2010-07-23 00:57:18

by Tirumala Reddy Marri

[permalink] [raw]
Subject: [PATCH] Adding ADMA support for PPC460EX DMA engine.

From: Tirumala Marri <[email protected]>

This patch will add ADMA support for DMA engine and HW offload for
XOR/ADG (RAID-5/6) functionalities.
1. It supports memcpy, xor, GF(2) based RAID-6.
2. It supports interrupt based DMA completions.
3. Also supports memcpy in RAID-1 case.

Kernel version: 2.6.35-rc5

Testing:
Created RAID-5/6 arrays usign mdadm.
And ran raw IO and filesystem IO to the RAID array.
Chunk size 4k,64k was tested.
RAID rebuild , disk fail, resync tested.

File names:
This code is similar to ppc440spe . So I named the files as
drivers/dma/ppc4xx/adma1.c and drivers/dma/ppc4xx/adma1.h

Signed-off-by: Tirumala R Marri <[email protected]>
---
arch/powerpc/boot/dts/canyonlands.dts | 23 +-
drivers/dma/Kconfig | 10 +
drivers/dma/Makefile | 1 +
drivers/dma/ppc4xx/Makefile | 1 +
drivers/dma/ppc4xx/adma1.c | 4119 +++++++++++++++++++++++++++++++++
drivers/dma/ppc4xx/adma1.h | 192 ++
drivers/dma/ppc4xx/dma.h | 20 +-
7 files changed, 4362 insertions(+), 4 deletions(-)
create mode 100644 drivers/dma/ppc4xx/adma1.c
create mode 100644 drivers/dma/ppc4xx/adma1.h

diff --git a/arch/powerpc/boot/dts/canyonlands.dts b/arch/powerpc/boot/dts/canyonlands.dts
index cd56bb5..eb3ca8c 100644
--- a/arch/powerpc/boot/dts/canyonlands.dts
+++ b/arch/powerpc/boot/dts/canyonlands.dts
@@ -114,7 +114,10 @@
interrupt-parent = <&UIC1>;
interrupts = <11 1>;
};
-
+ MQ0: mq {
+ compatible = "ibm,mq-460ex";
+ dcr-reg = <0x040 0x020>;
+ };
plb {
compatible = "ibm,plb-460ex", "ibm,plb4";
#address-cells = <2>;
@@ -162,6 +165,24 @@
interrupt-parent = <&UIC2>;
interrupts = <0x1e 4>;
};
+ I2O: i2o@400100000 {
+ compatible = "ibm,i2o-460ex";
+ reg = <0x00000004 0x00100000 0x100>;
+ dcr-reg = <0x060 0x020>;
+ };
+ ADMA: adma {
+ compatible = "amcc,dma-460ex";
+ device_type = "dma";
+ reg = <0x00000004 0x00100200 0x100>;
+ interrupt-parent = <&ADMA>;
+ interrupts =<0 1 2>;
+ #interrupt-cells = <1>;
+ #address-cells = <0>;
+ #size-cells = <0>;
+ interrupt-map = </*FIFO need service */ 0 &UIC0 0x16 4
+ /*FIFO FULL */ 1 &UIC0 0x15 4
+ /*FIFO HSDMA err */ 2 &UIC1 0x16 4>;
+ };

POB0: opb {
compatible = "ibm,opb-460ex", "ibm,opb";
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9e01e96..77f9ea0 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -163,6 +163,16 @@ config TIMB_DMA
help
Enable support for the Timberdale FPGA DMA engine.

+config AMCC_PPC460EX_ADMA
+ tristate "AMCC PPC460Ex ADMA support"
+ depends on 460EX
+ select DMA_ENGINE
+ select ARCH_HAS_ASYNC_TX_FIND_CHANNEL
+ help
+ Enable support for the AMCC PPC460Ex RAID engines.
+ Also adds HW acceleration for memset and memcpy.
+ Enabling RAID-5/6 would also need HW key.
+
config ARCH_HAS_ASYNC_TX_FIND_CHANNEL
bool

diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 0fe5ebb..1d0ccfc 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -20,6 +20,7 @@ obj-$(CONFIG_TXX9_DMAC) += txx9dmac.o
obj-$(CONFIG_SH_DMAE) += shdma.o
obj-$(CONFIG_COH901318) += coh901318.o coh901318_lli.o
obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += ppc4xx/
+obj-$(CONFIG_AMCC_PPC460EX_ADMA) += ppc4xx/
obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
diff --git a/drivers/dma/ppc4xx/Makefile b/drivers/dma/ppc4xx/Makefile
index b3d259b..435a086 100644
--- a/drivers/dma/ppc4xx/Makefile
+++ b/drivers/dma/ppc4xx/Makefile
@@ -1 +1,2 @@
obj-$(CONFIG_AMCC_PPC440SPE_ADMA) += adma.o
+obj-$(CONFIG_AMCC_PPC460EX_ADMA) += adma1.o
diff --git a/drivers/dma/ppc4xx/adma1.c b/drivers/dma/ppc4xx/adma1.c
new file mode 100644
index 0000000..30c2229
--- /dev/null
+++ b/drivers/dma/ppc4xx/adma1.c
@@ -0,0 +1,4119 @@
+/*
+ * Copyright(c) 2010 Applied Micro (APM). All rights reserved.
+ *
+ * Author: Tirumala Reddy Marri [email protected]
+ *
+ * This driver follows Dan Williams and Yuri Tikhonovs implementations.
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+ *
+ * The full GNU General Public License is included in this distribution in the
+ * file called COPYING.
+ *
+ */
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/async_tx.h>
+#include <linux/delay.h>
+#include <linux/dma-mapping.h>
+#include <linux/spinlock.h>
+#include <linux/interrupt.h>
+#include <linux/platform_device.h>
+#include <linux/uaccess.h>
+#include <linux/of_platform.h>
+#include <linux/proc_fs.h>
+#include <asm/dcr.h>
+#include <asm/dcr-regs.h>
+#include "dma.h"
+#include "adma1.h"
+
+enum ppc_adma_init_code {
+ PPC_ADMA_INIT_OK = 0,
+ PPC_ADMA_INIT_MEMRES,
+ PPC_ADMA_INIT_MEMREG,
+ PPC_ADMA_INIT_ALLOC,
+ PPC_ADMA_INIT_COHERENT,
+ PPC_ADMA_INIT_CHANNEL,
+ PPC_ADMA_INIT_IRQ1,
+ PPC_ADMA_INIT_IRQ2,
+ PPC_ADMA_INIT_REGISTER
+};
+
+static char *ppc_adma_errors[] = {
+ [PPC_ADMA_INIT_OK] = "ok",
+ [PPC_ADMA_INIT_MEMRES] = "failed to get memory resource",
+ [PPC_ADMA_INIT_MEMREG] = "failed to request memory region",
+ [PPC_ADMA_INIT_ALLOC] = "failed to allocate memory for adev "
+ "structure",
+ [PPC_ADMA_INIT_COHERENT] = "failed to allocate coherent memory for "
+ "hardware descriptors",
+ [PPC_ADMA_INIT_CHANNEL] = "failed to allocate memory for channel",
+ [PPC_ADMA_INIT_IRQ1] = "failed to request first irq",
+ [PPC_ADMA_INIT_IRQ2] = "failed to request second irq",
+ [PPC_ADMA_INIT_REGISTER] = "failed to register dma async device",
+};
+
+static enum ppc_adma_init_code
+ppc460ex_adma_devices[PPC460EX_ADMA_ENGINES_NUM];
+
+struct ppc_dma_chan_ref {
+ struct dma_chan *chan;
+ struct list_head node;
+};
+
+/*
+ * The list of channels exported by ppc460ex ADMA
+ */
+struct list_head
+ppc460ex_adma_chan_list = LIST_HEAD_INIT(ppc460ex_adma_chan_list);
+
+/*
+ * Pointer to DMA0 CP/CS FIFO
+ */
+static void *ppc460ex_dma_fifo_buf;
+
+/*
+ * Pointers to last submitted to DMA0
+ */
+static struct ppc460ex_adma_desc_slot *chan_last_sub[1];
+static struct ppc460ex_adma_desc_slot *chan_first_cdb[1];
+
+/*
+ * Since RXOR operations use the common register (MQ0_CF2H) for setting-up
+ * the block size in transactions, then we do not allow to activate more than
+ * only one RXOR transactions simultaneously. So use this var to store
+ * the information about is RXOR currently active (PPC460EX_RXOR_RUN bit is
+ * set) or not (PPC460EX_RXOR_RUN is clear).
+ */
+static unsigned long ppc460ex_rxor_state;
+/*
+ * This array is used in data-check operations for storing a pattern
+ */
+static char ppc460ex_qword[16];
+
+
+/*
+ * These are used in enable & check routines
+ */
+static u32 ppc460ex_r6_enabled;
+static u32 ppc460ex_r5_enabled;
+
+static struct pc460ex_adma_chan *ppc460ex_r6_tchan;
+static struct pc460ex_adma_chan *ppc460ex_r5_tchan;
+static struct completion ppc460ex_r6_test_comp;
+static struct completion ppc460ex_r5_test_comp;
+static atomic_t ppc460ex_adma_err_irq_ref;
+static dcr_host_t ppc460ex_mq_dcr_host;
+static unsigned int ppc460ex_mq_dcr_len;
+
+static int ppc460ex_adma_alloc_chan_resources(struct dma_chan *chan);
+static struct ppc460ex_adma_desc_slot *ppc460ex_adma_alloc_slots(
+ struct ppc460ex_adma_chan *chan, int num_slots,
+ int slots_per_op);
+
+/******************************************************************************
+ * Command (Descriptor) Blocks low-level routines
+ ******************************************************************************/
+/*
+ * ppc460ex_desc_init_interrupt - initialize the descriptor for INTERRUPT
+ * pseudo operation
+ */
+static inline void ppc460ex_desc_init_interrupt(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan)
+{
+ memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+ /*
+ * NOP with interrupt
+ */
+ set_bit(PPC460EX_DESC_INT, &desc->flags);
+}
+/*
+ * ppc460ex_desc_init_pqzero_sum - initialize the descriptor
+ * for PQ_ZERO_SUM operation
+ */
+static void ppc460ex_desc_init_pqzero_sum(
+ struct ppc460ex_adma_desc_slot *desc,
+ int dst_cnt, int src_cnt)
+{
+ struct dma_cdb *hw_desc;
+ struct ppc460ex_adma_desc_slot *iter;
+ int i = 0;
+ u8 dopc = (dst_cnt == 2) ? DMA_CDB_OPC_MULTICAST :
+ DMA_CDB_OPC_MV_SG1_SG2;
+ /*
+ * Initialize starting from 2nd or 3rd descriptor dependent
+ * on dst_cnt. First one or two slots are for cloning P
+ * and/or Q to chan->pdest and/or chan->qdest as we have
+ * to preserve original P/Q.
+ */
+ iter = list_first_entry(&desc->group_list,
+ struct ppc460ex_adma_desc_slot, chain_node);
+ iter = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot, chain_node);
+
+ if (dst_cnt > 1) {
+ iter = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot, chain_node);
+ }
+ /*
+ * initialize each source descriptor in chain
+ */
+ list_for_each_entry_from(iter, &desc->group_list, chain_node) {
+ hw_desc = iter->hw_desc;
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->src_cnt = 0;
+ iter->dst_cnt = 0;
+
+ /* This is a ZERO_SUM operation:
+ * - <src_cnt> descriptors starting from 2nd or 3rd
+ * descriptor are for GF-XOR operations;
+ * - remaining <dst_cnt> descriptors are for checking the result
+ */
+ if (i++ < src_cnt)
+ /* MV_SG1_SG2 if only Q is being verified
+ * MULTICAST if both P and Q are being verified
+ */
+ hw_desc->opc = dopc;
+ else
+ /*
+ * DMA_CDB_OPC_DCHECK128 operation
+ */
+ hw_desc->opc = DMA_CDB_OPC_DCHECK128;
+
+ if (likely(!list_is_last(&iter->chain_node,
+ &desc->group_list))) {
+ /*
+ * set 'next' pointer
+ */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ } else {
+ /* this is the last descriptor.
+ * this slot will be pasted from ADMA level
+ * each time it wants to configure parameters
+ * of the transaction (src, dst, ...)
+ */
+ iter->hw_next = NULL;
+ /* always enable interrupt generation since we get
+ * the status of pqzero from the handler
+ */
+ set_bit(PPC460EX_DESC_INT, &iter->flags);
+ }
+ }
+ desc->src_cnt = src_cnt;
+ desc->dst_cnt = dst_cnt;
+}
+/*
+ * ppc460ex_desc_init_memcpy - initialize the descriptor for MEMCPY operation
+ */
+static inline void ppc460ex_desc_init_memcpy(
+ struct ppc460ex_adma_desc_slot *desc,
+ unsigned long flags)
+{
+ struct dma_cdb *hw_desc = desc->hw_desc;
+
+ memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+ desc->hw_next = NULL;
+ desc->src_cnt = 1;
+ desc->dst_cnt = 1;
+
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC460EX_DESC_INT, &desc->flags);
+ else
+ clear_bit(PPC460EX_DESC_INT, &desc->flags);
+
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+}
+
+/*
+ * ppc460ex_desc_init_memset - initialize the descriptor for MEMSET operation
+ */
+static inline void ppc460ex_desc_init_memset(
+ struct ppc460ex_adma_desc_slot *desc, int value,
+ unsigned long flags)
+{
+ struct dma_cdb *hw_desc = desc->hw_desc;
+
+ memset(desc->hw_desc, 0, sizeof(struct dma_cdb));
+ desc->hw_next = NULL;
+ desc->src_cnt = 1;
+ desc->dst_cnt = 1;
+
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC460EX_DESC_INT, &desc->flags);
+ else
+ clear_bit(PPC460EX_DESC_INT, &desc->flags);
+
+ hw_desc->sg1u = hw_desc->sg1l = cpu_to_le32((u32)value);
+ hw_desc->sg3u = hw_desc->sg3l = cpu_to_le32((u32)value);
+ hw_desc->opc = DMA_CDB_OPC_DFILL128;
+}
+/*
+ * ppc460ex_desc_assign_cookie - assign a cookie
+ */
+static dma_cookie_t ppc460ex_desc_assign_cookie(struct ppc460ex_adma_chan *chan,
+ struct ppc460ex_adma_desc_slot *desc)
+{
+ dma_cookie_t cookie = chan->common.cookie;
+ cookie++;
+ if (cookie < 0)
+ cookie = 1;
+ chan->common.cookie = desc->async_tx.cookie = cookie;
+ return cookie;
+}
+/*
+ * ppc460ex_desc_set_src_addr - set source address into the descriptor
+ */
+static inline void ppc460ex_desc_set_src_addr(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan, int src_idx,
+ dma_addr_t addrh, dma_addr_t addrl)
+{
+ struct dma_cdb *dma_hw_desc;
+ phys_addr_t addr64, tmplow, tmphi;
+
+ if (!addrh) {
+ addr64 = addrl;
+ tmphi = (addr64 >> 32);
+ tmplow = (addr64 & 0xFFFFFFFF);
+ } else {
+ tmphi = addrh;
+ tmplow = addrl;
+ }
+ dma_hw_desc = desc->hw_desc;
+ dma_hw_desc->sg1l = cpu_to_le32((u32)tmplow);
+ dma_hw_desc->sg1u = cpu_to_le32((u32)tmphi);
+}
+
+/*
+ * ppc460ex_desc_set_src_mult - set source address mult into the descriptor
+ */
+static inline void ppc460ex_desc_set_src_mult(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan, u32 mult_index, int sg_index,
+ unsigned char mult_value)
+{
+ struct dma_cdb *dma_hw_desc;
+ u32 *psgu;
+
+ dma_hw_desc = desc->hw_desc;
+
+ switch (sg_index) {
+ /*
+ * for RXOR operations set multiplier
+ * into source cued address
+ */
+ case DMA_CDB_SG_SRC:
+ psgu = &dma_hw_desc->sg1u;
+ break;
+ /*
+ * for WXOR operations set multiplier
+ * into destination cued address(es)
+ */
+ case DMA_CDB_SG_DST1:
+ psgu = &dma_hw_desc->sg2u;
+ break;
+ case DMA_CDB_SG_DST2:
+ psgu = &dma_hw_desc->sg3u;
+ break;
+ default:
+ BUG();
+ }
+
+ *psgu |= cpu_to_le32(mult_value << mult_index);
+}
+
+/*
+ * ppc460ex_desc_set_dest_addr - set destination address into the descriptor
+ */
+static inline void ppc460ex_desc_set_dest_addr(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan,
+ dma_addr_t addrh, dma_addr_t addrl,
+ u32 dst_idx)
+{
+ struct dma_cdb *dma_hw_desc;
+ phys_addr_t addr64, tmphi, tmplow;
+ u32 *psgu, *psgl;
+
+ if (!addrh) {
+ addr64 = addrl;
+ tmphi = (addr64 >> 32);
+ tmplow = (addr64 & 0xFFFFFFFF);
+ } else {
+ tmphi = addrh;
+ tmplow = addrl;
+ }
+ dma_hw_desc = desc->hw_desc;
+
+ psgu = dst_idx ? &dma_hw_desc->sg3u : &dma_hw_desc->sg2u;
+ psgl = dst_idx ? &dma_hw_desc->sg3l : &dma_hw_desc->sg2l;
+
+ *psgl = cpu_to_le32((u32)tmplow);
+ *psgu |= cpu_to_le32((u32)tmphi);
+}
+
+/*
+ * ppc460ex_desc_set_byte_count - set number of data bytes involved
+ * into the operation
+ */
+static inline void ppc460ex_desc_set_byte_count(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan, u32 byte_count)
+{
+ struct dma_cdb *dma_hw_desc;
+
+ dma_hw_desc = desc->hw_desc;
+ dma_hw_desc->cnt = cpu_to_le32(byte_count);
+}
+/*
+ * ppc460ex_desc_set_rxor_block_size - set RXOR block size
+ */
+static inline void ppc460ex_desc_set_rxor_block_size(u32 byte_count)
+{
+ /*
+ * assume that byte_count is aligned on the 512-boundary;
+ * thus write it directly to the register (bits 23:31 are
+ * reserved there).
+ */
+ mtdcr(DCRN_MQ0_CF2H, byte_count);
+}
+/*
+ * ppc460ex_desc_set_dcheck - set CHECK pattern
+ */
+static inline void ppc460ex_desc_set_dcheck(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan, u8 *qword)
+{
+ struct dma_cdb *dma_hw_desc;
+
+ dma_hw_desc = desc->hw_desc;
+ out_le32(&dma_hw_desc->sg3l, qword[0]);
+ out_le32(&dma_hw_desc->sg3u, qword[4]);
+ out_le32(&dma_hw_desc->sg2l, qword[8]);
+ out_le32(&dma_hw_desc->sg2u, qword[12]);
+}
+/*
+ * ppc460ex_desc_get_src_num - extract the number of source addresses from
+ * the descriptor
+ */
+static inline u32 ppc460ex_desc_get_src_num(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan)
+{
+ struct dma_cdb *dma_hw_desc;
+
+ dma_hw_desc = desc->hw_desc;
+
+ switch (dma_hw_desc->opc) {
+ case DMA_CDB_OPC_NO_OP:
+ case DMA_CDB_OPC_DFILL128:
+ return 0;
+ case DMA_CDB_OPC_DCHECK128:
+ return 1;
+ case DMA_CDB_OPC_MV_SG1_SG2:
+ case DMA_CDB_OPC_MULTICAST:
+ /*
+ * Only for RXOR operations we have more than
+ * one source
+ */
+ if (le32_to_cpu(dma_hw_desc->sg1u) &
+ DMA_CUED_XOR_WIN_MSK) {
+ /* RXOR op, there are 2 or 3 sources */
+ if (((le32_to_cpu(dma_hw_desc->sg1u) >>
+ DMA_CUED_REGION_OFF) &
+ DMA_CUED_REGION_MSK) == DMA_RXOR12) {
+ /* RXOR 1-2 */
+ return 2;
+ } else {
+ /* RXOR 1-2-3/1-2-4/1-2-5 */
+ return 3;
+ }
+ }
+ return 1;
+ default:
+ dev_dbg(chan->device->common.dev, "%s: unknown OPC 0x%02x\n",
+ __func__, dma_hw_desc->opc);
+ BUG();
+ }
+
+ return 0;
+}
+
+/*
+ * ppc460ex_desc_get_dst_num - get the number of destination addresses in
+ * this descriptor
+ */
+static inline u32 ppc460ex_desc_get_dst_num(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan)
+{
+ struct dma_cdb *dma_hw_desc;
+
+ /*
+ * May be 1 or 2 destinations
+ */
+ dma_hw_desc = desc->hw_desc;
+ switch (dma_hw_desc->opc) {
+ case DMA_CDB_OPC_NO_OP:
+ case DMA_CDB_OPC_DCHECK128:
+ return 0;
+ case DMA_CDB_OPC_MV_SG1_SG2:
+ case DMA_CDB_OPC_DFILL128:
+ return 1;
+ case DMA_CDB_OPC_MULTICAST:
+ return 2;
+ default:
+ dev_dbg(chan->device->common.dev, "%s: unknown OPC 0x%02x\n",
+ __func__, dma_hw_desc->opc);
+ BUG();
+ }
+ return 0;
+}
+/*
+ * ppc460ex_desc_get_src_addr - extract the source address from the descriptor
+ */
+static inline u32 ppc460ex_desc_get_src_addr(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan, int src_idx)
+{
+ struct dma_cdb *dma_hw_desc;
+ u32 sg11;
+
+ dma_hw_desc = desc->hw_desc;
+ /*
+ * May have 0, 1, 2, or 3 sources
+ */
+ switch (dma_hw_desc->opc) {
+ case DMA_CDB_OPC_NO_OP:
+ case DMA_CDB_OPC_DFILL128:
+ return 0;
+ case DMA_CDB_OPC_DCHECK128:
+ if (unlikely(src_idx)) {
+ dev_dbg(chan->device->common.dev,
+ "%s: try to get %d source for"
+ " DCHECK128\n", __func__, src_idx);
+ BUG();
+ }
+ return le32_to_cpu(dma_hw_desc->sg1l);
+ case DMA_CDB_OPC_MULTICAST:
+ case DMA_CDB_OPC_MV_SG1_SG2:
+ if (unlikely(src_idx > 2)) {
+ dev_dbg(chan->device->common.dev,
+ "%s: try to get %d source from"
+ " DMA descr\n", __func__, src_idx);
+ BUG();
+ }
+ if (src_idx) {
+ if (le32_to_cpu(dma_hw_desc->sg1u) &
+ DMA_CUED_XOR_WIN_MSK) {
+ u8 region;
+
+ if (src_idx == 1)
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ desc->unmap_len;
+
+ region = (le32_to_cpu(
+ dma_hw_desc->sg1u)) >>
+ DMA_CUED_REGION_OFF;
+
+ region &= DMA_CUED_REGION_MSK;
+ switch (region) {
+ case DMA_RXOR123:
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ (desc->unmap_len << 1);
+ case DMA_RXOR124:
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ (desc->unmap_len * 3);
+ case DMA_RXOR125:
+ return le32_to_cpu(
+ dma_hw_desc->sg1l) +
+ (desc->unmap_len << 2);
+ default:
+ dev_dbg(chan->device->common.dev,
+ "%s: try to"
+ " get src3 for region %02x"
+ "PPC460EX_DESC_RXOR12?\n",
+ __func__, region);
+ BUG();
+ }
+ } else {
+ dev_dbg(chan->device->common.dev,
+ "%s: try to get %d"
+ " source for non-cued descr\n",
+ __func__, src_idx);
+ BUG();
+ }
+ }
+ return le32_to_cpu(dma_hw_desc->sg1l);
+ default:
+ dev_dbg(chan->device->common.dev, "%s: unknown OPC 0x%02x\n",
+ __func__, dma_hw_desc->opc);
+ BUG();
+ }
+ sg11 = le32_to_cpu(dma_hw_desc->sg1l);
+ return sg11;
+}
+
+/*
+ * ppc460ex_desc_get_dest_addr - extract the destination address from the
+ * descriptor
+ */
+static inline u32 ppc460ex_desc_get_dest_addr(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan, int idx)
+{
+ struct dma_cdb *dma_hw_desc;
+
+ dma_hw_desc = desc->hw_desc;
+
+ if (likely(!idx))
+ return le32_to_cpu(dma_hw_desc->sg2l);
+ return le32_to_cpu(dma_hw_desc->sg3l);
+}
+/*
+ * ppc460ex_desc_get_link - get the address of the descriptor that
+ * follows this one
+ */
+static inline u32 ppc460ex_desc_get_link(struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan)
+{
+ if (!desc->hw_next)
+ return 0;
+
+ return desc->hw_next->phys;
+}
+/******************************************************************************
+ * ADMA channel low-level routines
+ ******************************************************************************/
+static inline u32 ppc460ex_chan_get_current_descriptor(
+ struct ppc460ex_adma_chan *chan);
+static inline void ppc460ex_chan_append(struct ppc460ex_adma_chan *chan);
+static void ppc460ex_dma_put_desc(struct ppc460ex_adma_chan *chan,
+ struct ppc460ex_adma_desc_slot *desc);
+
+/*
+ * ppc460ex_adma_device_clear_eot_status - interrupt ack to XOR or DMA engine
+ */
+static inline void ppc460ex_adma_device_clear_eot_status(
+ struct ppc460ex_adma_chan *chan)
+{
+ struct dma_regs *dma_reg;
+ u8 *p = chan->device->dma_desc_pool_virt;
+ struct dma_cdb *cdb;
+ u32 rv, i;
+
+ /*
+ * read FIFO to ack
+ */
+ dma_reg = (struct dma_regs *)chan->device->dma_reg;
+ while ((rv = in_le32(&dma_reg->csfpl))) {
+ i = rv & DMA_CDB_ADDR_MSK;
+ cdb = (struct dma_cdb *)&p[i -
+ (u32)chan->device->dma_desc_pool];
+
+ /* Clear opcode to ack. This is necessary for
+ * ZeroSum operations only
+ */
+ cdb->opc = 0;
+
+ if (test_bit(PPC460EX_RXOR_RUN,
+ &ppc460ex_rxor_state)) {
+ /* probably this is a completed RXOR op,
+ * get pointer to CDB using the fact that
+ * physical and virtual addresses of CDB
+ * in pools have the same offsets
+ */
+ if (le32_to_cpu(cdb->sg1u) &
+ DMA_CUED_XOR_BASE) {
+ /* this is a RXOR */
+ clear_bit(PPC460EX_RXOR_RUN,
+ &ppc460ex_rxor_state);
+ }
+ }
+
+ if (rv & DMA_CDB_STATUS_MSK) {
+ /*
+ * ZeroSum check failed
+ */
+ struct ppc460ex_adma_desc_slot *iter;
+ dma_addr_t phys = rv & ~DMA_CDB_MSK;
+
+ /*
+ * Update the status of corresponding
+ * descriptor.
+ */
+ list_for_each_entry(iter, &chan->chain,
+ chain_node) {
+ if (iter->phys == phys)
+ break;
+ }
+ /*
+ * if cannot find the corresponding
+ * slot it's a bug
+ */
+ BUG_ON(&iter->chain_node == &chan->chain);
+
+ if (iter->xor_check_result)
+ *iter->xor_check_result |=
+ rv & DMA_CDB_STATUS_MSK;
+ }
+ }
+
+ rv = in_le32(&dma_reg->dsts);
+ if (rv) {
+ dev_dbg(chan->device->common.dev,
+ "DMA%d err status: 0x%x\n", chan->device->id,
+ rv);
+ /*
+ * write back to clear
+ */
+ out_le32(&dma_reg->dsts, rv);
+ }
+}
+/*
+ * ppc460ex_chan_is_busy - get the channel status
+ */
+static inline int ppc460ex_chan_is_busy(struct ppc460ex_adma_chan *chan)
+{
+ int busy = 0;
+ struct dma_regs *dma_reg;
+
+ dma_reg = (struct dma_regs *)chan->device->dma_reg;
+ /*
+ * if command FIFO's head and tail pointers are equal and
+ * status tail is the same as command, then channel is free
+ */
+ if (dma_reg->cpfhp != dma_reg->cpftp ||
+ dma_reg->cpftp != dma_reg->csftp)
+ busy = 1;
+
+ return busy;
+}
+/*
+ * ppc460ex_chan_append - update the h/w chain in the channel
+ */
+void ppc460ex_chan_append(struct ppc460ex_adma_chan *chan)
+{
+ struct dma_regs *dma_reg;
+ struct ppc460ex_adma_desc_slot *iter;
+ u32 cur_desc;
+ unsigned long flags;
+
+ local_irq_save(flags);
+ dma_reg = (struct dma_regs *)chan->device->dma_reg;
+ cur_desc = ppc460ex_chan_get_current_descriptor(chan);
+
+ if (likely(cur_desc)) {
+ iter = chan_last_sub[chan->device->id];
+ BUG_ON(!iter);
+ } else {
+ /*
+ * first peer
+ */
+ iter = chan_first_cdb[chan->device->id];
+ BUG_ON(!iter);
+ ppc460ex_dma_put_desc(chan, iter);
+ chan->hw_chain_inited = 1;
+ }
+ /*
