[PATCH 0/5] powerpc: Get rid of redundant arch specific swab functions

[PATCH 0/5] powerpc: Get rid of redundant arch specific swab functions

[David Gibson]

arch/powerpc/include/asm/swab.h includes some powerpc specific
byteswapping functions, which are implemented in terms of powerpc's
built in byte reversed load/store instructions.  There are two problems with this:

1) They're not necessary - gcc is perfectly capable of generating the
   byte-reversed load and store instructions when using the normal,
   generic byteswapping functions (tested with gcc (GCC) 4.8.3
   20140911 (Red Hat 4.8.3-9))

2) They've become poorly named.  The ld_le*() and st_le*() functions
   in fact *always* byteswap, even in a little-endian powerpc kernel
   build, in which case they'll actually be performing BE accesses.

This series removes the existing users of these arch-specific
functions, replacing them with calls to the generic byteswappers.  5/5
then removes the function definitions.

I've compile tested this series with pmac32_defconfig,
mpc512x_defconfig and ppc64_defconfig, and also those configs tweaked
to explicitly enable the BT8XX and MXC MMC drivers where possible.

I've tested the KVM patch (4/5) with both BE and LE guests, however I
don't have hardware to do any real testing of the drivers affected in
1..3/5.

David Gibson (5):
  powerpc: Move Power Macintosh drivers to generic byteswappers
  powerpc: Remove powerpc specific byteswap from bt8xx DVB driver
  powerpc: Remove arch specific byteswappers from the MXC MMC driver
  powerpc: Cleanup KVM emulated load/store endian handling
  powerpc: Remove unused st_le*() and ld_le* functions

 arch/powerpc/include/asm/dbdma.h    | 12 ++++----
 arch/powerpc/include/asm/kvm_host.h |  2 +-
 arch/powerpc/include/asm/swab.h     | 26 -----------------
 arch/powerpc/include/asm/vga.h      |  4 +--
 arch/powerpc/kvm/powerpc.c          | 38 ++++++++++++------------
 drivers/ata/pata_macio.c            | 10 +++----
 drivers/block/swim3.c               | 12 ++++----
 drivers/ide/pmac.c                  | 10 +++----
 drivers/macintosh/rack-meter.c      | 30 +++++++++----------
 drivers/media/pci/bt8xx/bt878.h     |  4 +--
 drivers/mmc/host/mxcmmc.c           |  2 +-
 drivers/net/ethernet/apple/bmac.c   | 30 +++++++++----------
 drivers/net/ethernet/apple/mace.c   | 44 ++++++++++++++--------------
 drivers/scsi/mac53c94.c             | 10 +++----
 drivers/scsi/mesh.c                 | 14 ++++-----
 drivers/video/fbdev/controlfb.c     |  2 +-
 drivers/video/fbdev/platinumfb.c    |  2 +-
 sound/ppc/pmac.c                    | 58 ++++++++++++++++++-------------------
 18 files changed, 141 insertions(+), 169 deletions(-)

-- 
2.1.0

[PATCH 1/5] powerpc: Move Power Macintosh drivers to generic byteswappers

[David Gibson]

ppc has special instruction forms to efficiently load and store values
in non-native endianness.  These can be accessed via the arch-specific
{ld,st}_le{16,32}() inlines in arch/powerpc/include/asm/swab.h.

However, gcc is perfectly capable of generating the byte-reversing
load/store instructions when using the normal, generic cpu_to_le*() and
le*_to_cpu() functions eaning the arch-specific functions don't have much
point.

Worse the "le" in the names of the arch specific functions is now
misleading, because they always generate byte-reversing forms, but some
ppc machines can now run a little-endian kernel.

