6_DMA_INTRAW1) & mask;
+	writel(tmp, dmac + S6_DMA_INTCLEAR1);
+	ret |= ((tmp >> channel) & 1) << 5;
+	ret |= ((tmp >> S6_DMA_INT1_MASTER) & S6_DMA_INT1_MASTER_MASK) << 6;
+
+	return ret;
+}
+
+void s6dmac_release_chan(u32 dmac, int chan)
+{
+	if (chan >= 0)
+		s6dmac_disable_chan(dmac, chan);
+}
+
+
+/* global init */
+
+static inline void __init dmac_init(u32 dmac, u8 chan_nb)
+{
+	s6dmac_ctrl[S6_DMAC_INDEX(dmac)].dmac = dmac;
+	spin_lock_init(&s6dmac_ctrl[S6_DMAC_INDEX(dmac)].lock);
+	s6dmac_ctrl[S6_DMAC_INDEX(dmac)].chan_nb = chan_nb;
+	writel(S6_DMA_INT1_MASTER_MASK << S6_DMA_INT1_MASTER,
+		dmac + S6_DMA_INTCLEAR1);
+}
+
+static inline void __init dmac_master(u32 dmac,
+	u32 m0start, u32 m0end, u32 m1start, u32 m1end)
+{
+	writel(m0start, dmac + S6_DMA_MASTER0START);
+	writel(m0end - 1, dmac + S6_DMA_MASTER0END);
+	writel(m1start, dmac + S6_DMA_MASTER1START);
+	writel(m1end - 1, dmac + S6_DMA_MASTER1END);
+}
+
+static void __init s6_dmac_init(void)
+{
+	dmac_init(S6_REG_LMSDMA, S6_LMSDMA_NB);
+	dmac_master(S6_REG_LMSDMA,
+		S6_MEM_DDR, S6_MEM_PCIE_APER, S6_MEM_EFI, S6_MEM_GMAC);
+	dmac_init(S6_REG_NIDMA, S6_NIDMA_NB);
+	dmac_init(S6_REG_DPDMA, S6_DPDMA_NB);
+	dmac_master(S6_REG_DPDMA,
+		S6_MEM_DDR, S6_MEM_PCIE_APER, S6_REG_DP, S6_REG_DPDMA);
+	dmac_init(S6_REG_HIFDMA, S6_HIFDMA_NB);
+	dmac_master(S6_REG_HIFDMA,
+		S6_MEM_GMAC, S6_MEM_PCIE_CFG, S6_MEM_PCIE_APER, S6_MEM_AUX);
+}
+
+arch_initcall(s6_dmac_init);