+ * is there something new to append
+ */
+ if (!iter->hw_next)
+ goto out;
+ /*
+ * flush descriptors from the s/w queue to fifo
+ */
+ list_for_each_entry_continue(iter, &chan->chain, chain_node) {
+ ppc460ex_dma_put_desc(chan, iter);
+ if (!iter->hw_next)
+ break;
+ }
+out:
+ local_irq_restore(flags);
+}
+
+/*
+ * ppc460ex_chan_get_current_descriptor - get the currently executed descriptor
+ */
+u32 ppc460ex_chan_get_current_descriptor(struct ppc460ex_adma_chan *chan)
+{
+ struct dma_regs *dma_reg;
+
+ if (unlikely(!chan->hw_chain_inited))
+ /*
+ * h/w descriptor chain is not initialized yet
+ */
+ return 0;
+
+ dma_reg = (struct dma_regs *)chan->device->dma_reg;
+ return (le32_to_cpu(dma_reg->acpl)) & (~DMA_CDB_MSK);
+}
+/*
+ * ppc460ex_dma_put_desc - put DMA0,1 descriptor to FIFO
+ */
+static void ppc460ex_dma_put_desc(struct ppc460ex_adma_chan *chan,
+ struct ppc460ex_adma_desc_slot *desc)
+{
+ u32 pcdb;
+ struct dma_regs *dma_reg =
+ dma_reg = (struct dma_regs *)chan->device->dma_reg;
+
+ pcdb = desc->phys;
+ if (!test_bit(PPC460EX_DESC_INT, &desc->flags))
+ pcdb |= DMA_CDB_NO_INT;
+
+ chan_last_sub[chan->device->id] = desc;
+
+ out_le32(&dma_reg->cpfpl, pcdb);
+
+}
+/******************************************************************************
+ * ADMA device level
+ *****************************************************************************/
+
+static dma_cookie_t ppc460ex_adma_tx_submit(
+ struct dma_async_tx_descriptor *tx);
+static inline void ppc460ex_adma_set_dest(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index);
+
+static void ppc460ex_adma_pqzero_sum_set_src_mult(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ unsigned char mult, int index, int dst_pos);
+static void ppc460ex_adma_pq_set_src_mult(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ unsigned char mult, int index, int dst_pos);
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_mq_xor(
+ struct dma_chan *chan, dma_addr_t dst,
+ dma_addr_t *src, unsigned int src_cnt,
+ size_t len, unsigned long flags);
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_p(
+ struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
+ unsigned int src_cnt, unsigned char *scf,
+ size_t len, unsigned long flags);
+static struct ppc460ex_adma_desc_slot *ppc460ex_dma01_prep_mult(
+ struct ppc460ex_adma_chan *ppc460ex_chan,
+ dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+ const unsigned char *scf, size_t len, unsigned long flags);
+static struct ppc460ex_adma_desc_slot *ppc460ex_dma_prep_pq(
+ struct ppc460ex_adma_chan *ppc460ex_chan,
+ dma_addr_t *dst, unsigned int dst_cnt,
+ dma_addr_t *src, unsigned int src_cnt, unsigned char *scf,
+ size_t len, unsigned long flags);
+
+static void ppc460ex_adma_pqxor_set_src_mult(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ unsigned char mult, int index, int dst_pos);
+static void ppc460ex_adma_pqxor_set_src(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index);
+
+static void ppc460ex_adma_pqxor_set_dest(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t *addrs, unsigned long flags);
+static void ppc460ex_adma_pqzero_sum_set_dest(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t paddr, dma_addr_t qaddr);
+static void ppc460ex_desc_init_pq(struct ppc460ex_adma_desc_slot *desc,
+ int dst_cnt, int src_cnt, unsigned long flags,
+ unsigned long op);
+static inline void ppc460ex_adma_memcpy_xor_set_src(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index);
+#define DMA_CTRL_FLAGS_LAST DMA_PREP_FENCE
+#define DMA_PREP_ZERO_P (DMA_CTRL_FLAGS_LAST << 1)
+#define DMA_PREP_ZERO_Q (DMA_PREP_ZERO_P << 1)
+/*
+ * ppc460ex_can_rxor - check if the operands may be processed with RXOR
+ */
+int ppc460ex_can_rxor(struct page **srcs, int src_cnt, size_t len)
+{
+ int i, order = 0, state = 0;
+
+ if (unlikely(!(src_cnt > 1)))
+ return 0;
+
+ for (i = 1; i < src_cnt; i++) {
+ char *cur_addr = page_address(srcs[i]);
+ char *old_addr = page_address(srcs[i-1]);
+ switch (state) {
+ case 0:
+ if (cur_addr == old_addr + len) {
+ /* direct RXOR */
+ order = 1;
+ state = 1;
+ } else
+ if (old_addr == cur_addr + len) {
+ /* reverse RXOR */
+ order = -1;
+ state = 1;
+ } else
+ goto out;
+ break;
+ case 1:
+ if ((i == src_cnt-2) ||
+ (order == -1 && cur_addr != old_addr - len)) {
+ order = 0;
+ state = 0;
+ } else
+ if ((cur_addr == old_addr + len*order) ||
+ (cur_addr == old_addr + 2*len) ||
+ (cur_addr == old_addr + 3*len)) {
+ state = 2;
+ } else {
+ order = 0;
+ state = 0;
+ }
+ break;
+ case 2:
+ order = 0;
+ state = 0;
+ break;
+ }
+ }
+
+out:
+ if (state == 1 || state == 2)
+ return 1;
+
+ return 0;
+}
+/*
+ * ppc460ex_adma_device_estimate - estimate the efficiency of processing
+ * the operation given on this channel. It's assumed that 'chan' is
+ * capable to process 'cap' type of operation.
+ * @chan: channel to use
+ * @cap: type of transaction
+ * @dst_lst: array of destination pointers
+ * @src_lst: array of source pointers
+ * @src_cnt: number of source operands
+ * @src_sz: size of each source operand
+ */
+int ppc460ex_adma_estimate(struct dma_chan *chan,
+ enum dma_transaction_type cap, struct page **dst_lst, int dst_cnt,
+ struct page **src_lst, int src_cnt, size_t src_sz)
+{
+ int ef = 1;
+
+ if (cap == DMA_PQ || cap == DMA_PQ_VAL) {
+ /*
+ * If RAID-6 capabilities were not activated don't try
+ * to use them
+ */
+ if (unlikely(!ppc460ex_r6_enabled))
+ return -1;
+ }
+ if (cap == DMA_XOR) {
+ if (unlikely(!ppc460ex_r5_enabled))
+ return -1;
+ }
+ /* in the current implementation of ppc460ex ADMA driver it
+ * makes sense to pick out only pqxor case, because it may be
+ * processed:
+ * (1) either using Biskup method on DMA2;
+ * (2) or on DMA0/1.
+ * Thus we give a favour to (1) if the sources are suitable;
+ * else let it be processed on one of the DMA0/1 engines.
+ */
+ if (cap == DMA_PQ) {
+ if (ppc460ex_can_rxor(src_lst, src_cnt, src_sz))
+ ef = 3; /* override (dma0 + idle) */
+ else
+ ef = 0; /* can't process on DMA2 if !rxor */
+ }
+
+ /*
+ * channel idleness increases the priority
+ */
+ if (likely(ef) &&
+ !ppc460ex_chan_is_busy(to_ppc460ex_adma_chan(chan)))
+ ef++;
+
+ return ef;
+}
+struct dma_chan *
+ppc460ex_async_tx_find_best_channel(enum dma_transaction_type cap,
+ struct page **dst_lst, int dst_cnt, struct page **src_lst,
+ int src_cnt, size_t src_sz)
+{
+ struct dma_chan *best_chan = NULL;
+ struct ppc_dma_chan_ref *ref;
+ int best_rank = -1;
+
+ if (unlikely(!src_sz))
+ return NULL;
+ if (src_sz > PAGE_SIZE) {
+ switch (cap) {
+ case DMA_PQ:
+ if (src_cnt == 1 && dst_lst[1] == src_lst[0])
+ return NULL;
+ if (src_cnt == 2 && dst_lst[1] == src_lst[1])
+ return NULL;
+ break;
+ default:
+ break;
+ }
+ }
+ list_for_each_entry(ref, &ppc460ex_adma_chan_list, node) {
+ if (dma_has_cap(cap, ref->chan->device->cap_mask)) {
+ int rank;
+
+ rank = ppc460ex_adma_estimate(ref->chan, cap, dst_lst,
+ dst_cnt, src_lst, src_cnt, src_sz);
+ if (rank > best_rank) {
+ best_rank = rank;
+ best_chan = ref->chan;
+ }
+ }
+ }
+ return best_chan;
+}
+/*
+ * ppc460ex_get_group_entry - get group entry with index idx
+ * @tdesc: is the last allocated slot in the group.
+ */
+struct ppc460ex_adma_desc_slot *
+ppc460ex_get_group_entry(struct ppc460ex_adma_desc_slot *tdesc,
+ u32 entry_idx)
+{
+ struct ppc460ex_adma_desc_slot *iter = tdesc->group_head;
+ int i = 0;
+
+ if (entry_idx < 0 || entry_idx >= (tdesc->src_cnt + tdesc->dst_cnt)) {
+ pr_debug("%s: entry_idx %d, src_cnt %d, dst_cnt %d\n",
+ __func__, entry_idx, tdesc->src_cnt,
+ tdesc->dst_cnt);
+ BUG();
+ }
+ list_for_each_entry(iter, &tdesc->group_list, chain_node) {
+ if (i++ == entry_idx)
+ break;
+ }
+ return iter;
+}
+/*
+ * ppc460ex_adma_free_slots - flags descriptor slots for reuse
+ * @slot: Slot to free
+ * Caller must hold &ppc460ex_chan->lock while calling this function
+ */
+void ppc460ex_adma_free_slots(struct ppc460ex_adma_desc_slot *slot,
+ struct ppc460ex_adma_chan *chan)
+{
+ int stride = slot->slots_per_op;
+
+ while (stride--) {
+ slot->slots_per_op = 0;
+ slot = list_entry(slot->slot_node.next,
+ struct ppc460ex_adma_desc_slot,
+ slot_node);
+ }
+}
+
+void ppc460ex_adma_unmap(struct ppc460ex_adma_chan *chan,
+ struct ppc460ex_adma_desc_slot *desc)
+{
+ u32 src_cnt, dst_cnt;
+ dma_addr_t addr;
+ /*
+ * get the number of sources & destination
+ * included in this descriptor and unmap
+ * them all
+ */
+ src_cnt = ppc460ex_desc_get_src_num(desc, chan);
+ dst_cnt = ppc460ex_desc_get_dst_num(desc, chan);
+
+ /*
+ * unmap destinations
+ */
+ if (!(desc->async_tx.flags & DMA_COMPL_SKIP_DEST_UNMAP)) {
+ while (dst_cnt--) {
+ addr = ppc460ex_desc_get_dest_addr(
+ desc, chan, dst_cnt);
+ dma_unmap_page(chan->device->dev,
+ addr, desc->unmap_len,
+ DMA_FROM_DEVICE);
+ }
+ }
+
+ /*
+ * unmap sources
+ */
+ if (!(desc->async_tx.flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
+ while (src_cnt--) {
+ addr = ppc460ex_desc_get_src_addr(
+ desc, chan, src_cnt);
+ dma_unmap_page(chan->device->dev,
+ addr, desc->unmap_len,
+ DMA_TO_DEVICE);
+ }
+ }
+
+}
+/*
+ * ppc460ex_adma_run_tx_complete_actions - call functions to be called
+ * upon complete
+ */
+dma_cookie_t ppc460ex_adma_run_tx_complete_actions(
+ struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan,
+ dma_cookie_t cookie)
+{
+ int i;
+
+ BUG_ON(desc->async_tx.cookie < 0);
+ if (desc->async_tx.cookie > 0) {
+ cookie = desc->async_tx.cookie;
+ desc->async_tx.cookie = 0;
+
+ /*
+ * call the callback (must not sleep or submit new
+ * operations to this channel)
+ */
+ if (desc->async_tx.callback)
+ desc->async_tx.callback(
+ desc->async_tx.callback_param);
+
+ /* unmap dma addresses
+ * (unmap_single vs unmap_page?)
+ *
+ * actually, ppc's dma_unmap_page() functions are empty, so
+ * the following code is just for the sake of completeness
+ */
+ if (chan && chan->needs_unmap && desc->group_head &&
+ desc->unmap_len) {
+ struct ppc460ex_adma_desc_slot *unmap =
+ desc->group_head;
+ /*
+ * assume 1 slot per op always
+ */
+ u32 slot_count = unmap->slot_cnt;
+
+ /*
+ * Run through the group list and unmap addresses
+ */
+ for (i = 0; i < slot_count; i++) {
+ BUG_ON(!unmap);
+ ppc460ex_adma_unmap(chan, unmap);
+ unmap = unmap->hw_next;
+ }
+ }
+ }
+
+ /*
+ * run dependent operations
+ */
+ dma_run_dependencies(&desc->async_tx);
+
+ return cookie;
+}
+/*
+ * ppc460ex_adma_clean_slot - clean up CDB slot (if ack is set)
+ */
+int ppc460ex_adma_clean_slot(struct ppc460ex_adma_desc_slot *desc,
+ struct ppc460ex_adma_chan *chan)
+{
+ struct dma_cdb *cdb ;
+ /*
+ * the client is allowed to attach dependent operations
+ * until 'ack' is set
+ */
+ if (!async_tx_test_ack(&desc->async_tx))
+ return 0;
+
+ /*
+ * leave the last descriptor in the chain
+ * so we can append to it
+ */
+ if (list_is_last(&desc->chain_node, &chan->chain) ||
+ desc->phys == ppc460ex_chan_get_current_descriptor(chan))
+ return 1;
+
+ /* our DMA interrupt handler clears opc field of
+ * each processed descriptor. For all types of
+ * operations except for ZeroSum we do not actually
+ * need ack from the interrupt handler. ZeroSum is a
+ * gtcial case since the result of this operation
+ * is available from the handler only, so if we see
+ * such type of descriptor (which is unprocessed yet)
+ * then leave it in chain.
+ */
+ cdb = desc->hw_desc;
+ if (cdb->opc == DMA_CDB_OPC_DCHECK128)
+ return 1;
+
+ dev_dbg(chan->device->common.dev, "\tfree slot %lx: %d stride: %d\n",
+ (ulong)desc->phys, desc->idx, desc->slots_per_op);
+
+ list_del(&desc->chain_node);
+ ppc460ex_adma_free_slots(desc, chan);
+ return 0;
+}
+/*
+ *__ppc460ex_adma_slot_cleanup - this is the common clean-up routine
+ *which runs through the channel CDBs list until reach the descriptor
+ *currently processed. When routine determines that all CDBs of group
+ *are completed then corresponding callbacks (if any) are called and slots
+ *are freed.
+ */
+void __ppc460ex_adma_slot_cleanup(struct ppc460ex_adma_chan *chan)
+{
+ struct ppc460ex_adma_desc_slot *iter, *_iter, *group_start = NULL;
+ dma_cookie_t cookie = 0;
+ u32 current_desc = ppc460ex_chan_get_current_descriptor(chan);
+ int busy = ppc460ex_chan_is_busy(chan);
+ int seen_current = 0, slot_cnt = 0, slots_per_op = 0;
+
+ dev_dbg(chan->device->common.dev, "ppc460ex adma%d: %s\n",
+ chan->device->id, __func__);
+
+ if (!current_desc) {
+ /*
+ * There were no transactions yet, so
+ * nothing to clean
+ */
+ return;
+ }
+
+ /*
+ * free completed slots from the chain starting with
+ * the oldest descriptor
+ */
+ list_for_each_entry_safe(iter, _iter, &chan->chain,
+ chain_node) {
+ dev_dbg(chan->device->common.dev, "\tcookie: %d slot: %d "
+ "busy: %d this_desc: %#x next_desc: %#x cur: %#x ack: %d\n",
+ iter->async_tx.cookie, iter->idx, busy, (u32)iter->phys,
+ ppc460ex_desc_get_link(iter, chan), current_desc,
+ async_tx_test_ack(&iter->async_tx));
+ prefetch(_iter);
+ prefetch(&_iter->async_tx);
+
+ /*
+ * do not advance past the current descriptor loaded into the
+ * hardware channel,subsequent descriptors are either in process
+ * or have not been submitted
+ */
+ if (seen_current)
+ break;
+
+ /*
+ * stop the search if we reach the current descriptor and the
+ * channel is busy, or if it appears that the current descriptor
+ * needs to be re-read (i.e. has been appended to)
+ */
+ if (iter->phys == current_desc) {
+ BUG_ON(seen_current++);
+ if (busy || ppc460ex_desc_get_link(iter, chan)) {
+ /* not all descriptors of the group have
+ * been completed; exit.
+ */
+ break;
+ }
+ }
+
+ /*
+ * detect the start of a group transaction
+ */
+ if (!slot_cnt && !slots_per_op) {
+ slot_cnt = iter->slot_cnt;
+ slots_per_op = iter->slots_per_op;
+ if (slot_cnt <= slots_per_op) {
+ slot_cnt = 0;
+ slots_per_op = 0;
+ }
+ }
+
+ if (slot_cnt) {
+ if (!group_start)
+ group_start = iter;
+ slot_cnt -= slots_per_op;
+ }
+
+ /*
+ * all the members of a group are complete
+ */
+ if (slots_per_op != 0 && slot_cnt == 0) {
+ struct ppc460ex_adma_desc_slot *grp_iter, *_grp_iter;
+ int end_of_chain = 0;
+
+ /*
+ * clean up the group
+ */
+ slot_cnt = group_start->slot_cnt;
+ grp_iter = group_start;
+ list_for_each_entry_safe_from(grp_iter, _grp_iter,
+ &chan->chain, chain_node) {
+
+ cookie = ppc460ex_adma_run_tx_complete_actions(
+ grp_iter, chan, cookie);
+
+ slot_cnt -= slots_per_op;
+ end_of_chain = ppc460ex_adma_clean_slot(
+ grp_iter, chan);
+ if (end_of_chain && slot_cnt) {
+ /*
+ * Should wait for ZeroSum complete
+ */
+ if (cookie > 0)
+ chan->completed_cookie = cookie;
+ return;
+ }
+
+ if (slot_cnt == 0 || end_of_chain)
+ break;
+ }
+
+ /*
+ * the group should be complete at this point
+ */
+ BUG_ON(slot_cnt);
+
+ slots_per_op = 0;
+ group_start = NULL;
+ if (end_of_chain)
+ break;
+ else
+ continue;
+ /*
+ * wait for group completion
+ */
+ } else if (slots_per_op)
+ continue;
+
+ cookie = ppc460ex_adma_run_tx_complete_actions(iter, chan,
+ cookie);
+
+ if (ppc460ex_adma_clean_slot(iter, chan))
+ break;
+ }
+
+ BUG_ON(!seen_current);
+}
+/*
+ * ppc460ex_adma_tasklet - clean up watch-dog initiator
+ */
+void ppc460ex_adma_tasklet(unsigned long data)
+{
+ struct ppc460ex_adma_chan *chan = (struct ppc460ex_adma_chan *) data;
+ spin_lock_nested(&chan->lock, SINGLE_DEPTH_NESTING);
+ __ppc460ex_adma_slot_cleanup(chan);
+ spin_unlock(&chan->lock);
+}
+
+/*
+ * ppc460ex_adma_slot_cleanup - clean up scheduled initiator
+ */
+void ppc460ex_adma_slot_cleanup(struct ppc460ex_adma_chan *chan)
+{
+ spin_lock_bh(&chan->lock);
+ __ppc460ex_adma_slot_cleanup(chan);
+ spin_unlock_bh(&chan->lock);
+}
+/*
+ * ppc460ex_adma_alloc_slots - allocate free slots (if any)
+ */
+struct ppc460ex_adma_desc_slot *ppc460ex_adma_alloc_slots(
+ struct ppc460ex_adma_chan *chan, int num_slots,
+ int slots_per_op)
+{
+ struct ppc460ex_adma_desc_slot *iter = NULL, *_iter;
+ struct ppc460ex_adma_desc_slot *alloc_start = NULL;
+ struct list_head chain = LIST_HEAD_INIT(chain);
+ int slots_found, retry = 0;
+
+ BUG_ON(!num_slots || !slots_per_op);
+ /* start search from the last allocated descrtiptor
+ * if a contiguous allocation can not be found start searching
+ * from the beginning of the list
+ */
+retry:
+ slots_found = 0;
+ if (retry == 0)
+ iter = chan->last_used;
+ else
+ iter = list_entry(&chan->all_slots,
+ struct ppc460ex_adma_desc_slot,
+ slot_node);
+ list_for_each_entry_safe_continue(iter, _iter, &chan->all_slots,
+ slot_node) {
+ prefetch(_iter);
+ prefetch(&_iter->async_tx);
+ if (iter->slots_per_op) {
+ slots_found = 0;
+ continue;
+ }
+
+ /*
+ * start the allocation if the slot is correctly aligned
+ */
+ if (!slots_found++)
+ alloc_start = iter;
+ if (slots_found == num_slots) {
+ struct ppc460ex_adma_desc_slot *alloc_tail = NULL;
+ struct ppc460ex_adma_desc_slot *last_used = NULL;
+
+ iter = alloc_start;
+ while (num_slots) {
+ int i;
+ /*
+ * pre-ack all but the last descriptor
+ */
+ if (num_slots != slots_per_op)
+ async_tx_ack(&iter->async_tx);
+
+ list_add_tail(&iter->chain_node, &chain);
+ alloc_tail = iter;
+ iter->async_tx.cookie = 0;
+ iter->hw_next = NULL;
+ iter->flags = 0;
+ iter->slot_cnt = num_slots;
+ iter->xor_check_result = NULL;
+ for (i = 0; i < slots_per_op; i++) {
+ iter->slots_per_op = slots_per_op - i;
+ last_used = iter;
+ iter = list_entry(iter->slot_node.next,
+ struct ppc460ex_adma_desc_slot,
+ slot_node);
+ }
+ num_slots -= slots_per_op;
+ }
+ alloc_tail->group_head = alloc_start;
+ alloc_tail->async_tx.cookie = -EBUSY;
+ list_splice(&chain, &alloc_tail->group_list);
+ chan->last_used = last_used;
+ return alloc_tail;
+ }
+ }
+ if (!retry++)
+ goto retry;
+
+ /*
+ * try to free some slots if the allocation fails
+ */
+ tasklet_schedule(&chan->irq_tasklet);
+ return NULL;
+}
+/*
+ * ppc460ex_adma_alloc_chan_resources - allocate pools for CDB slots
+ */
+int ppc460ex_adma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan;
+ struct ppc460ex_adma_desc_slot *slot = NULL;
+ char *hw_desc;
+ int i, db_sz;
+ int init;
+
+ ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ init = ppc460ex_chan->slots_allocated ? 0 : 1;
+ chan->chan_id = ppc460ex_chan->device->id;
+
+ /*
+ * Allocate descriptor slots
+ */
+ i = ppc460ex_chan->slots_allocated;
+ db_sz = sizeof(struct dma_cdb);
+
+ for (; i < (ppc460ex_chan->device->pool_size / db_sz); i++) {
+ slot = kzalloc(sizeof(struct ppc460ex_adma_desc_slot),
+ GFP_KERNEL);
+ if (!slot) {
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "EX/GT ADMA Channel only initialized"
+ " %d descriptor slots", i--);
+ break;
+ }
+ hw_desc = (char *) ppc460ex_chan->device->dma_desc_pool_virt;
+ slot->hw_desc = (void *) &hw_desc[i * db_sz];
+ dma_async_tx_descriptor_init(&slot->async_tx, chan);
+ slot->async_tx.tx_submit = ppc460ex_adma_tx_submit;
+ INIT_LIST_HEAD(&slot->chain_node);
+ INIT_LIST_HEAD(&slot->slot_node);
+ INIT_LIST_HEAD(&slot->group_list);
+ slot->phys = ppc460ex_chan->device->dma_desc_pool + i * db_sz;
+ slot->idx = i;
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+ ppc460ex_chan->slots_allocated++;
+ list_add_tail(&slot->slot_node, &ppc460ex_chan->all_slots);
+ spin_unlock_bh(&ppc460ex_chan->lock);
+ }
+ if (i && !ppc460ex_chan->last_used) {
+ ppc460ex_chan->last_used =
+ list_entry(ppc460ex_chan->all_slots.next,
+ struct ppc460ex_adma_desc_slot,
+ slot_node);
+ }
+
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: allocated %d descriptor slots\n",
+ ppc460ex_chan->device->id, i);