To start getting rid of the arch-specific forms, this patch removes them
from all the old Power Macintosh drivers, replacing them with the
generic byteswappers.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
 arch/powerpc/include/asm/dbdma.h  | 12 ++++----
 arch/powerpc/include/asm/vga.h    |  4 +--
 drivers/ata/pata_macio.c          | 10 +++----
 drivers/block/swim3.c             | 12 ++++----
 drivers/ide/pmac.c                | 10 +++----
 drivers/macintosh/rack-meter.c    | 30 ++++++++++----------
 drivers/net/ethernet/apple/bmac.c | 30 ++++++++++----------
 drivers/net/ethernet/apple/mace.c | 44 ++++++++++++++---------------
 drivers/scsi/mac53c94.c           | 10 +++----
 drivers/scsi/mesh.c               | 14 +++++-----
 drivers/video/fbdev/controlfb.c   |  2 +-
 drivers/video/fbdev/platinumfb.c  |  2 +-
 sound/ppc/pmac.c                  | 58 +++++++++++++++++++--------------------
 13 files changed, 119 insertions(+), 119 deletions(-)

diff --git a/arch/powerpc/include/asm/dbdma.h b/arch/powerpc/include/asm/dbdma.h
index e23f07e..6c69836 100644
--- a/arch/powerpc/include/asm/dbdma.h
+++ b/arch/powerpc/include/asm/dbdma.h
@@ -42,12 +42,12 @@ struct dbdma_regs {
  * DBDMA command structure.  These fields are all little-endian!
  */
 struct dbdma_cmd {
-    unsigned short req_count;	/* requested byte transfer count */
-    unsigned short command;	/* command word (has bit-fields) */
-    unsigned int   phy_addr;	/* physical data address */
-    unsigned int   cmd_dep;	/* command-dependent field */
-    unsigned short res_count;	/* residual count after completion */
-    unsigned short xfer_status;	/* transfer status */
+	__le16 req_count;	/* requested byte transfer count */
+	__le16 command;		/* command word (has bit-fields) */
+	__le32 phy_addr;	/* physical data address */
+	__le32 cmd_dep;		/* command-dependent field */
+	__le16 res_count;	/* residual count after completion */
+	__le16 xfer_status;	/* transfer status */
 };
 
 /* DBDMA command values in command field */
diff --git a/arch/powerpc/include/asm/vga.h b/arch/powerpc/include/asm/vga.h
index e5f8dd3..ab3acd2 100644
--- a/arch/powerpc/include/asm/vga.h
+++ b/arch/powerpc/include/asm/vga.h
@@ -25,12 +25,12 @@
 
 static inline void scr_writew(u16 val, volatile u16 *addr)
 {
-    st_le16(addr, val);
+	*addr = cpu_to_le16(val);
 }
 
 static inline u16 scr_readw(volatile const u16 *addr)
 {
-    return ld_le16(addr);
+	return le16_to_cpu(*addr);
 }
 
 #define VT_BUF_HAVE_MEMCPYW
diff --git a/drivers/ata/pata_macio.c b/drivers/ata/pata_macio.c
index a02f76f..b002858 100644
--- a/drivers/ata/pata_macio.c
+++ b/drivers/ata/pata_macio.c
@@ -540,9 +540,9 @@ static void pata_macio_qc_prep(struct ata_queued_cmd *qc)
 			BUG_ON (pi++ >= MAX_DCMDS);
 
 			len = (sg_len < MAX_DBDMA_SEG) ? sg_len : MAX_DBDMA_SEG;
-			st_le16(&table->command, write ? OUTPUT_MORE: INPUT_MORE);
-			st_le16(&table->req_count, len);
-			st_le32(&table->phy_addr, addr);
+			table->command = cpu_to_le16(write ? OUTPUT_MORE: INPUT_MORE);
+			table->req_count = cpu_to_le16(len);
+			table->phy_addr = cpu_to_le32(addr);
 			table->cmd_dep = 0;
 			table->xfer_status = 0;
 			table->res_count = 0;
@@ -557,12 +557,12 @@ static void pata_macio_qc_prep(struct ata_queued_cmd *qc)
 