+
+ /*
+ * initialize the channel and the chain with a null operation
+ */
+ if (init) {
+ ppc460ex_chan->hw_chain_inited = 0;
+ /*
+ * Use WXOR for self-testing
+ */
+ if (!ppc460ex_r6_tchan)
+ ppc460ex_r6_tchan = ppc460ex_chan;
+ ppc460ex_chan->needs_unmap = 1;
+ }
+ return (i > 0) ? i : -ENOMEM;
+}
+/*
+ * ppc460ex_desc_set_link - set the address of descriptor following this
+ * descriptor in chain
+ */
+static inline void ppc460ex_desc_set_link(struct ppc460ex_adma_chan *chan,
+ struct ppc460ex_adma_desc_slot *prev_desc,
+ struct ppc460ex_adma_desc_slot *next_desc)
+{
+ unsigned long flags;
+
+ if (unlikely(!prev_desc || !next_desc ||
+ (prev_desc->hw_next && prev_desc->hw_next != next_desc))) {
+ /* If previous next is overwritten something is wrong.
+ * though we may refetch from append to initiate list
+ * processing; in this case - it's ok.
+ */
+ dev_dbg(chan->device->common.dev,
+ "%s: prev_desc=0x%p; next_desc=0x%p; "
+ "prev->hw_next=0x%p\n", __func__, prev_desc,
+ next_desc, prev_desc ? prev_desc->hw_next : 0);
+ BUG();
+ }
+
+ local_irq_save(flags);
+
+ /*
+ * do s/w chaining both for DMA and XOR descriptors
+ */
+ prev_desc->hw_next = next_desc;
+ local_irq_restore(flags);
+}
+/*
+ * ppc460ex_adma_check_threshold - append CDBs to h/w chain if threshold
+ * has been achieved
+ */
+static void ppc460ex_adma_check_threshold(struct ppc460ex_adma_chan *chan)
+{
+ dev_dbg(chan->device->common.dev, "ppc460ex adma%d: pending: %d\n",
+ chan->device->id, chan->pending);
+
+ if (chan->pending >= PPC460EX_ADMA_THRESHOLD) {
+ chan->pending = 0;
+ ppc460ex_chan_append(chan);
+ }
+}
+/*
+ * ppc460ex_adma_tx_submit - submit new descriptor group to the channel
+ * (it's not necessary that descriptors will be submitted to the h/w
+ * chains too right now)
+ */
+static dma_cookie_t ppc460ex_adma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct ppc460ex_adma_desc_slot *sw_desc = tx_to_ppc460ex_adma_slot(tx);
+ struct ppc460ex_adma_chan *chan = to_ppc460ex_adma_chan(tx->chan);
+ struct ppc460ex_adma_desc_slot *group_start, *old_chain_tail;
+ int slot_cnt;
+ int slots_per_op;
+ dma_cookie_t cookie;
+
+ group_start = sw_desc->group_head;
+ slot_cnt = group_start->slot_cnt;
+ slots_per_op = group_start->slots_per_op;
+
+ spin_lock_bh(&chan->lock);
+
+ cookie = ppc460ex_desc_assign_cookie(chan, sw_desc);
+
+ if (unlikely(list_empty(&chan->chain))) {
+ /*
+ *first peer
+ */
+ list_splice_init(&sw_desc->group_list, &chan->chain);
+ chan_first_cdb[chan->device->id] = group_start;
+ } else {
+ /*
+ * isn't first peer, bind CDBs to chain
+ */
+ old_chain_tail = list_entry(chan->chain.prev,
+ struct ppc460ex_adma_desc_slot, chain_node);
+ list_splice_init(&sw_desc->group_list,
+ &old_chain_tail->chain_node);
+ /*
+ * fix up the hardware chain
+ */
+ ppc460ex_desc_set_link(chan, old_chain_tail, group_start);
+ }
+
+ /*
+ * increment the pending count by the number of operations
+ */
+ chan->pending += slot_cnt / slots_per_op;
+ ppc460ex_adma_check_threshold(chan);
+ spin_unlock_bh(&chan->lock);
+
+ dev_dbg(chan->device->common.dev,
+ "ppc460ex adma%d: %s cookie: %d slot: %d tx %p\n",
+ chan->device->id, __func__,
+ sw_desc->async_tx.cookie, sw_desc->idx, sw_desc);
+ return cookie;
+}
+/*
+ * ppc460ex_adma_prep_dma_interrupt - prepare CDB for a pseudo DMA operation
+ */
+struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_interrupt(
+ struct dma_chan *chan, unsigned long flags)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ struct ppc460ex_adma_desc_slot *sw_desc, *group_start;
+ int slot_cnt, slots_per_op;
+
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: %s\n", ppc460ex_chan->device->id,
+ __func__);
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+ slot_cnt = slots_per_op = 1;
+ sw_desc = ppc460ex_adma_alloc_slots(ppc460ex_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ ppc460ex_desc_init_interrupt(group_start, ppc460ex_chan);
+ group_start->unmap_len = 0;
+ sw_desc->async_tx.flags = flags;
+ }
+ spin_unlock_bh(&ppc460ex_chan->lock);
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+/*
+ * ppc460ex_adma_prep_dma_pqzero_sum - prepare CDB group for
+ * a PQ_ZERO_SUM operation
+ */
+struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_pqzero_sum(
+ struct dma_chan *chan, dma_addr_t *pq, dma_addr_t *src,
+ unsigned int src_cnt, const unsigned char *scf, size_t len,
+ enum sum_check_flags *pqres, unsigned long flags)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan;
+ struct ppc460ex_adma_desc_slot *sw_desc, *iter;
+ dma_addr_t pdest, qdest;
+ int slot_cnt, slots_per_op, idst, dst_cnt;
+
+ ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+
+ if (flags & DMA_PREP_PQ_DISABLE_P)
+ pdest = 0;
+ else
+ pdest = pq[0];
+
+ if (flags & DMA_PREP_PQ_DISABLE_Q)
+ qdest = 0;
+ else
+ qdest = pq[1];
+
+
+ /* Always use WXOR for P/Q calculations (two destinations).
+ * Need 1 or 2 extra slots to verify results are zero.
+ */
+ idst = dst_cnt = (pdest && qdest) ? 2 : 1;
+
+ /* One additional slot per destination to clone P/Q
+ * before calculation (we have to preserve destinations).
+ */
+ slot_cnt = src_cnt + dst_cnt * 2;
+ slots_per_op = 1;
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+ sw_desc = ppc460ex_adma_alloc_slots(ppc460ex_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ ppc460ex_desc_init_pqzero_sum(sw_desc, dst_cnt, src_cnt);
+
+ /*
+ * Setup byte count for each slot just allocated
+ */
+ sw_desc->async_tx.flags = flags;
+ list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan,
+ len);
+ iter->unmap_len = len;
+ }
+
+ if (pdest) {
+ struct dma_cdb *hw_desc;
+ struct ppc460ex_adma_chan *chan;
+
+ iter = sw_desc->group_head;
+ chan = to_ppc460ex_adma_chan(iter->async_tx.chan);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter->src_cnt = 0;
+ iter->dst_cnt = 0;
+ ppc460ex_desc_set_dest_addr(iter, chan, 0,
+ ppc460ex_chan->pdest, 0);
+ ppc460ex_desc_set_src_addr(iter, chan, 0, 0, pdest);
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan,
+ len);
+ iter->unmap_len = 0;
+ /*
+ * override pdest to preserve original P
+ */
+ pdest = ppc460ex_chan->pdest;
+ }
+ if (qdest) {
+ struct dma_cdb *hw_desc;
+ struct ppc460ex_adma_chan *chan;
+
+ iter = list_first_entry(&sw_desc->group_list,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ chan = to_ppc460ex_adma_chan(iter->async_tx.chan);
+
+ if (pdest) {
+ iter = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ }
+
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter->src_cnt = 0;
+ iter->dst_cnt = 0;
+ ppc460ex_desc_set_dest_addr(iter, chan, 0,
+ ppc460ex_chan->qdest, 0);
+ ppc460ex_desc_set_src_addr(iter, chan, 0, 0, qdest);
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan,
+ len);
+ iter->unmap_len = 0;
+ /*
+ * override qdest to preserve original Q
+ */
+ qdest = ppc460ex_chan->qdest;
+ }
+
+ /*
+ * Setup destinations for P/Q ops
+ */
+ ppc460ex_adma_pqzero_sum_set_dest(sw_desc, pdest, qdest);
+
+ /*
+ * Setup zero QWORDs into DCHECK CDBs
+ */
+ idst = dst_cnt;
+ list_for_each_entry_reverse(iter, &sw_desc->group_list,
+ chain_node) {
+ /*
+ * The last CDB corresponds to Q-parity check,
+ * the one before last CDB corresponds
+ * P-parity check
+ */
+ if (idst == DMA_DEST_MAX_NUM) {
+ if (idst == dst_cnt) {
+ set_bit(PPC460EX_DESC_QCHECK,
+ &iter->flags);
+ } else {
+ set_bit(PPC460EX_DESC_PCHECK,
+ &iter->flags);
+ }
+ } else {
+ if (qdest) {
+ set_bit(PPC460EX_DESC_QCHECK,
+ &iter->flags);
+ } else {
+ set_bit(PPC460EX_DESC_PCHECK,
+ &iter->flags);
+ }
+ }
+ iter->xor_check_result = pqres;
+
+ /*
+ * set it to zero, if check fail then result will
+ * be updated
+ */
+ *iter->xor_check_result = 0;
+ ppc460ex_desc_set_dcheck(iter, ppc460ex_chan,
+ ppc460ex_qword);
+
+ if (!(--dst_cnt))
+ break;
+ }
+
+ /*
+ * Setup sources and mults for P/Q ops
+ */
+ list_for_each_entry_continue_reverse(iter, &sw_desc->group_list,
+ chain_node) {
+ struct ppc460ex_adma_chan *chan;
+ u32 mult_dst;
+
+ chan = to_ppc460ex_adma_chan(iter->async_tx.chan);
+ ppc460ex_desc_set_src_addr(iter, chan, 0,
+ DMA_CUED_XOR_HB,
+ src[src_cnt - 1]);
+ if (qdest) {
+ mult_dst = (dst_cnt - 1) ? DMA_CDB_SG_DST2 :
+ DMA_CDB_SG_DST1;
+ ppc460ex_desc_set_src_mult(iter, chan,
+ DMA_CUED_MULT1_OFF,
+ mult_dst,
+ scf[src_cnt - 1]);
+ }
+ if (!(--src_cnt))
+ break;
+ }
+ }
+ spin_unlock_bh(&ppc460ex_chan->lock);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+/*
+ * ppc460ex_adma_prep_dma_xor_zero_sum - prepare CDB group for
+ * XOR ZERO_SUM operation
+ */
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_xor_zero_sum(
+ struct dma_chan *chan, dma_addr_t *src, unsigned int src_cnt,
+ size_t len, enum sum_check_flags *result, unsigned long flags)
+{
+ struct dma_async_tx_descriptor *tx;
+ dma_addr_t pq[2];
+
+ /*
+ * validate P, disable Q
+ */
+ pq[0] = src[0];
+ pq[1] = 0;
+ flags |= DMA_PREP_PQ_DISABLE_Q;
+
+ tx = ppc460ex_adma_prep_dma_pqzero_sum(chan, pq, &src[1],
+ src_cnt - 1, 0, len,
+ result, flags);
+ return tx;
+}
+/*
+ * ppc460ex_desc_init_pq - initialize the descriptor for PQ_XOR operation
+ */
+static inline void ppc460ex_desc_init_pq(struct ppc460ex_adma_desc_slot *desc,
+ int dst_cnt, int src_cnt, unsigned long flags,
+ unsigned long op)
+{
+ struct dma_cdb *hw_desc;
+ struct ppc460ex_adma_desc_slot *iter;
+ u8 dopc;
+
+
+ /*
+ * Common initialization of a PQ descriptors chain
+ */
+
+ set_bits(op, &desc->flags);
+ desc->src_cnt = src_cnt;
+ desc->dst_cnt = dst_cnt;
+
+ dopc = (desc->dst_cnt == DMA_DEST_MAX_NUM) ?
+ DMA_CDB_OPC_MULTICAST : DMA_CDB_OPC_MV_SG1_SG2;
+
+ list_for_each_entry(iter, &desc->group_list, chain_node) {
+ hw_desc = iter->hw_desc;
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+
+ if (likely(!list_is_last(&iter->chain_node,
+ &desc->group_list))) {
+ /*
+ * set 'next' pointer
+ */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot, chain_node);
+ clear_bit(PPC460EX_DESC_INT, &iter->flags);
+ } else {
+ /* this is the last descriptor.
+ * this slot will be pasted from ADMA level
+ * each time it wants to configure parameters
+ * of the transaction (src, dst, ...)
+ */
+ iter->hw_next = NULL;
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC460EX_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC460EX_DESC_INT, &iter->flags);
+ }
+ }
+
+ /* Set OPS depending on WXOR/RXOR type of operation */
+ if (!test_bit(PPC460EX_DESC_RXOR, &desc->flags)) {
+ /* This is a WXOR only chain:
+ * - first descriptors are for zeroing destinations
+ * if PPC460EX_ZERO_P/Q set;
+ * - descriptors remained are for GF-XOR operations.
+ */
+ iter = list_first_entry(&desc->group_list,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+
+ if (test_bit(PPC460EX_ZERO_P, &desc->flags)) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ }
+
+ if (test_bit(PPC460EX_ZERO_Q, &desc->flags)) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ }
+
+ list_for_each_entry_from(iter, &desc->group_list, chain_node) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = dopc;
+ }
+ } else {
+ /* This is either RXOR-only or mixed RXOR/WXOR
+ * The first 1 or 2 slots in chain are always RXOR,
+ * if need to calculate P & Q, then there are two
+ * RXOR slots; if only P or only Q, then there is one
+ */
+ iter = list_first_entry(&desc->group_list,
+ struct ppc460ex_adma_desc_slot, chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+
+ if (desc->dst_cnt == DMA_DEST_MAX_NUM) {
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ }
+
+ /*
+ * The remain descs (if any) are WXORs
+ */
+ if (test_bit(PPC460EX_DESC_WXOR, &desc->flags)) {
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ list_for_each_entry_from(iter, &desc->group_list,
+ chain_node) {
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = dopc;
+ }
+ }
+ }
+}
+
+/*
+ * ppc460ex_adma_prep_dma_memcpy - prepare CDB for a MEMCPY operation
+ */
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_memcpy(
+ struct dma_chan *chan, dma_addr_t dma_dest,
+ dma_addr_t dma_src, size_t len, unsigned long flags)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ struct ppc460ex_adma_desc_slot *sw_desc, *group_start;
+ int slot_cnt, slots_per_op;
+ if (unlikely(!len))
+ return NULL;
+ BUG_ON(unlikely(len > PPC460EX_ADMA_DMA_MAX_BYTE_COUNT));
+ spin_lock_bh(&ppc460ex_chan->lock);
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: %s len: %u int_en %d\n",
+ ppc460ex_chan->device->id, __func__,
+ len, flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ slot_cnt = slots_per_op = 1;
+ sw_desc = ppc460ex_adma_alloc_slots(ppc460ex_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ ppc460ex_desc_init_memcpy(group_start, flags);
+ ppc460ex_adma_set_dest(group_start, dma_dest, 0);
+ ppc460ex_adma_memcpy_xor_set_src(group_start, dma_src, 0);
+ ppc460ex_desc_set_byte_count(group_start, ppc460ex_chan, len);
+ sw_desc->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+ spin_unlock_bh(&ppc460ex_chan->lock);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+
+/*
+ * ppc460ex_adma_prep_dma_memset - prepare CDB for a MEMSET operation
+ */
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_memset(
+ struct dma_chan *chan, dma_addr_t dma_dest, int value,
+ size_t len, unsigned long flags)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ struct ppc460ex_adma_desc_slot *sw_desc, *group_start;
+ int slot_cnt, slots_per_op;
+ if (unlikely(!len))
+ return NULL;
+ BUG_ON(unlikely(len > PPC460EX_ADMA_DMA_MAX_BYTE_COUNT));
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: %s cal: %u len: %u int_en %d\n",
+ ppc460ex_chan->device->id, __func__, value, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ slot_cnt = slots_per_op = 1;
+ sw_desc = ppc460ex_adma_alloc_slots(ppc460ex_chan, slot_cnt,
+ slots_per_op);
+ if (sw_desc) {
+ group_start = sw_desc->group_head;
+ ppc460ex_desc_init_memset(group_start, value, flags);
+ ppc460ex_adma_set_dest(group_start, dma_dest, 0);
+ ppc460ex_desc_set_byte_count(group_start, ppc460ex_chan, len);
+ sw_desc->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+ spin_unlock_bh(&ppc460ex_chan->lock);
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+static void ppc460ex_adma_pq_zero_op(struct ppc460ex_adma_desc_slot *iter,
+ struct ppc460ex_adma_chan *chan, dma_addr_t addr)
+{
+ /*
+ * To clear destinations update the descriptor
+ * (P or Q depending on index) as follows:
+ * addr is destination (0 corresponds to SG2):
+ */
+ ppc460ex_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE, addr, 0);
+
+ /* ... and the addr is source: */
+ ppc460ex_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB, addr);
+
+}
+/*
+ * ppc460ex_adma_set_dest - set destination address into descriptor
+ */
+static inline void ppc460ex_adma_set_dest(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index)
+{
+ struct ppc460ex_adma_chan *chan;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+ BUG_ON(index >= sw_desc->dst_cnt);
+
+ /*
+ * to do: support transfers lengths
+ * PPC460EX_ADMA_DMA/XOR_MAX_BYTE_COUNT
+ */
+ ppc460ex_desc_set_dest_addr(sw_desc->group_head,
+ chan, 0, addr, index);
+}
+/*
+ * ppc460ex_adma_pq_xor_set_dest - set destination address into descriptor
+ * for the PQXOR operation
+ */
+static void ppc460ex_adma_pqxor_set_dest(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t *addrs, unsigned long flags)
+{
+ struct ppc460ex_adma_desc_slot *iter;
+ struct ppc460ex_adma_chan *chan;
+ dma_addr_t paddr, qaddr;
+ dma_addr_t addr = 0, ppath, qpath;
+ int index = 0;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+ if (flags & DMA_PREP_PQ_DISABLE_P)
+ paddr = 0;
+ else
+ paddr = addrs[0];
+
+ if (flags & DMA_PREP_PQ_DISABLE_Q)
+ qaddr = 0;
+ else
+ qaddr = addrs[1];
+
+ if (!paddr || !qaddr)
+ addr = paddr ? paddr : qaddr;
+
+ /*
+ * walk through the WXOR source list and set P/Q-destinations
+ * for each slot:
+ */
+ if (!test_bit(PPC460EX_DESC_RXOR, &sw_desc->flags)) {
+ /*
+ * This is WXOR-only chain; may have 1/2 zero descs
+ */
+ if (test_bit(PPC460EX_ZERO_P, &sw_desc->flags))
+ index++;
+ if (test_bit(PPC460EX_ZERO_Q, &sw_desc->flags))
+ index++;
+
+ iter = ppc460ex_get_group_entry(sw_desc, index);
+ if (addr) {
+ /*
+ * one destination
+ */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list, chain_node)
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, addr, 0);
+ } else {
+ /*
+ * two destinations
+ */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list, chain_node) {
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, paddr, 0);
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, qaddr, 1);
+ }
+ }
+
+ if (index) {
+ /* To clear destinations update the descriptor
+ * (1st,2nd, or both depending on flags)
+ */
+ index = 0;
+ if (test_bit(PPC460EX_ZERO_P,
+ &sw_desc->flags)) {
+ iter = ppc460ex_get_group_entry(
+ sw_desc, index++);
+ ppc460ex_adma_pq_zero_op(iter, chan,
+ paddr);
+ }
+
+ if (test_bit(PPC460EX_ZERO_Q,
+ &sw_desc->flags)) {
+ iter = ppc460ex_get_group_entry(
+ sw_desc, index++);
+ ppc460ex_adma_pq_zero_op(iter, chan,
+ qaddr);
+ }
+
+ return;
+ }
+ } else {
+ /*
+ * This is RXOR-only or RXOR/WXOR mixed chain
+ *
+ * If we want to include destination into calculations,
+ * then make dest addresses cued with mult=1 (XOR).
+ */
+ ppath = test_bit(PPC460EX_ZERO_P, &sw_desc->flags) ?
+ DMA_CUED_XOR_HB :
+ DMA_CUED_XOR_BASE |
+ (1 << DMA_CUED_MULT1_OFF);
+ qpath = test_bit(PPC460EX_ZERO_Q, &sw_desc->flags) ?
+ DMA_CUED_XOR_HB :
+ DMA_CUED_XOR_BASE |
+ (1 << DMA_CUED_MULT1_OFF);
+
+ /*
+ * Setup destination(s) in RXOR slot(s)
+ */
+ iter = ppc460ex_get_group_entry(sw_desc, index++);
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ paddr ? ppath : qpath,
+ paddr ? paddr : qaddr, 0);
+ if (!addr) {
+ /*
+ * two destinations
+ */
+ iter = ppc460ex_get_group_entry(sw_desc,
+ index++);
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ qpath, qaddr, 0);
+ }
+
+ if (test_bit(PPC460EX_DESC_WXOR, &sw_desc->flags)) {
+ /* Setup destination(s) in remaining WXOR
+ * slots
+ */
+ iter = ppc460ex_get_group_entry(sw_desc,
+ index);
+ if (addr) {
+ /*
+ * one destination
+ */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list,
+ chain_node)
+ ppc460ex_desc_set_dest_addr(
+ iter, chan,
+ DMA_CUED_XOR_BASE,
+ addr, 0);
+
+ } else {
+ /*
+ * two destinations
+ */
+ list_for_each_entry_from(iter,
+ &sw_desc->group_list,
+ chain_node) {
+ ppc460ex_desc_set_dest_addr(
+ iter, chan,
+ DMA_CUED_XOR_BASE,
+ paddr, 0);
+ ppc460ex_desc_set_dest_addr(
+ iter, chan,
+ DMA_CUED_XOR_BASE,
+ qaddr, 1);
+ }
+ }
+ }
+
+ }
+}
+
+/*
+ * ppc460ex_adma_pqxor_set_src - set source address into descriptor
+ */
+static void ppc460ex_adma_pqxor_set_src(struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t addr, int index)
+{
+ struct ppc460ex_adma_chan *chan;
+ dma_addr_t haddr = 0;
+ struct ppc460ex_adma_desc_slot *iter = NULL;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+ /* DMA0,1 may do: WXOR, RXOR, RXOR+WXORs chain
+ */
+ if (test_bit(PPC460EX_DESC_RXOR, &sw_desc->flags)) {
+ /*
+ * RXOR-only or RXOR/WXOR operation
+ */
+ int iskip = test_bit(PPC460EX_DESC_RXOR12,
+ &sw_desc->flags) ? 2 : 3;
+
+ if (index == 0) {
+ /* 1st slot (RXOR)
+ * setup sources region (R1-2-3, R1-2-4,or R1-2-5)
+ */
+ if (test_bit(PPC460EX_DESC_RXOR12,
+ &sw_desc->flags))
+ haddr = DMA_RXOR12 <<
+ DMA_CUED_REGION_OFF;
+ else if (test_bit(PPC460EX_DESC_RXOR123,
+ &sw_desc->flags))
+ haddr = DMA_RXOR123 <<
+ DMA_CUED_REGION_OFF;
+ else if (test_bit(PPC460EX_DESC_RXOR124,
+ &sw_desc->flags))
+ haddr = DMA_RXOR124 <<
+ DMA_CUED_REGION_OFF;
+ else if (test_bit(PPC460EX_DESC_RXOR125,
+ &sw_desc->flags))
+ haddr = DMA_RXOR125 <<
+ DMA_CUED_REGION_OFF;
+ else
+ BUG();
+ haddr |= DMA_CUED_XOR_BASE;
+ sw_desc = sw_desc->group_head;
+ } else if (index < iskip) {
+ /* 1st slot (RXOR)
+ * shall actually set source address only once
+ * instead of first <iskip>
+ */
+ iter = NULL;
+ } else {
+ /* second and next slots (WXOR);
+ * skip first slot with RXOR
+ */
+ haddr = DMA_CUED_XOR_HB;
+ sw_desc = ppc460ex_get_group_entry(sw_desc,
+ index - iskip + 1);
+ }
+ } else {
+ int znum = 0;
+ /* WXOR-only operation;
+ * skip first slots with destinations
+ */
+ if (test_bit(PPC460EX_ZERO_P, &sw_desc->flags))
+ znum++;
+ if (test_bit(PPC460EX_ZERO_Q, &sw_desc->flags))
+ znum++;
+
+ haddr = DMA_CUED_XOR_HB;
+ iter = ppc460ex_get_group_entry(sw_desc,
+ index + znum);
+ }
+
+ if (likely(iter)) {
+ ppc460ex_desc_set_src_addr(iter, chan, 0, haddr, addr);
+ if (!index &&
+ test_bit(PPC460EX_DESC_RXOR, &sw_desc->flags) &&
+ sw_desc->dst_cnt == 2) {
+ /* if we have two destinations for RXOR, then
+ * setup source in the second descr too
+ */
+ iter = ppc460ex_get_group_entry(sw_desc, 1);
+ ppc460ex_desc_set_src_addr(iter, chan, 0,
+ haddr, addr);
+ }
+ }
+}
+/*
+ * ppc460ex_adma_pqxor_set_src_mult - set multiplication coefficient into
+ * descriptor for the PQXOR operation
+ */
+static void ppc460ex_adma_pqxor_set_src_mult(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ unsigned char mult, int index, int dst_pos)
+{
+ struct ppc460ex_adma_chan *chan;
+ u32 mult_idx, mult_dst;
+ struct ppc460ex_adma_desc_slot *iter = NULL, *iter1 = NULL;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+ if (test_bit(PPC460EX_DESC_RXOR, &sw_desc->flags)) {
+ int region = test_bit(PPC460EX_DESC_RXOR12,
+ &sw_desc->flags) ? 2 : 3;
+
+ if (index < region) {
+ /*
+ * RXOR multipliers
+ */
+
+ iter = ppc460ex_get_group_entry(sw_desc,
+ sw_desc->dst_cnt - 1);
+ if (sw_desc->dst_cnt == 2)
+ iter1 = ppc460ex_get_group_entry(sw_desc, 0);
+ mult_idx = DMA_CUED_MULT1_OFF + (index << 3);
+ mult_dst = DMA_CDB_SG_SRC;
+ } else {
+ /*
+ * WXOR multiplier
+ */
+ iter = ppc460ex_get_group_entry(sw_desc,
+ index - region +
+ sw_desc->dst_cnt);
+ mult_idx = DMA_CUED_MULT1_OFF;
+ mult_dst = dst_pos ? DMA_CDB_SG_DST2 :
+ DMA_CDB_SG_DST1;
+ }
+ } else {
+ int znum = 0;
+
+ /* WXOR-only;
+ * skip first slots with destinations (if ZERO_DST has
+ * place)
+ */
+ if (test_bit(PPC460EX_ZERO_P, &sw_desc->flags))
+ znum++;
+ if (test_bit(PPC460EX_ZERO_Q, &sw_desc->flags))
+ znum++;
+ iter = ppc460ex_get_group_entry(sw_desc, index + znum);
+ mult_idx = DMA_CUED_MULT1_OFF;
+ mult_dst = dst_pos ? DMA_CDB_SG_DST2 : DMA_CDB_SG_DST1;
+ }
+
+ if (likely(iter)) {
+ ppc460ex_desc_set_src_mult(iter, chan,
+ mult_idx, mult_dst, mult);
+
+ if (unlikely(iter1)) {
+ /* if we have two destinations for RXOR, then
+ * we've just set Q mult. Set-up P now.
+ */
+ ppc460ex_desc_set_src_mult(iter1, chan,
+ mult_idx, mult_dst, 1);
+ }
+
+ }
+}
+
+
+static inline struct ppc460ex_adma_desc_slot *ppc460ex_dma_prep_pq(
+ struct ppc460ex_adma_chan *ppc460ex_chan,
+ dma_addr_t *dst, unsigned int dst_cnt,
+ dma_addr_t *src, unsigned int src_cnt, unsigned char *scf,
+ size_t len, unsigned long flags)
+{
+ int slot_cnt;
+ struct ppc460ex_adma_desc_slot *sw_desc = NULL, *iter;
+ unsigned long op = 0;
+ unsigned char mult = 1;
+
+ /* select operations WXOR/RXOR depending on the
+ * source addresses of operators and the number
+ * of destinations (RXOR support only Q-parity calculations)
+ */
+ set_bit(PPC460EX_DESC_WXOR, &op);
+ if (!test_and_set_bit(PPC460EX_RXOR_RUN, &ppc460ex_rxor_state)) {
+ /* no active RXOR;
+ * do RXOR if:
+ * - destination os only one,
+ * - there are more than 1 source,
+ * - len is aligned on 512-byte boundary,
+ * - source addresses fit to one of 4 possible regions.
+ */
+ if (dst_cnt == 3 && src_cnt > 1 &&
+ !(len & ~MQ0_CF2H_RXOR_BS_MASK) &&
+ (src[0] + len) == src[1]) {
+ /* may do RXOR R1 R2 */
+ set_bit(PPC460EX_DESC_RXOR, &op);
+ if (src_cnt != 2) {
+ /* may try to enhance region of RXOR */
+ if ((src[1] + len) == src[2]) {
+ /* do RXOR R1 R2 R3 */
+ set_bit(PPC460EX_DESC_RXOR123,
+ &op);
+ } else if ((src[1] + len * 2) == src[2]) {
+ /* do RXOR R1 R2 R4 */
+ set_bit(PPC460EX_DESC_RXOR124, &op);
+ } else if ((src[1] + len * 3) == src[2]) {
+ /* do RXOR R1 R2 R5 */
+ set_bit(PPC460EX_DESC_RXOR125,
+ &op);
+ } else {
+ /* do RXOR R1 R2 */
+ set_bit(PPC460EX_DESC_RXOR12,
+ &op);
+ }
+ } else {
+ /* do RXOR R1 R2 */
+ set_bit(PPC460EX_DESC_RXOR12, &op);
+ }
+ }
+
+ if (!test_bit(PPC460EX_DESC_RXOR, &op)) {
+ /* can not do this operation with RXOR */
+ clear_bit(PPC460EX_RXOR_RUN,
+ &ppc460ex_rxor_state);
+ } else {
+ /* can do; set block size right now */
+ ppc460ex_desc_set_rxor_block_size(len);
+ }
+ }
+
+ /*
+ * Number of necessary slots depends on operation type selected
+ */
+ if (!test_bit(PPC460EX_DESC_RXOR, &op)) {
+ /* This is a WXOR only chain. Need descriptors for each
+ * source to GF-XOR them with WXOR, and need descriptors
+ * for each destination to zero them with WXOR
+ */
+ slot_cnt = src_cnt;
+
+ if (flags & DMA_PREP_ZERO_P) {
+ slot_cnt++;
+ set_bit(PPC460EX_ZERO_P, &op);
+ }
+ if (flags & DMA_PREP_ZERO_Q) {
+ slot_cnt++;
+ set_bit(PPC460EX_ZERO_Q, &op);
+ }
+ } else {
+ /* Need 1/2 descriptor for RXOR operation, and
+ * need (src_cnt - (2 or 3)) for WXOR of sources
+ * remained (if any)
+ */
+ slot_cnt = dst_cnt;
+
+ if (flags & DMA_PREP_ZERO_P)
+ set_bit(PPC460EX_ZERO_P, &op);
+ if (flags & DMA_PREP_ZERO_Q)
+ set_bit(PPC460EX_ZERO_Q, &op);
+
+ if (test_bit(PPC460EX_DESC_RXOR12, &op))
+ slot_cnt += src_cnt - 2;
+ else
+ slot_cnt += src_cnt - 3;
+
+ /*
+ * Thus we have either RXOR only chain or
+ * mixed RXOR/WXOR
+ */
+ if (slot_cnt == dst_cnt) {
+ /* RXOR only chain */
+ clear_bit(PPC460EX_DESC_WXOR, &op);
+ }
+ }
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+ /*
+ * for both RXOR/WXOR each descriptor occupies one slot
+ */
+ sw_desc = ppc460ex_adma_alloc_slots(ppc460ex_chan, slot_cnt, 1);
+ if (sw_desc) {
+ ppc460ex_desc_init_pq(sw_desc, dst_cnt, src_cnt,
+ flags, op);
+
+ /*
+ * setup dst/src/mult
+ */
+ ppc460ex_adma_pqxor_set_dest(sw_desc,
+ dst, flags);
+ while (src_cnt--) {
+ ppc460ex_adma_pqxor_set_src(sw_desc, src[src_cnt],
+ src_cnt);
+ if (!(flags & DMA_PREP_PQ_DISABLE_Q))
+ mult = scf[src_cnt];
+ ppc460ex_adma_pqxor_set_src_mult(sw_desc, mult,
+ src_cnt, dst_cnt - 1);
+ }
+
+ /*
+ * Setup byte count foreach slot just allocated
+ */
+ sw_desc->async_tx.flags = flags;
+ list_for_each_entry(iter, &sw_desc->group_list,
+ chain_node) {
+ ppc460ex_desc_set_byte_count(iter,
+ ppc460ex_chan, len);
+ iter->unmap_len = len;
+ }
+ }
+ spin_unlock_bh(&ppc460ex_chan->lock);
+
+ return sw_desc;
+}
+/*
+ * ppc460ex_dma01_prep_mult -
+ * for Q operation where destination is also the source
+ */
+static struct ppc460ex_adma_desc_slot *ppc460ex_dma01_prep_mult(
+ struct ppc460ex_adma_chan *ppc460ex_chan,
+ dma_addr_t *dst, int dst_cnt, dma_addr_t *src, int src_cnt,
+ const unsigned char *scf, size_t len, unsigned long flags)
+{
+ struct ppc460ex_adma_desc_slot *sw_desc = NULL;
+ unsigned long op = 0;
+ int slot_cnt;
+
+ set_bit(PPC460EX_DESC_WXOR, &op);
+ slot_cnt = 2;
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+
+ /*
+ * use WXOR, each descriptor occupies one slot
+ */
+ sw_desc = ppc460ex_adma_alloc_slots(ppc460ex_chan, slot_cnt, 1);
+ if (sw_desc) {
+ struct ppc460ex_adma_chan *chan;
+ struct ppc460ex_adma_desc_slot *iter;
+ struct dma_cdb *hw_desc;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+ set_bits(op, &sw_desc->flags);
+ sw_desc->src_cnt = src_cnt;
+ sw_desc->dst_cnt = dst_cnt;
+ /*
+ * First descriptor, zero data in the destination and copy it
+ * to q page using MULTICAST transfer.
+ */
+ iter = list_first_entry(&sw_desc->group_list,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ /*
+ * set 'next' pointer
+ */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ clear_bit(PPC460EX_DESC_INT, &iter->flags);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MULTICAST;
+
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, dst[0], 0);
+ ppc460ex_desc_set_dest_addr(iter, chan, 0, dst[1], 1);
+ ppc460ex_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ src[0]);
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan, len);
+ iter->unmap_len = len;
+
+ /*
+ * Second descriptor, multiply data from the q page
+ * and store the result in real destination.
+ */
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = NULL;
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC460EX_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC460EX_DESC_INT, &iter->flags);
+
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ ppc460ex_desc_set_src_addr(iter, chan, 0,
+ DMA_CUED_XOR_HB, dst[1]);
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, dst[0], 0);
+
+ ppc460ex_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+ DMA_CDB_SG_DST1, scf[0]);
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan, len);
+ iter->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+
+ spin_unlock_bh(&ppc460ex_chan->lock);
+
+ return sw_desc;
+}
+/*
+ * ppc460ex_dma01_prep_sum_product -
+ * Dx = A*(P+Pxy) + B*(Q+Qxy) operation where destination is also
+ * the source.
+ */
+static struct ppc460ex_adma_desc_slot *ppc460ex_dma01_prep_sum_product(
+ struct ppc460ex_adma_chan *ppc460ex_chan,
+ dma_addr_t *dst, dma_addr_t *src, int src_cnt,
+ const unsigned char *scf, size_t len, unsigned long flags)
+{
+ struct ppc460ex_adma_desc_slot *sw_desc = NULL;
+ unsigned long op = 0;
+ int slot_cnt;
+
+ set_bit(PPC460EX_DESC_WXOR, &op);
+ slot_cnt = 3;
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+
+ /*
+ * WXOR, each descriptor occupies one slot
+ */
+ sw_desc = ppc460ex_adma_alloc_slots(ppc460ex_chan, slot_cnt, 1);
+ if (sw_desc) {
+ struct ppc460ex_adma_chan *chan;
+ struct ppc460ex_adma_desc_slot *iter;
+ struct dma_cdb *hw_desc;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+ set_bits(op, &sw_desc->flags);
+ sw_desc->src_cnt = src_cnt;
+ sw_desc->dst_cnt = 1;
+ /*
+ * 1st descriptor, src[1] data to q page and zero destination
+ */
+ iter = list_first_entry(&sw_desc->group_list,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ clear_bit(PPC460EX_DESC_INT, &iter->flags);
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MULTICAST;
+
+ ppc460ex_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+ *dst, 0);
+ ppc460ex_desc_set_dest_addr(iter, chan, 0,
+ ppc460ex_chan->qdest, 1);
+ ppc460ex_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ src[1]);
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan, len);
+ iter->unmap_len = len;
+
+ /*
+ * 2nd descriptor, multiply src[1] data and store the
+ * result in destination */
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ /* set 'next' pointer */
+ iter->hw_next = list_entry(iter->chain_node.next,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC460EX_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC460EX_DESC_INT, &iter->flags);
+
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ ppc460ex_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ ppc460ex_chan->qdest);
+ ppc460ex_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+ *dst, 0);
+ ppc460ex_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+ DMA_CDB_SG_DST1, scf[1]);
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan, len);
+ iter->unmap_len = len;
+
+ /*
+ * 3rd descriptor, multiply src[0] data and xor it
+ * with destination
+ */
+ iter = list_first_entry(&iter->chain_node,
+ struct ppc460ex_adma_desc_slot,
+ chain_node);
+ memset(iter->hw_desc, 0, sizeof(struct dma_cdb));
+ iter->hw_next = NULL;
+ if (flags & DMA_PREP_INTERRUPT)
+ set_bit(PPC460EX_DESC_INT, &iter->flags);
+ else
+ clear_bit(PPC460EX_DESC_INT, &iter->flags);
+
+ hw_desc = iter->hw_desc;
+ hw_desc->opc = DMA_CDB_OPC_MV_SG1_SG2;
+ ppc460ex_desc_set_src_addr(iter, chan, 0, DMA_CUED_XOR_HB,
+ src[0]);
+ ppc460ex_desc_set_dest_addr(iter, chan, DMA_CUED_XOR_BASE,
+ *dst, 0);
+ ppc460ex_desc_set_src_mult(iter, chan, DMA_CUED_MULT1_OFF,
+ DMA_CDB_SG_DST1, scf[0]);
+ ppc460ex_desc_set_byte_count(iter, ppc460ex_chan, len);
+ iter->unmap_len = len;
+ sw_desc->async_tx.flags = flags;
+ }
+
+ spin_unlock_bh(&ppc460ex_chan->lock);
+
+ return sw_desc;
+}
+/*
+ * ppc460ex_adma_prep_dma_pq- prepare CDB (group) for a GF-XOR operation
+ */
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_pq(
+ struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
+ unsigned int src_cnt, unsigned char *scf,
+ size_t len, unsigned long flags)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ struct ppc460ex_adma_desc_slot *sw_desc = NULL;
+ int dst_cnt = 0;
+
+ BUG_ON(!len);
+ BUG_ON(unlikely(len > PPC460EX_ADMA_XOR_MAX_BYTE_COUNT));
+ BUG_ON(!src_cnt);
+
+ if (src_cnt == 1 && dst[1] == src[0]) {
+ dma_addr_t dest[2];
+
+ /* dst[1] is real destination (Q) */
+ dest[0] = dst[1];
+ /* this is the page to multicast source data to */
+ dest[1] = ppc460ex_chan->qdest;
+ sw_desc = ppc460ex_dma01_prep_mult(ppc460ex_chan,
+ dest, 2, src, src_cnt, scf, len, flags);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+ }
+
+ if (src_cnt == 2 && dst[1] == src[1]) {
+ sw_desc = ppc460ex_dma01_prep_sum_product(ppc460ex_chan,
+ &dst[1], src, 2, scf, len, flags);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+ }
+ if (!(flags & DMA_PREP_PQ_DISABLE_P)) {
+ BUG_ON(!dst[0]);
+ dst_cnt++;
+ flags |= DMA_PREP_ZERO_P;
+ }
+
+ if (!(flags & DMA_PREP_PQ_DISABLE_Q)) {
+ BUG_ON(!dst[1]);
+ dst_cnt++;
+ flags |= DMA_PREP_ZERO_Q;
+ }
+ BUG_ON(!dst_cnt);
+
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: %s src_cnt: %d len: %u int_en: %d\n",
+ ppc460ex_chan->device->id, __func__, src_cnt, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ sw_desc = ppc460ex_dma_prep_pq(ppc460ex_chan,
+ dst, dst_cnt, src, src_cnt, scf,
+ len, flags);
+
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+/*
+ * ppc460ex_adma_prep_dma_p- prepare CDB (group) for a GF-XOR operation
+ */
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_p(
+ struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
+ unsigned int src_cnt, unsigned char *scf,
+ size_t len, unsigned long flags)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ struct ppc460ex_adma_desc_slot *sw_desc = NULL;
+ int dst_cnt = 0;
+
+ BUG_ON(!len);
+ BUG_ON(unlikely(len > PPC460EX_ADMA_XOR_MAX_BYTE_COUNT));
+ BUG_ON(!src_cnt);
+
+ if (src_cnt == 1 && dst[1] == src[0]) {
+ dma_addr_t dest[2];
+
+ /* dst[1] is real destination (Q) */
+ dest[0] = dst[1];
+ /* this is the page to multicast source data to */
+ dest[1] = ppc460ex_chan->qdest;
+ sw_desc = ppc460ex_dma01_prep_mult(ppc460ex_chan,
+ dest, 2, src, src_cnt, scf, len, flags);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+ }
+
+ if (src_cnt == 2 && dst[1] == src[1]) {
+ sw_desc = ppc460ex_dma01_prep_sum_product(ppc460ex_chan,
+ &dst[1], src, 2, scf, len, flags);
+ return sw_desc ? &sw_desc->async_tx : NULL;
+ }
+ if (!(flags & DMA_PREP_PQ_DISABLE_P)) {
+ BUG_ON(!dst[0]);
+ dst_cnt++;
+ if (flags & DMA_ZERO_P)
+ flags |= DMA_PREP_ZERO_P;
+ }
+
+ if (!(flags & DMA_PREP_PQ_DISABLE_Q)) {
+ BUG_ON(!dst[1]);
+ dst_cnt++;
+ flags |= DMA_PREP_ZERO_Q;
+ }
+ BUG_ON(!dst_cnt);
+
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: %s src_cnt: %d len: %u int_en: %d\n",
+ ppc460ex_chan->device->id, __func__, src_cnt, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ sw_desc = ppc460ex_dma_prep_pq(ppc460ex_chan,
+ dst, dst_cnt, src, src_cnt, scf,
+ len, flags);
+
+ return sw_desc ? &sw_desc->async_tx : NULL;
+}
+/*
+ * ppc460ex_adma_prep_dma_mq_xor - prepare CDB (group) for a GF-XOR operation
+ */
+static struct dma_async_tx_descriptor *ppc460ex_adma_prep_dma_mq_xor(
+ struct dma_chan *chan, dma_addr_t dst,
+ dma_addr_t *src, unsigned int src_cnt,
+ size_t len, unsigned long flags)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ struct dma_async_tx_descriptor *tx;
+ dma_addr_t pq[2];
+ unsigned char scf = 0;
+
+ /* validate P, disable Q */
+ pq[0] = dst;
+ pq[1] = 0;
+ flags |= DMA_PREP_PQ_DISABLE_Q;
+
+ BUG_ON(!len);
+ BUG_ON(unlikely(len > PPC460EX_ADMA_XOR_MAX_BYTE_COUNT));
+ BUG_ON(!src_cnt);
+
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: %s src_cnt: %d len: %u int_en: %d\n",
+ ppc460ex_chan->device->id, __func__, src_cnt, len,
+ flags & DMA_PREP_INTERRUPT ? 1 : 0);
+
+ tx = ppc460ex_adma_prep_dma_p(chan, &pq[0], src,
+ src_cnt, &scf, len, flags);
+ return tx;
+
+}
+/*
+ * ppc460ex_adma_memcpy_xor_set_src - set source address into descriptor
+ */
+static inline void ppc460ex_adma_memcpy_xor_set_src(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t addr,
+ int index)
+{
+ struct ppc460ex_adma_chan *chan;
+
+ sw_desc = sw_desc->group_head;
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+
+ if (likely(sw_desc))
+ ppc460ex_desc_set_src_addr(sw_desc, chan, index, 0, addr);
+}
+/*
+ * ppc460ex_adma_pqzero_sum_set_src_mult - set multiplication coefficient
+ * into descriptor for the PQZERO_SUM operation
+ */
+static void ppc460ex_adma_pqzero_sum_set_src_mult(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ unsigned char mult, int index, int dst_pos)
+{
+ struct ppc460ex_adma_chan *chan;
+ u32 mult_idx, mult_dst;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+ /* set mult for sources only */
+ BUG_ON(index >= sw_desc->src_cnt);
+
+ /* get pointed slot */
+ sw_desc = ppc460ex_get_group_entry(sw_desc, index);
+
+ mult_idx = DMA_CUED_MULT1_OFF;
+ mult_dst = dst_pos ? DMA_CDB_SG_DST2 : DMA_CDB_SG_DST1;
+
+ if (likely(sw_desc))
+ ppc460ex_desc_set_src_mult(sw_desc, chan, mult_idx, mult_dst,
+ mult);
+}
+/*
+ * ppc460ex_adma_pq_set_src_mult - set multiplication coefficient into
+ * descriptor for the PQXOR operation
+ */
+static void ppc460ex_adma_pq_set_src_mult(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ unsigned char mult, int index, int dst_pos)
+{
+ struct ppc460ex_adma_chan *chan;
+ u32 mult_idx, mult_dst;
+ struct ppc460ex_adma_desc_slot *iter = NULL, *iter1 = NULL;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+
+ if (test_bit(PPC460EX_DESC_RXOR, &sw_desc->flags)) {
+ int region = test_bit(PPC460EX_DESC_RXOR12,
+ &sw_desc->flags) ? 2 : 3;
+
+ if (index < region) {
+ /* RXOR multipliers */
+ iter = ppc460ex_get_group_entry(sw_desc,
+ sw_desc->dst_cnt - 1);
+ if (sw_desc->dst_cnt == 2)
+ iter1 = ppc460ex_get_group_entry(
+ sw_desc, 0);
+
+ mult_idx = DMA_CUED_MULT1_OFF + (index << 3);
+ mult_dst = DMA_CDB_SG_SRC;
+ } else {
+ /* WXOR multiplier */
+ iter = ppc460ex_get_group_entry(sw_desc,
+ index - region +
+ sw_desc->dst_cnt);
+ mult_idx = DMA_CUED_MULT1_OFF;
+ mult_dst = dst_pos ? DMA_CDB_SG_DST2 :
+ DMA_CDB_SG_DST1;
+ }
+ } else {
+ int znum = 0;
+
+ /* WXOR-only;
+ * skip first slots with destinations (if ZERO_DST has
+ * place)
+ */
+ if (test_bit(PPC460EX_ZERO_P, &sw_desc->flags))
+ znum++;
+ if (test_bit(PPC460EX_ZERO_Q, &sw_desc->flags))
+ znum++;
+
+ iter = ppc460ex_get_group_entry(sw_desc, index + znum);
+ mult_idx = DMA_CUED_MULT1_OFF;
+ mult_dst = dst_pos ? DMA_CDB_SG_DST2 : DMA_CDB_SG_DST1;
+ }
+
+ if (likely(iter)) {
+ ppc460ex_desc_set_src_mult(iter, chan,
+ mult_idx, mult_dst, mult);
+
+ if (unlikely(iter1)) {
+ /* if we have two destinations for RXOR, then
+ * we've just set Q mult. Set-up P now.
+ */
+ ppc460ex_desc_set_src_mult(iter1, chan,
+ mult_idx, mult_dst, 1);
+ }
+
+ }
+
+}
+/*
+ * ppc460ex_adma_pq_zero_sum_set_dest - set destination address into descriptor
+ * for the PQ_ZERO_SUM operation
+ */
+void ppc460ex_adma_pqzero_sum_set_dest(
+ struct ppc460ex_adma_desc_slot *sw_desc,
+ dma_addr_t paddr, dma_addr_t qaddr)
+{
+ struct ppc460ex_adma_desc_slot *iter, *end;
+ struct ppc460ex_adma_chan *chan;
+ dma_addr_t addr = 0;
+ int idx;
+
+ chan = to_ppc460ex_adma_chan(sw_desc->async_tx.chan);
+
+ /* walk through the WXOR source list and set P/Q-destinations
+ * for each slot
+ */
+ idx = (paddr && qaddr) ? 2 : 1;
+ /* set end */
+ list_for_each_entry_reverse(end, &sw_desc->group_list,
+ chain_node) {
+ if (!(--idx))
+ break;
+ }
+ /* set start */
+ idx = (paddr && qaddr) ? 2 : 1;
+ iter = ppc460ex_get_group_entry(sw_desc, idx);
+
+ if (paddr && qaddr) {
+ /* two destinations */
+ list_for_each_entry_from(iter, &sw_desc->group_list,
+ chain_node) {
+ if (unlikely(iter == end))
+ break;
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, paddr, 0);
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, qaddr, 1);
+ }
+ } else {
+ /* one destination */
+ addr = paddr ? paddr : qaddr;
+ list_for_each_entry_from(iter, &sw_desc->group_list,
+ chain_node) {
+ if (unlikely(iter == end))
+ break;
+ ppc460ex_desc_set_dest_addr(iter, chan,
+ DMA_CUED_XOR_BASE, addr, 0);
+ }
+ }
+
+ /* The remaining descriptors are DATACHECK. These have no need in
+ * destination. Actually, these destinations are used there
+ * as sources for check operation. So, set addr as source.
+ */
+ ppc460ex_desc_set_src_addr(end, chan, 0, 0, addr ? addr : paddr);
+
+ if (!addr) {
+ end = list_entry(end->chain_node.next,
+ struct ppc460ex_adma_desc_slot, chain_node);
+ ppc460ex_desc_set_src_addr(end, chan, 0, 0, qaddr);
+ }
+}
+/*
+ * ppc460ex_adma_free_chan_resources - free the resources allocated
+ */
+void ppc460ex_adma_free_chan_resources(struct dma_chan *chan)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ struct ppc460ex_adma_desc_slot *iter, *_iter;
+ int in_use_descs = 0;
+
+ ppc460ex_adma_slot_cleanup(ppc460ex_chan);
+
+ spin_lock_bh(&ppc460ex_chan->lock);
+ list_for_each_entry_safe(iter, _iter, &ppc460ex_chan->chain,
+ chain_node) {
+ in_use_descs++;
+ list_del(&iter->chain_node);
+ }
+ list_for_each_entry_safe_reverse(iter, _iter,
+ &ppc460ex_chan->all_slots, slot_node) {
+ list_del(&iter->slot_node);
+ kfree(iter);
+ ppc460ex_chan->slots_allocated--;
+ }
+ ppc460ex_chan->last_used = NULL;
+
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d %s slots_allocated %d\n",
+ ppc460ex_chan->device->id,
+ __func__, ppc460ex_chan->slots_allocated);
+ spin_unlock_bh(&ppc460ex_chan->lock);
+
+ /* one is ok since we left it on there on purpose */
+ if (in_use_descs > 1)
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "GT: Freeing %d in use descriptors!\n",
+ in_use_descs - 1);
+}
+
+/**
+ * ppc460ex_adma_tx_status - poll the status of an ADMA transaction
+ * @chan: ADMA channel handle
+ * @cookie: ADMA transaction identifier
+ * @txstate: a holder for the current state of the channel
+ */
+static enum dma_status ppc460ex_adma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie, struct dma_tx_state *txstate)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan;
+ dma_cookie_t last_used;
+ dma_cookie_t last_complete;
+ enum dma_status ret;
+
+ ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ last_used = chan->cookie;
+ last_complete = ppc460ex_chan->completed_cookie;
+
+ dma_set_tx_state(txstate, last_complete, last_used, 0);
+
+ ret = dma_async_is_complete(cookie, last_complete, last_used);
+ if (ret == DMA_SUCCESS)
+ return ret;
+
+ ppc460ex_adma_slot_cleanup(ppc460ex_chan);
+
+ last_used = chan->cookie;
+ last_complete = ppc460ex_chan->completed_cookie;
+
+ dma_set_tx_state(txstate, last_complete, last_used, 0);
+
+ return dma_async_is_complete(cookie, last_complete, last_used);
+}
+/*
+ * ppc460ex_adma_is_complete - poll the status of an ADMA transaction
+ * @chan: ADMA channel handle
+ * @cookie: ADMA transaction identifier
+ */
+enum dma_status ppc460ex_adma_is_complete(struct dma_chan *chan,
+ dma_cookie_t cookie, dma_cookie_t *done, dma_cookie_t *used)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ dma_cookie_t last_used;
+ dma_cookie_t last_complete;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+ last_complete = ppc460ex_chan->completed_cookie;
+
+ if (done)
+ *done = last_complete;
+ if (used)
+ *used = last_used;
+
+ ret = dma_async_is_complete(cookie, last_complete, last_used);
+ if (ret == DMA_SUCCESS)
+ return ret;
+
+ ppc460ex_adma_slot_cleanup(ppc460ex_chan);
+
+ last_used = chan->cookie;
+ last_complete = ppc460ex_chan->completed_cookie;
+
+ if (done)
+ *done = last_complete;
+ if (used)
+ *used = last_used;
+
+ return dma_async_is_complete(cookie, last_complete, last_used);
+}
+/*
+ * ppc460ex_adma_eot_handler - end of transfer interrupt handler
+ */
+static irqreturn_t ppc460ex_adma_eot_handler(int irq, void *data)
+{
+ struct ppc460ex_adma_chan *chan = data;
+
+ dev_dbg(chan->device->common.dev,
+ "ppc460ex adma%d: %s\n", chan->device->id, __func__);
+
+ tasklet_schedule(&chan->irq_tasklet);
+ ppc460ex_adma_device_clear_eot_status(chan);
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * ppc460ex_adma_err_handler - DMA error interrupt handler;
+ * do the same things as a eot handler
+ */
+static irqreturn_t ppc460ex_adma_err_handler(int irq, void *data)
+{
+ struct ppc460ex_adma_chan *chan = data;
+ dev_dbg(chan->device->common.dev,
+ "ppc460ex adma%d: %s\n", chan->device->id, __func__);
+ tasklet_schedule(&chan->irq_tasklet);
+ ppc460ex_adma_device_clear_eot_status(chan);
+
+ return IRQ_HANDLED;
+}
+
+static void ppc460ex_test_rad6_callback(void *unused)
+{
+ complete(&ppc460ex_r6_test_comp);
+}
+/*
+ * ppc460ex_test_callback - called when test operation has been done
+ */
+static void ppc460ex_test_raid5_callback(void *unused)
+{
+ complete(&ppc460ex_r5_test_comp);
+}
+/*
+ * ppc460ex_adma_issue_pending - flush all pending descriptors to h/w
+ */
+static void ppc460ex_adma_issue_pending(struct dma_chan *chan)
+{
+ struct ppc460ex_adma_chan *ppc460ex_chan;
+
+ ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ dev_dbg(ppc460ex_chan->device->common.dev,
+ "ppc460ex adma%d: %s %d\n", ppc460ex_chan->device->id, __func__,
+ ppc460ex_chan->pending);
+ if (ppc460ex_chan->pending) {
+ ppc460ex_chan->pending = 0;
+ ppc460ex_chan_append(ppc460ex_chan);
+ }
+}
+/*
+ * ppc460ex_test_raid6 - test are RAID-6 capabilities enabled successfully.
+ * For this we just perform one WXOR operation with the same source
+ * and destination addresses, the GF-multiplier is 1; so if RAID-6
+ o/of_platform_driver_unregister(&ppc460ex_adma_driver);
+ * capabilities are enabled then we'll get src/dst filled with zero.
+ */
+static int ppc460ex_test_raid6(struct ppc460ex_adma_chan *chan)
+{
+ struct ppc460ex_adma_desc_slot *sw_desc, *iter;
+ struct page *pg;
+ char *a;
+ unsigned long op = 0;
+ int rval = 0;
+ dma_addr_t dma_addr, addrs[2];;
+
+ if (!ppc460ex_r6_tchan)
+ return -1;
+
+ set_bit(PPC460EX_DESC_WXOR, &op);
+
+ pg = alloc_page(GFP_KERNEL);
+ if (!pg)
+ return -ENOMEM;
+
+ spin_lock_bh(&chan->lock);
+ sw_desc = ppc460ex_adma_alloc_slots(chan, 1, 1);
+ if (sw_desc) {
+ /* 1 src, 1 dsr, int_ena, WXOR */
+ ppc460ex_desc_init_pq(sw_desc, 1, 1, 1, op);
+ list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+ ppc460ex_desc_set_byte_count(iter, chan, PAGE_SIZE);
+ iter->unmap_len = PAGE_SIZE;
+ }
+ } else {
+ rval = -EFAULT;
+ spin_unlock_bh(&chan->lock);
+ goto exit;
+ }
+ spin_unlock_bh(&chan->lock);
+
+ /* Fill the test page with ones */
+ memset(page_address(pg), 0xFF, PAGE_SIZE);
+ dma_addr = dma_map_page(chan->device->dev, pg, 0, PAGE_SIZE,
+ DMA_BIDIRECTIONAL);
+
+ /* Setup adresses */
+ ppc460ex_adma_pqxor_set_src(sw_desc, dma_addr, 0);
+ ppc460ex_adma_pqxor_set_src_mult(sw_desc, 1, 0, 0);
+ addrs[0] = dma_addr;
+ addrs[1] = 0;
+ ppc460ex_adma_pqxor_set_dest(sw_desc, addrs, DMA_PREP_PQ_DISABLE_Q);
+
+ async_tx_ack(&sw_desc->async_tx);
+ sw_desc->async_tx.callback = ppc460ex_test_rad6_callback;
+ sw_desc->async_tx.callback_param = NULL;
+
+ init_completion(&ppc460ex_r6_test_comp);
+
+ ppc460ex_adma_tx_submit(&sw_desc->async_tx);
+ ppc460ex_adma_issue_pending(&chan->common);
+
+ wait_for_completion(&ppc460ex_r6_test_comp);
+
+ /* Now check is the test page zeroed */
+ a = page_address(pg);
+ if ((*(u32 *)a) == 0 && memcmp(a, a+4, PAGE_SIZE-4) == 0) {
+ /* page is zero - RAID-6 enabled */
+ rval = 0;
+ } else {
+ /* RAID-6 was not enabled */
+ rval = -EINVAL;
+ }
+exit:
+ __free_page(pg);
+ return rval;
+}
+/*
+ * ppc460ex_test_raid5 - test are RAID-5 capabilities enabled successfully.
+ * For this we just perform one WXOR operation with the same source
+ * and destination addresses, the GF-multiplier is 1; so if RAID-5
+ o/of_platform_driver_unregister(&ppc460ex_adma_driver);
+ * capabilities are enabled then we'll get src/dst filled with zero.
+ */
+static int ppc460ex_test_raid5(struct ppc460ex_adma_chan *chan)
+{
+ struct ppc460ex_adma_desc_slot *sw_desc, *iter;
+ struct page *pg;
+ char *a;
+ dma_addr_t dma_addr, addrs[2];
+ unsigned long op = 0;
+ int rval = 0;
+
+ if (!ppc460ex_r5_tchan)
+ return -1;
+
+ set_bit(PPC460EX_DESC_WXOR, &op);
+
+ pg = alloc_page(GFP_KERNEL);
+ if (!pg)
+ return -ENOMEM;
+
+ spin_lock_bh(&chan->lock);
+ sw_desc = ppc460ex_adma_alloc_slots(chan, 1, 1);
+ if (sw_desc) {
+ /* 1 src, 1 dsr, int_ena, WXOR */
+ ppc460ex_desc_init_pq(sw_desc, 1, 1, 1, op);
+ list_for_each_entry(iter, &sw_desc->group_list, chain_node) {
+ ppc460ex_desc_set_byte_count(iter, chan, PAGE_SIZE);
+ iter->unmap_len = PAGE_SIZE;
+ }
+ } else {
+ rval = -EFAULT;
+ spin_unlock_bh(&chan->lock);
+ goto exit;
+ }
+ spin_unlock_bh(&chan->lock);
+
+ /* Fill the test page with ones */
+ memset(page_address(pg), 0xFF, PAGE_SIZE);
+ dma_addr = dma_map_page(chan->device->dev, pg, 0, PAGE_SIZE,
+ DMA_BIDIRECTIONAL);
+
+ /* Setup adresses */
+ ppc460ex_adma_pqxor_set_src(sw_desc, dma_addr, 0);
+ ppc460ex_adma_pqxor_set_src_mult(sw_desc, 1, 0, 0);
+ addrs[0] = dma_addr;
+ addrs[1] = 0;
+ ppc460ex_adma_pqxor_set_dest(sw_desc, addrs, DMA_PREP_PQ_DISABLE_Q);
+
+ async_tx_ack(&sw_desc->async_tx);
+ sw_desc->async_tx.callback = ppc460ex_test_raid5_callback;
+ sw_desc->async_tx.callback_param = NULL;
+
+ init_completion(&ppc460ex_r5_test_comp);
+
+ ppc460ex_adma_tx_submit(&sw_desc->async_tx);
+ ppc460ex_adma_issue_pending(&chan->common);
+
+ wait_for_completion(&ppc460ex_r5_test_comp);
+
+ /*Make sure cache is flushed to memory*/
+ dma_addr = dma_map_page(chan->device->dev, pg, 0, PAGE_SIZE,
+ DMA_BIDIRECTIONAL);
+ /* Now check is the test page zeroed */
+ a = page_address(pg);
+ if ((*(u32 *)a) == 0 && memcmp(a, a+4, PAGE_SIZE-4) == 0) {
+ /* page is zero - RAID-5 enabled */
+ rval = 0;
+ } else {
+ /* RAID-5 was not enabled */
+ rval = -EINVAL;
+ }
+exit:
+ __free_page(pg);
+ return rval;
+}
+/*
+ * /sysfs interface
+ */
+static ssize_t ppc460ex_poly_read(struct device_driver *dev, char *buf)
+{
+ ssize_t size = 0;
+ u32 reg;
+
+ reg = dcr_read(ppc460ex_mq_dcr_host, DCRN_MQ0_CFBHL);
+ reg >>= MQ0_CFBHL_POLY;
+ reg &= 0xFF;
+
+ size = sprintf(buf, PAGE_SIZE,
+ "PPC460EX RAID-6 driver uses 0x1%02x polynomial.\n",
+ reg);
+ return size;
+}
+
+static ssize_t ppc460ex_poly_write(struct device_driver *dev,
+ const char *buf, size_t count)
+{
+ unsigned long val, rval;
+
+ if (!count || count > 6)
+ return -EINVAL;
+
+ sscanf(buf, "%lx", &val);
+ if (val & ~0x1FF)
+ return -EINVAL;
+
+ val &= 0xFF;
+ rval = dcr_read(ppc460ex_mq_dcr_host, DCRN_MQ0_CFBHL);
+ rval &= ~(0xFF << MQ0_CFBHL_POLY);
+ rval |= val << MQ0_CFBHL_POLY;
+ dcr_write(ppc460ex_mq_dcr_host, DCRN_MQ0_CFBHL, rval);
+
+ return count;
+}
+static ssize_t show_ppc460ex_devices(struct device_driver *dev, char *buf)
+{
+ ssize_t size = 0;
+ int i;
+
+ for (i = 0; i < PPC460EX_ADMA_ENGINES_NUM; i++) {
+ if (ppc460ex_adma_devices[i] == -1)
+ continue;
+ size += snprintf(buf + size, PAGE_SIZE - size,
+ "PPC460EX-ADMA.%d: %s\n", i,
+ ppc_adma_errors[ppc460ex_adma_devices[i]]);
+ }
+ return size;
+}
+static ssize_t ppc460ex_r6ena_read(struct device_driver *dev, char *buf)
+{
+ ssize_t size = 0;
+ size = snprintf(buf, PAGE_SIZE,
+ "PPC460EX RAID-6 capabilities are %sABLED.\n",
+ ppc460ex_r6_enabled ? "EN" : "DIS");
+ return size;
+}
+
+static ssize_t ppc460ex_r6ena_write(struct device_driver *dev,
+ const char *buf, size_t count)
+{
+ unsigned long val;
+
+ if (!count || count > 11)
+ return -EINVAL;
+
+ if (!ppc460ex_r6_tchan)
+ return -EFAULT;
+
+ /* Write a key */
+ sscanf(buf, "%lx", &val);
+ dcr_write(ppc460ex_mq_dcr_host, DCRN_MQ0_XORBA, val);
+ isync();
+
+ /* Verify does it really work now */
+ if (ppc460ex_test_raid6(ppc460ex_r6_tchan) == 0) {
+ /*
+ * PPC460Ex RAID-6 has been activated successfully
+ */;
+ dev_dbg(dev, "PPC460Ex RAID-6 has been activated "
+ "successfully\n");
+ ppc460ex_r6_enabled = 1;
+ } else {
+ /*
+ * PPC460Ex RAID-6 hasn't been activated! Error key ?
+ */
+ dev_dbg(dev, "PPC460Ex RAID-6 hasn't been activated!"
+ " Error key ?\n");
+ ppc460ex_r6_enabled = 0;
+ }
+
+ return count;
+}
+static ssize_t ppc460ex_r5ena_read(struct device_driver *dev, char *buf)
+{
+ ssize_t size = 0;
+ u32 reg;
+ reg = dcr_read(ppc460ex_mq_dcr_host, DCRN_MQ0_CFBHL);
+ reg >>= MQ0_CFBHL_POLY;
+ reg &= 0xFF;
+ size = snprintf(buf, PAGE_SIZE,
+ "PPC460EX RAID-5 capabilities are %sABLED.\n",
+ ppc460ex_r6_enabled ? "EN" : "DIS");
+ return size;
+}
+
+static ssize_t ppc460ex_r5ena_write(struct device_driver *dev,
+ const char *buf, size_t count)
+{
+ unsigned long val;
+
+ if (!count || count > 11)
+ return -EINVAL;
+
+ if (!ppc460ex_r6_tchan)
+ return -EFAULT;
+
+ /* Write a key */
+ sscanf(buf, "%lx", &val);
+ dcr_write(ppc460ex_mq_dcr_host, DCRN_MQ0_XORBA, val);
+ isync();
+
+ /* Verify does it really work now */
+ if (ppc460ex_test_raid5(ppc460ex_r5_tchan) == 0) {
+ /*
+ * PPC460Ex RAID-5 has been activated successfully
+ */
+ dev_dbg(dev, "PPC460Ex RAID-5 has been activated "
+ "successfully\n");
+ ppc460ex_r5_enabled = 1;
+ } else {
+ /*
+ * PPC460Ex RAID-5 hasn't been activated! Error key ?
+ */
+ dev_dbg(dev, "PPC460Ex RAID-5 hasn't been activated!"
+ " Error key ?\n");
+ ppc460ex_r5_enabled = 0;
+ }
+
+ return count;
+}
+static DRIVER_ATTR(devices, S_IRUGO, show_ppc460ex_devices, NULL);
+static DRIVER_ATTR(raid6_enable, S_IRUGO | S_IWUSR, ppc460ex_r6ena_read,
+ ppc460ex_r6ena_write);
+static DRIVER_ATTR(poly, S_IRUGO | S_IWUSR, ppc460ex_poly_read,
+ ppc460ex_poly_write);
+static DRIVER_ATTR(raid5_enable, S_IRUGO | S_IWUSR, ppc460ex_r5ena_read,
+ ppc460ex_r5ena_write);
+static void ppc460ex_adma_init_capabilities(struct ppc460ex_adma_device *adev)
+{
+ dma_cap_set(DMA_MEMCPY, adev->common.cap_mask);
+ dma_cap_set(DMA_INTERRUPT, adev->common.cap_mask);
+ dma_cap_set(DMA_MEMSET, adev->common.cap_mask);
+ dma_cap_set(DMA_PQ, adev->common.cap_mask);
+ dma_cap_set(DMA_PQ_VAL, adev->common.cap_mask);
+ dma_cap_set(DMA_XOR_VAL, adev->common.cap_mask);
+
+ /* Set base routines */
+
+ adev->common.device_alloc_chan_resources =
+ ppc460ex_adma_alloc_chan_resources;
+ adev->common.device_free_chan_resources =
+ ppc460ex_adma_free_chan_resources;
+ adev->common.device_tx_status = ppc460ex_adma_tx_status;
+ adev->common.device_issue_pending =
+ ppc460ex_adma_issue_pending;
+
+
+ /*Setup routines based on capability*/
+ if (dma_has_cap(DMA_MEMCPY, adev->common.cap_mask)) {
+ adev->common.device_prep_dma_memcpy =
+ ppc460ex_adma_prep_dma_memcpy;
+ }
+ if (dma_has_cap(DMA_MEMSET, adev->common.cap_mask)) {
+ adev->common.device_prep_dma_memset =
+ ppc460ex_adma_prep_dma_memset;
+ }
+ if (dma_has_cap(DMA_XOR, adev->common.cap_mask)) {
+ adev->common.max_xor = XOR_MAX_OPS;
+ adev->common.device_prep_dma_xor =
+ ppc460ex_adma_prep_dma_mq_xor;
+ }
+ if (dma_has_cap(DMA_PQ, adev->common.cap_mask)) {
+ dma_set_maxpq(&adev->common,
+ DMA0_FIFO_SIZE / sizeof(struct dma_cdb), 0);
+ adev->common.device_prep_dma_pq =
+ ppc460ex_adma_prep_dma_pq;
+ }
+ if (dma_has_cap(DMA_PQ_VAL, adev->common.cap_mask)) {
+ adev->common.max_pq = DMA0_FIFO_SIZE / sizeof(struct dma_cdb);
+ adev->common.device_prep_dma_pq_val =
+ ppc460ex_adma_prep_dma_pqzero_sum;
+ }
+ if (dma_has_cap(DMA_XOR_VAL, adev->common.cap_mask)) {
+ adev->common.max_xor = DMA0_FIFO_SIZE /
+ sizeof(struct dma_cdb);
+ adev->common.device_prep_dma_xor_val =
+ ppc460ex_adma_prep_dma_xor_zero_sum;
+ }
+ if (dma_has_cap(DMA_INTERRUPT, adev->common.cap_mask)) {
+ adev->common.device_prep_dma_interrupt =
+ ppc460ex_adma_prep_dma_interrupt;
+ }
+ pr_info("%s: APM ppc460ex ADMA engine:"
+ "( %s%s%s%s%s%s%s)\n",
+ dev_name(adev->dev),
+ dma_has_cap(DMA_PQ, adev->common.cap_mask) ?
+ "pq " : "",
+ dma_has_cap(DMA_PQ_VAL, adev->common.cap_mask) ?
+ "pq_val " : "",
+ dma_has_cap(DMA_XOR, adev->common.cap_mask) ?
+ "xor " : "",
+ dma_has_cap(DMA_XOR_VAL, adev->common.cap_mask) ?
+ "xor_val " : "",
+ dma_has_cap(DMA_MEMCPY, adev->common.cap_mask) ?
+ "memcpy " : "",
+ dma_has_cap(DMA_MEMSET, adev->common.cap_mask) ?
+ "memset " : "",
+ dma_has_cap(DMA_INTERRUPT, adev->common.cap_mask) ?
+ "intr " : "");
+}
+
+static int ppc460ex_setup_irqs(struct ppc460ex_adma_device *adev,
+ struct ppc460ex_adma_chan *chan, int *initcode)
+{
+ struct device_node *np;
+ int ret;
+
+ np = adev->dev->of_node;
+
+ adev->err_irq = irq_of_parse_and_map(np, 2);
+ if (adev->err_irq == NO_IRQ) {
+ dev_warn(adev->dev, "no err irq resource?\n");
+ *initcode = PPC_ADMA_INIT_IRQ2;
+ adev->err_irq = -ENXIO;
+ } else
+ atomic_inc(&ppc460ex_adma_err_irq_ref);
+
+ adev->irq = irq_of_parse_and_map(np, 0);
+ if (adev->irq == NO_IRQ) {
+ dev_err(adev->dev, "no irq resource\n");
+ *initcode = PPC_ADMA_INIT_IRQ1;
+ ret = -ENXIO;
+ goto err_irq_map;
+ }
+ dev_dbg(adev->dev, "irq %d, err irq %d\n",
+ adev->irq, adev->err_irq);
+ ret = request_irq(adev->irq, ppc460ex_adma_eot_handler,
+ 0, dev_driver_string(adev->dev), chan);
+ if (ret) {
+ dev_err(adev->dev, "can't request irq %d\n",
+ adev->irq);
+ *initcode = PPC_ADMA_INIT_IRQ1;
+ ret = -EIO;
+ goto err_req1;
+ }
+ if (adev->err_irq > 0) {
+ u32 mask, enable;
+ ret = request_irq(adev->err_irq,
+ ppc460ex_adma_err_handler,
+ IRQF_SHARED,
+ dev_driver_string(adev->dev),
+ chan);
+ if (ret) {
+ dev_err(adev->dev, "can't request irq %d\n",
+ adev->err_irq);
+ *initcode = PPC_ADMA_INIT_IRQ2;
+ ret = -EIO;
+ goto err_req2;
+ }
+ np = of_find_compatible_node(NULL, NULL, "ibm,i2o-460ex");
+ if (!np) {
+ pr_err("%s: can't find I2O device tree node\n",
+ __func__);
+ return -ENODEV;
+ }
+ adev->i2o_reg = of_iomap(np, 0);
+ if (!adev->i2o_reg) {
+ pr_err("%s: failed to map I2O registers\n", __func__);
+ of_node_put(np);
+ ret = -EINVAL;
+ goto err_req2;
+ }
+ of_node_put(np);
+ /* Unmask 'CS FIFO Attention' interrupts and
+ * enable generating interrupts on errors
+ */
+ enable = ~(I2O_IOPIM_P1EM | I2O_IOPIM_P1SNE);
+ mask = ioread32(&adev->i2o_reg->iopim) & enable;
+ iowrite32(mask, &adev->i2o_reg->iopim);
+ }
+ return 0;
+
+err_req2:
+ free_irq(adev->irq, chan);
+err_req1:
+ irq_dispose_mapping(adev->irq);
+err_irq_map:
+ if (adev->err_irq > 0) {
+ if (atomic_dec_and_test(&ppc460ex_adma_err_irq_ref))
+ irq_dispose_mapping(adev->err_irq);
+ }
+ return ret;
+}
+static void ppc460ex_adma_release_irqs(struct ppc460ex_adma_device *adev,
+ struct ppc460ex_adma_chan *chan)
+{
+ u32 mask;
+
+ /* disable DMAx engine interrupts */
+ mask = ioread32(&adev->i2o_reg->iopim) | I2O_IOPIM_P1SNE
+ | I2O_IOPIM_P1EM ;
+ iowrite32(mask, &adev->i2o_reg->iopim);
+
+ free_irq(adev->irq, chan);
+ irq_dispose_mapping(adev->irq);
+ if (adev->err_irq > 0) {
+ free_irq(adev->err_irq, chan);
+ if (atomic_dec_and_test(&ppc460ex_adma_err_irq_ref)) {
+ irq_dispose_mapping(adev->err_irq);
+ iounmap(adev->i2o_reg);
+ }
+ }
+}
+/*
+ * ppc460ex_adma_remove - remove the asynch device
+ */
+static int __devexit ppc460ex_adma_remove(struct of_device *ofdev)
+{
+ struct ppc460ex_adma_device *device = dev_get_drvdata(&ofdev->dev);
+ struct device_node *np = ofdev->dev.of_node;
+ struct resource res;
+ struct dma_chan *chan, *_chan;
+ struct ppc460ex_adma_chan *ppc460ex_chan;
+
+ dev_set_drvdata(&ofdev->dev, NULL);
+
+ dma_async_device_unregister(&device->common);
+
+ dma_free_coherent(device->dev, device->pool_size,
+ device->dma_desc_pool_virt, device->dma_desc_pool);
+
+ list_for_each_entry_safe(chan, _chan, &device->common.channels,
+ device_node) {
+ ppc460ex_chan = to_ppc460ex_adma_chan(chan);
+ dma_unmap_page(&ofdev->dev, ppc460ex_chan->pdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ dma_unmap_page(&ofdev->dev, ppc460ex_chan->qdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ __free_page(ppc460ex_chan->pdest_page);
+ __free_page(ppc460ex_chan->qdest_page);
+ list_del(&chan->device_node);
+ kfree(ppc460ex_chan);
+ }
+
+ dma_free_coherent(device->dev, device->pool_size,
+ device->dma_desc_pool_virt, device->dma_desc_pool);
+
+
+ iounmap(device->dma_reg);
+ of_address_to_resource(np, 0, &res);
+ release_mem_region(res.start, resource_size(&res));
+ kfree(device);
+ return 0;
+}
+/*
+ * ppc460ex_adma_probe - Probe the DMA engine for features.
+ */
+static int __devinit ppc460ex_adma_probe(struct of_device *ofdev,
+ const struct of_device_id *match)
+{
+ struct device_node *np = ofdev->dev.of_node;
+ struct resource res;
+ struct ppc_dma_chan_ref *ref, *_ref;
+ struct ppc460ex_adma_device *adev;
+ struct ppc460ex_adma_chan *chan;
+ int ret = 0, initcode = PPC_ADMA_INIT_OK;
+ void *regs;
+
+ if (of_address_to_resource(np, 0, &res)) {
+ dev_err(&ofdev->dev, "Failed to get memory resource\n");
+ ret = -ENODEV;
+ goto out;
+ }
+
+ if (!request_mem_region(res.start, resource_size(&res),
+ dev_driver_string(&ofdev->dev))) {
+ dev_err(&ofdev->dev, "failed to request memory region "
+ "(0x%016llx-0x%016llx)\n",
+ (u64)res.start, (u64)res.end);
+ ret = -EBUSY;
+ goto out;
+ }
+
+ /* create a device */
+ adev = kzalloc(sizeof(*adev), GFP_KERNEL);
+ if (!adev) {
+ dev_err(&ofdev->dev, "failed to allocate device\n");
+ initcode = PPC_ADMA_INIT_ALLOC;
+ ret = -ENOMEM;
+ goto err_adev_alloc;
+ }
+
+ adev->id = 0;
+ adev->pool_size = DMA_FIFO_SIZE << 2;
+ /*
+ * allocate coherent memory for hardware descriptors
+ */
+ adev->dma_desc_pool_virt = dma_alloc_coherent(&ofdev->dev,
+ adev->pool_size, &adev->dma_desc_pool,
+ GFP_KERNEL);
+ if (adev->dma_desc_pool_virt == NULL) {
+ dev_err(&ofdev->dev, "failed to allocate %d bytes of coherent "
+ "memory for hardware descriptors\n",
+ adev->pool_size);
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ regs = ioremap(res.start, resource_size(&res));
+ if (!regs) {
+ dev_err(&ofdev->dev, "failed to ioremap regs!\n");
+ goto err_regs_alloc;
+ }
+
+ adev->dma_reg = regs;
+ /*
+ *DMA FIFO length = CSlength + CPlength;
+ */
+ iowrite32(DMA_FIFO_ENABLE | ((DMA_FIFO_SIZE >> 3) - 2),
+ &adev->dma_reg->fsiz);
+ /* Configure DMA engine */
+ iowrite32(DMA_CFG_DXEPR_HP | DMA_CFG_DFMPP_HP | DMA_CFG_FALGN,
+ &adev->dma_reg->cfg);
+ /* Clear Status */
+ iowrite32(~0, &adev->dma_reg->dsts);
+
+ adev->dev = &ofdev->dev;
+ adev->common.dev = &ofdev->dev;
+ INIT_LIST_HEAD(&adev->common.channels);
+ dev_set_drvdata(&ofdev->dev, adev);
+
+ /* create a channel */
+ chan = kzalloc(sizeof(*chan), GFP_KERNEL);
+ if (!chan) {
+ dev_err(&ofdev->dev, "can't allocate channel structure\n");
+ ret = -ENOMEM;
+ goto err_chan_alloc;
+ }
+ spin_lock_init(&chan->lock);
+ INIT_LIST_HEAD(&chan->chain);
+ INIT_LIST_HEAD(&chan->all_slots);
+ chan->device = adev;
+ chan->common.device = &adev->common;
+ list_add_tail(&chan->common.device_node, &adev->common.channels);
+ tasklet_init(&chan->irq_tasklet, ppc460ex_adma_tasklet,
+ (unsigned long)chan);
+ /*
+ * allocate and map helper pages for async validation or
+ * async_mult/async_sum_product operations on DMA0/1.
+ */
+ chan->pdest_page = alloc_page(GFP_KERNEL);
+ chan->qdest_page = alloc_page(GFP_KERNEL);
+ if (!chan->pdest_page ||
+ !chan->qdest_page) {
+ if (chan->pdest_page)
+ __free_page(chan->pdest_page);
+ if (chan->qdest_page)
+ __free_page(chan->qdest_page);
+ ret = -ENOMEM;
+ goto err_page_alloc;
+ }
+ chan->pdest = dma_map_page(&ofdev->dev, chan->pdest_page, 0,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ chan->qdest = dma_map_page(&ofdev->dev, chan->qdest_page, 0,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+
+ ref = kmalloc(sizeof(*ref), GFP_KERNEL);
+ if (ref) {
+ ref->chan = &chan->common;
+ INIT_LIST_HEAD(&ref->node);
+ list_add_tail(&ref->node, &ppc460ex_adma_chan_list);
+ } else {
+ dev_err(&ofdev->dev, "failed to allocate channel reference!\n");
+ ret = -ENOMEM;
+ goto err_ref_alloc;
+ }
+
+
+ ret = ppc460ex_setup_irqs(adev, chan, &initcode);
+ if (ret)
+ goto err_irq;
+ ppc460ex_adma_init_capabilities(adev);
+
+ ret = dma_async_device_register(&adev->common);
+ if (ret) {
+ dev_err(&ofdev->dev, "failed to register dma device\n");
+ goto err_dev_reg;
+ }
+ goto out;
+
+err_dev_reg:
+ ppc460ex_adma_release_irqs(adev, chan);
+err_irq:
+ list_for_each_entry_safe(ref, _ref, &ppc460ex_adma_chan_list, node) {
+ if (chan == to_ppc460ex_adma_chan(ref->chan)) {
+ list_del(&ref->node);
+ kfree(ref);
+ }
+ }
+
+err_ref_alloc:
+ dma_unmap_page(&ofdev->dev, chan->pdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ dma_unmap_page(&ofdev->dev, chan->qdest,
+ PAGE_SIZE, DMA_BIDIRECTIONAL);
+ __free_page(chan->pdest_page);
+ __free_page(chan->qdest_page);
+
+err_page_alloc:
+ kfree(chan);
+err_chan_alloc:
+ iounmap(adev->dma_reg);
+err_regs_alloc:
+ dma_free_coherent(adev->dev, adev->pool_size,
+ adev->dma_desc_pool_virt,
+ adev->dma_desc_pool);
+err_dma_alloc:
+ kfree(adev);
+err_adev_alloc:
+ release_mem_region(res.start, resource_size(&res));
+out:
+ return ret;
+
+}
+/*
+ * Create sys fs entries to enable DMA engine and select RAID-6 Poly
+ * nomial
+ */
+/*
+ *One time initialization which would decide some proparties like
+ *FIO depth, priority LL HB buses etc.
+ */
+static int ppc460ex_configure_raid_devices(void)
+{
+ struct device_node *np;
+ struct resource i2o_res;
+ struct i2o_regs __iomem *i2o_reg;
+ dcr_host_t i2o_dcr_host;
+ unsigned int dcr_base, dcr_len;
+ int i, ret;
+
+ np = of_find_compatible_node(NULL, NULL, "ibm,i2o-460ex");
+ if (!np) {
+ pr_err("%s: can't find I2O device tree node\n",
+ __func__);
+ return -ENODEV;
+ }
+
+ if (of_address_to_resource(np, 0, &i2o_res)) {
+ of_node_put(np);
+ return -EINVAL;
+ }
+
+ i2o_reg = of_iomap(np, 0);
+ if (!i2o_reg) {
+ pr_err("%s: failed to map I2O registers\n", __func__);
+ of_node_put(np);
+ return -EINVAL;
+ }
+
+ /* Get I2O DCRs base */
+ dcr_base = dcr_resource_start(np, 0);
+ dcr_len = dcr_resource_len(np, 0);
+ if (!dcr_base && !dcr_len) {
+ pr_err("%s: can't get DCR registers base/len!\n",
+ np->full_name);
+ of_node_put(np);
+ iounmap(i2o_reg);
+ return -ENODEV;
+ }
+
+ i2o_dcr_host = dcr_map(np, dcr_base, dcr_len);
+ if (!DCR_MAP_OK(i2o_dcr_host)) {
+ pr_err("%s: failed to map DCRs!\n", np->full_name);
+ of_node_put(np);
+ iounmap(i2o_reg);
+ return -ENODEV;
+ }
+ of_node_put(np);
+
+ ppc460ex_dma_fifo_buf = kmalloc(DMA_FIFO_SIZE, GFP_KERNEL);
+ if (!ppc460ex_dma_fifo_buf) {
+ pr_err("%s: DMA FIFO buffer allocation failed\n", __func__);
+ iounmap(i2o_reg);
+ dcr_unmap(i2o_dcr_host, dcr_len);
+ return -ENOMEM;
+ }
+
+ /*
+ * Confgiure HW
+ */
+ /* reset DMA */
+ mtdcri(SDR0, DCRN_SDR0_SRST, DCRN_SDR0_SRST_I2ODMA);
+ mtdcri(SDR0, DCRN_SDR0_SRST, 0);
+
+ /* Setup the base address of mmapped registers */
+ dcr_write(i2o_dcr_host, DCRN_I2O0_IBAH, (u32)(i2o_res.start >> 32));
+ dcr_write(i2o_dcr_host, DCRN_I2O0_IBAL, (u32)(i2o_res.start) |
+ I2O_REG_ENABLE);
+ dcr_unmap(i2o_dcr_host, dcr_len);
+
+ /* Setup FIFO memory space base address */
+ iowrite32(0, &i2o_reg->ifbah);
+ iowrite32(((u32)__pa(ppc460ex_dma_fifo_buf)), &i2o_reg->ifbal);
+
+ /* set zero FIFO size for I2O, so the whole
+ * ppc460ex_dma_fifo_buf is used by DMAs.
+ * DMAx_FIFOs will be configured while probe.
+ */
+ iowrite32(0, &i2o_reg->ifsiz);
+ iounmap(i2o_reg);
+
+ /* To prepare WXOR/RXOR functionality we need access to
+ * Memory Queue Module DCRs (finally it will be enabled
+ * via /sys interface of the ppc460ex ADMA driver).
+ */
+ np = of_find_compatible_node(NULL, NULL, "ibm,mq-460ex");
+ if (!np) {
+ pr_err("%s: can't find MQ device tree node\n",
+ __func__);
+ ret = -ENODEV;
+ goto out_free;
+ }
+
+ /* Get MQ DCRs base */
+ dcr_base = dcr_resource_start(np, 0);
+ dcr_len = dcr_resource_len(np, 0);
+ if (!dcr_base && !dcr_len) {
+ pr_err("%s: can't get DCR registers base/len!\n",
+ np->full_name);
+ ret = -ENODEV;
+ goto out_mq;
+ }
+
+ ppc460ex_mq_dcr_host = dcr_map(np, dcr_base, dcr_len);
+ if (!DCR_MAP_OK(ppc460ex_mq_dcr_host)) {
+ pr_err("%s: failed to map DCRs!\n", np->full_name);
+ ret = -ENODEV;
+ goto out_mq;
+ }
+ of_node_put(np);
+ ppc460ex_mq_dcr_len = dcr_len;
+
+ /* Set HB alias */
+ dcr_write(ppc460ex_mq_dcr_host, DCRN_MQ0_BAUH, DMA_CUED_XOR_HB);
+
+ /* Set:
+ * - LL transaction passing limit to 1;
+ * - Memory controller cycle limit to 1;
+ * - Galois Polynomial to 0x14d (default)
+ */
+ dcr_write(ppc460ex_mq_dcr_host, DCRN_MQ0_CFBHL,
+ (1 << MQ0_CFBHL_TPLM) | (1 << MQ0_CFBHL_HBCL) |
+ (PPC460EX_DEFAULT_POLY << MQ0_CFBHL_POLY));
+
+ atomic_set(&ppc460ex_adma_err_irq_ref, 0);
+ for (i = 0; i < PPC460EX_ADMA_ENGINES_NUM; i++)
+ ppc460ex_adma_devices[i] = -1;
+
+ return 0;
+
+out_mq:
+ of_node_put(np);
+out_free:
+ kfree(ppc460ex_dma_fifo_buf);
+ return ret;
+
+}
+
+static struct of_device_id adma_match[] = {
+ {
+ .compatible = "amcc,dma-460ex",
+ },
+ {},
+};
+static struct of_platform_driver ppc460ex_adma_driver = {
+ .probe = ppc460ex_adma_probe,
+ .remove = ppc460ex_adma_remove,
+ .driver = {
+ .name = "PPC460Ex-ADMA",
+ .owner = THIS_MODULE,
+ .of_match_table = adma_match,
+ },
+};
+static int __init ppc460ex_adma_init(void)
+{
+ int rval;
+
+ rval = ppc460ex_configure_raid_devices();
+ if (rval)
+ return rval;
+
+ rval = of_register_platform_driver(&ppc460ex_adma_driver);
+ if (rval) {
+ pr_err("%s: Driver reigstration failed\n", __func__);
+ goto out_reg;
+ }
+ /* Initialization status */
+ rval = driver_create_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_devices);
+ if (rval)
+ goto out_dev;
+
+ /* RAID-6 h/w enable entry */
+ rval = driver_create_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_raid6_enable);
+ if (rval)
+ goto out_en;
+ /* RAID-5 h/w enable entry */
+ rval = driver_create_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_raid5_enable);
+ if (rval)
+ goto out_en;
+
+ /* GF polynomial to use */
+ rval = driver_create_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_poly);
+ if (!rval)
+ return rval;
+
+out_en:
+ driver_remove_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_devices);
+out_dev:
+ /* User will not be able to enable h/w RAID-6 */
+ pr_err("%s: failed to create RAID-6 driver interface\n",
+ __func__);
+ of_unregister_platform_driver(&ppc460ex_adma_driver);
+out_reg:
+ dcr_unmap(ppc460ex_mq_dcr_host , ppc460ex_mq_dcr_len);
+ kfree(ppc460ex_dma_fifo_buf);
+ return rval;
+
+
+}
+static void __exit ppc460ex_adma_exit(void)
+{
+ driver_remove_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_poly);
+ driver_remove_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_raid5_enable);
+ driver_remove_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_raid6_enable);
+ driver_remove_file(&ppc460ex_adma_driver.driver,
+ &driver_attr_devices);
+ of_unregister_platform_driver(&ppc460ex_adma_driver);
+ dcr_unmap(ppc460ex_mq_dcr_host, ppc460ex_mq_dcr_len);
+}
+arch_initcall(ppc460ex_adma_init);
+module_exit(ppc460ex_adma_exit);
+
+MODULE_AUTHOR(" Tirumala R Marri <[email protected]>");
+MODULE_DESCRIPTION(" PPC460Ex ADMA Engine Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/dma/ppc4xx/adma1.h b/drivers/dma/ppc4xx/adma1.h
new file mode 100644
index 0000000..7a71f8d
--- /dev/null
+++ b/drivers/dma/ppc4xx/adma1.h
@@ -0,0 +1,192 @@
+/*
+ * 2010 (C) Applied Micro(APM).
+ *
+ * Author: Tirumala R Marri <[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public License
+ * version 2. This program is licensed "as is" without any warranty of
+ * any kind, whether express or implied.
+ */
+#ifndef PPC460EX_ADMA_H
+#define PPC460EX_ADMA_H
+
+#include <linux/types.h>
+#include "dma.h"
+
+#define to_ppc460ex_adma_chan(chan) \
+ container_of(chan, struct ppc460ex_adma_chan, common)
+#define to_ppc460ex_adma_device(dev) \
+ container_of(dev, struct ppc460ex_adma_device, common)
+#define tx_to_ppc460ex_adma_slot(tx) \
+ container_of(tx, struct ppc460ex_adma_desc_slot, async_tx)
+
+#define PPC460EX_R6_PROC_ROOT "driver/460ex_raid6"
+#define PPC460EX_R5_PROC_ROOT "driver/460ex_raid5"
+
+#define PPC460EX_DEFAULT_POLY 0x4d
+
+#define PPC460EX_ADMA_ENGINES_NUM 1
+#define PPC460EX_ADMA_WATCHDOG_MSEC 3
+#define PPC460EX_ADMA_THRESHOLD 1
+
+#define XOR_MAX_OPS 16
+
+
+#define PPC460EX_ADMA_DMA_MAX_BYTE_COUNT 0xFFFFFFUL
+/* this is the XOR_CBBCR width */
+#define PPC460EX_ADMA_XOR_MAX_BYTE_COUNT (1 << 31)
+#define PPC460EX_ADMA_ZERO_SUM_MAX_BYTE_COUNT PPC460EX_ADMA_XOR_MAX_BYTE_COUNT
+
+#define DMA_FIFO_SIZE 0x1000
+
+#define PPC460EX_RXOR_RUN 0
+#define MQ0_CF2H_RXOR_BS_MASK 0x1FF
+
+#define DMA_ZERO_P 7
+
+/**
+ * struct ppc460ex_adma_device - internal representation of an ADMA device
+ * @dev: device
+ * @dma_reg: DMA register base
+ * @i2o_reg: I2O register base
+ * @id: HW ADMA Device selector
+ * @dma_desc_pool_virt: base of DMA descriptor region (CPU address)
+ * @dma_desc_pool: base of DMA descriptor region (DMA address)
+ * @pool_size: Size of the descriptor pool.
+ * @irq: DMA completion interrupt
+ * @err_irq: DMA error interrupt
+ * @common: embedded struct dma_device
+ */
+struct ppc460ex_adma_device {
+ struct device *dev;
+ struct dma_regs __iomem *dma_reg;
+ struct i2o_regs __iomem *i2o_reg;
+ int id;
+ void *dma_desc_pool_virt;
+ dma_addr_t dma_desc_pool;
+ size_t pool_size;
+ int irq;
+ int err_irq;
+ struct dma_device common;
+};
+
+/**
+ * struct ppc460ex_adma_chan - internal representation of an ADMA channel
+ * @lock: serializes enqueue/dequeue operations to the slot pool
+ * @device: parent device
+ * @chain: device chain view of the descriptors
+ * @common: common dmaengine channel object members
+ * @all_slots: complete domain of slots usable by the channel
+ * @last_used:
+ * @pending: allows batching of hardware operations
+ * @completed_cookie: identifier for the most recently completed operation
+ * @slots_allocated: records the actual size of the descriptor slot pool
+ * @hw_chain_inited: h/w descriptor chain initialization flag
+ * @irq_tasklet: bottom half where ppc460ex_adma_slot_cleanup runs
+ * @needs_unmap: if buffers should not be unmapped upon final processing
+ * @pdest_page: P destination page for async validate operation
+ * @qdest_page: Q destination page for async validate operation
+ * @pdest: P dma addr for async validate operation
+ * @qdest: Q dma addr for async validate operation
+ */
+struct ppc460ex_adma_chan {
+ spinlock_t lock;
+ struct ppc460ex_adma_device *device;
+ struct list_head chain;
+ struct dma_chan common;
+ struct list_head all_slots;
+ struct ppc460ex_adma_desc_slot *last_used;
+ int pending;
+ dma_cookie_t completed_cookie;
+ int slots_allocated;
+ int hw_chain_inited;
+ struct tasklet_struct irq_tasklet;
+ u8 needs_unmap;
+ struct page *pdest_page;
+ struct page *qdest_page;
+ dma_addr_t pdest;
+ dma_addr_t qdest;
+};
+
+struct ppc460ex_rxor {
+ u32 addrl;
+ u32 addrh;
+ int len;
+ int xor_count;
+ int addr_count;
+ int desc_count;
+ int state;
+};
+
+/**
+ * struct ppc460ex_adma_desc_slot - PPC460EX-ADMA software descriptor
+ * @phys: hardware address of the hardware descriptor chain
+ * @group_head: first operation in a transaction
+ * @hw_next: pointer to the next descriptor in chain
+ * @async_tx: support for the async_tx api
+ * @slot_node: node on the iop_adma_chan.all_slots list
+ * @chain_node: node on the op_adma_chan.chain list
+ * @group_list: list of slots that make up a multi-descriptor transaction
+ * for example transfer lengths larger than the supported hw max
+ * @unmap_len: transaction bytecount
+ * @hw_desc: virtual address of the hardware descriptor chain
+ * @stride: currently chained or not
+ * @idx: pool index
+ * @slot_cnt: total slots used in an transaction (group of operations)
+ * @src_cnt: number of sources set in this descriptor
+ * @dst_cnt: number of destinations set in the descriptor
+ * @slots_per_op: number of slots per operation
+ * @descs_per_op: number of slot per P/Q operation see comment
+ * for ppc460ex_prep_dma_pqxor function
+ * @flags: desc state/type
+ * @reverse_flags: 1 if a corresponding rxor address uses reversed address order
+ * @xor_check_result: result of zero sum
+ * @crc32_result: result crc calculation
+ */
+struct ppc460ex_adma_desc_slot {
+ dma_addr_t phys;
+ struct ppc460ex_adma_desc_slot *group_head;
+ struct ppc460ex_adma_desc_slot *hw_next;
+ struct dma_async_tx_descriptor async_tx;
+ struct list_head slot_node;
+ struct list_head chain_node; /* node in channel ops list */
+ struct list_head group_list; /* list */
+ unsigned int unmap_len;
+ void *hw_desc;
+ u16 stride;
+ u16 idx;
+ u16 slot_cnt;
+ u8 src_cnt;
+ u8 dst_cnt;
+ u8 slots_per_op;
+ u8 descs_per_op;
+ unsigned long flags;
+ unsigned long reverse_flags[8];
+
+#define PPC460EX_DESC_INT 0 /* generate interrupt on complete */
+#define PPC460EX_ZERO_P 1 /* clear P destionaion */
+#define PPC460EX_ZERO_Q 2 /* clear Q destination */
+#define PPC460EX_COHERENT 3 /* src/dst are coherent */
+
+#define PPC460EX_DESC_WXOR 4 /* WXORs are in chain */
+#define PPC460EX_DESC_RXOR 5 /* RXOR is in chain */
+
+#define PPC460EX_DESC_RXOR123 8 /* CDB for RXOR123 operation */
+#define PPC460EX_DESC_RXOR124 9 /* CDB for RXOR124 operation */
+#define PPC460EX_DESC_RXOR125 10 /* CDB for RXOR125 operation */
+#define PPC460EX_DESC_RXOR12 11 /* CDB for RXOR12 operation */
+#define PPC460EX_DESC_RXOR_REV 12 /* CDB cont srcs in reversed order*/
+
+#define PPC460EX_DESC_PCHECK 13
+#define PPC460EX_DESC_QCHECK 14
+
+#define PPC460EX_DESC_RXOR_MSK 0x3
+
+
+ union {
+ u32 *xor_check_result;
+ u32 *crc32_result;
+ };
+};
+
+#endif /* PPC460EX_ADMA_H*/
diff --git a/drivers/dma/ppc4xx/dma.h b/drivers/dma/ppc4xx/dma.h
index bcde2df..9c05b1f 100644
--- a/drivers/dma/ppc4xx/dma.h
+++ b/drivers/dma/ppc4xx/dma.h
@@ -10,11 +10,23 @@
* kind, whether express or implied.
*/

-#ifndef _PPC440SPE_DMA_H
-#define _PPC440SPE_DMA_H
+#ifndef _PPC4XX_DMA_H
+#define _PPC4XX_DMA_H

#include <linux/types.h>

+#if defined(CONFIG_PPC460EX)
+
+/* Number of elements in the array with statical CDBs */
+#define MAX_STAT_DMA_CDBS 16
+/* Number of DMA engines available on the contoller */
+#define DMA_ENGINES_NUM 1
+/* Maximum h/w supported number of destinations */
+#define DMA_DEST_MAX_NUM 2
+
+#define DMA_FIFO_SIZE 0x1000
+#else
+
/* Number of elements in the array with statical CDBs */
#define MAX_STAT_DMA_CDBS 16
/* Number of DMA engines available on the contoller */
@@ -57,6 +69,8 @@
#define I2O_IOPIM_P1SNE (1<<6)
#define I2O_IOPIM_P1EM (1<<8)

+#endif /*defined(CONFIG_460EX)*/
+
/* DMA CDB fields */
#define DMA_CDB_MSK (0xF)
#define DMA_CDB_64B_ADDR (1<<2)
@@ -220,4 +234,4 @@ struct i2o_regs {
u32 iopt;
};

-#endif /* _PPC440SPE_DMA_H */
+#endif /* _PPC4XX_DMA_H */
--
1.6.1.rc3


2010-07-23 06:15:14

by Stefan Roese

[permalink] [raw]
Subject: Re: [PATCH] Adding ADMA support for PPC460EX DMA engine.

Hi Marri,

On Friday 23 July 2010 02:57:18 [email protected] wrote:
> From: Tirumala Marri <[email protected]>
>
> This patch will add ADMA support for DMA engine and HW offload for
> XOR/ADG (RAID-5/6) functionalities.
> 1. It supports memcpy, xor, GF(2) based RAID-6.
> 2. It supports interrupt based DMA completions.
> 3. Also supports memcpy in RAID-1 case.
>
> Kernel version: 2.6.35-rc5
>
> Testing:
> Created RAID-5/6 arrays usign mdadm.
> And ran raw IO and filesystem IO to the RAID array.
> Chunk size 4k,64k was tested.
> RAID rebuild , disk fail, resync tested.
>
> File names:
> This code is similar to ppc440spe . So I named the files as
> drivers/dma/ppc4xx/adma1.c and drivers/dma/ppc4xx/adma1.h

As you describe above, a lot of the code seems to be copied from
drivers/dma/ppc4xx/adma.c/h. Wouldn't it make more sense to factor out the
common code instead of duplicating it?

Thanks.

Cheers,
Stefan

--
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-0 Fax: (+49)-8142-66989-80 Email: [email protected]

2010-07-23 19:21:08

by Dan Williams

[permalink] [raw]
Subject: Re: [PATCH] Adding ADMA support for PPC460EX DMA engine.

On Thu, Jul 22, 2010 at 11:15 PM, Stefan Roese <[email protected]> wrote:
> Hi Marri,
>
> On Friday 23 July 2010 02:57:18 [email protected] wrote:
>> From: Tirumala Marri <[email protected]>
>>
>> ? This patch will add ADMA support for DMA engine and HW offload for
>> ? XOR/ADG (RAID-5/6) functionalities.
>> ? 1. It supports memcpy, xor, GF(2) based RAID-6.
>> ? 2. It supports interrupt based DMA completions.
>> ? 3. Also supports memcpy in RAID-1 case.
>>
>> ? Kernel version: 2.6.35-rc5
>>
>> ? Testing:
>> ? ? Created RAID-5/6 arrays usign mdadm.
>> ? ? And ran raw IO and filesystem IO to the RAID array.
>> ? ? Chunk size 4k,64k was tested.
>> ? ? RAID rebuild , disk fail, resync tested.
>>
>> ? File names:
>> ? ? This code is similar to ppc440spe . So I named the files as
>> ? ? drivers/dma/ppc4xx/adma1.c and drivers/dma/ppc4xx/adma1.h
>
> As you describe above, a lot of the code seems to be copied from
> drivers/dma/ppc4xx/adma.c/h. Wouldn't it make more sense to factor out the
> common code instead of duplicating it?
>

Yes, and you might look to drivers/dma/iop-adma.c as an example of a
way to support similar hardware with a single code base.

--
Dan

2010-07-23 21:39:24

by Tirumala Marri

[permalink] [raw]
Subject: RE: [PATCH] Adding ADMA support for PPC460EX DMA engine.

>As you describe above, a lot of the code seems to be copied from
>drivers/dma/ppc4xx/adma.c/h. Wouldn't it make more sense to factor out
the
>common code instead of duplicating it?



Hi Stefan,
Thanks for the review. There are definitely some functions can be
moved to a common file.

Hi Dan,
Could you also please review and see if there are any changes needed, so
I can include some changes as
Well in the modified patch.


Regards,
Marri