 	/* Convert the last command to an input/output */
 	table--;
-	st_le16(&table->command, write ? OUTPUT_LAST: INPUT_LAST);
+	table->command = cpu_to_le16(write ? OUTPUT_LAST: INPUT_LAST);
 	table++;
 
 	/* Add the stop command to the end of the list */
 	memset(table, 0, sizeof(struct dbdma_cmd));
-	st_le16(&table->command, DBDMA_STOP);
+	table->command = cpu_to_le16(DBDMA_STOP);
 
 	dev_dbgdma(priv->dev, "%s: %d DMA list entries\n", __func__, pi);
 }
diff --git a/drivers/block/swim3.c b/drivers/block/swim3.c
index 523ee8f..c264f2d 100644
--- a/drivers/block/swim3.c
+++ b/drivers/block/swim3.c
@@ -440,9 +440,9 @@ static inline void seek_track(struct floppy_state *fs, int n)
 static inline void init_dma(struct dbdma_cmd *cp, int cmd,
 			    void *buf, int count)
 {
-	st_le16(&cp->req_count, count);
-	st_le16(&cp->command, cmd);
-	st_le32(&cp->phy_addr, virt_to_bus(buf));
+	cp->req_count = cpu_to_le16(count);
+	cp->command = cpu_to_le16(cmd);
+	cp->phy_addr = cpu_to_le32(virt_to_bus(buf));
 	cp->xfer_status = 0;
 }
 
@@ -771,8 +771,8 @@ static irqreturn_t swim3_interrupt(int irq, void *dev_id)
 		}
 		/* turn off DMA */
 		out_le32(&dr->control, (RUN | PAUSE) << 16);
-		stat = ld_le16(&cp->xfer_status);
-		resid = ld_le16(&cp->res_count);
+		stat = le16_to_cpu(cp->xfer_status);
+		resid = le16_to_cpu(cp->res_count);
 		if (intr & ERROR_INTR) {
 			n = fs->scount - 1 - resid / 512;
 			if (n > 0) {
@@ -1170,7 +1170,7 @@ static int swim3_add_device(struct macio_dev *mdev, int index)
 
 	fs->dma_cmd = (struct dbdma_cmd *) DBDMA_ALIGN(fs->dbdma_cmd_space);
 	memset(fs->dma_cmd, 0, 2 * sizeof(struct dbdma_cmd));
-	st_le16(&fs->dma_cmd[1].command, DBDMA_STOP);
+	fs->dma_cmd[1].command = cpu_to_le16(DBDMA_STOP);
 
 	if (mdev->media_bay == NULL || check_media_bay(mdev->media_bay) == MB_FD)
 		swim3_mb_event(mdev, MB_FD);
diff --git a/drivers/ide/pmac.c b/drivers/ide/pmac.c
index 2db803c..d24a3f8 100644
--- a/drivers/ide/pmac.c
+++ b/drivers/ide/pmac.c
@@ -1497,9 +1497,9 @@ static int pmac_ide_build_dmatable(ide_drive_t *drive, struct ide_cmd *cmd)
 				       drive->name);
 				return 0;
 			}
-			st_le16(&table->command, wr? OUTPUT_MORE: INPUT_MORE);
-			st_le16(&table->req_count, tc);
-			st_le32(&table->phy_addr, cur_addr);
+			table->command = cpu_to_le16(wr? OUTPUT_MORE: INPUT_MORE);
+			table->req_count = cpu_to_le16(tc);
+			table->phy_addr = cpu_to_le32(cur_addr);
 			table->cmd_dep = 0;
 			table->xfer_status = 0;
 			table->res_count = 0;
@@ -1513,10 +1513,10 @@ static int pmac_ide_build_dmatable(ide_drive_t *drive, struct ide_cmd *cmd)
 
 	/* convert the last command to an input/output last command */
 	if (count) {
-		st_le16(&table[-1].command, wr? OUTPUT_LAST: INPUT_LAST);
+		table[-1].command = cpu_to_le16(wr? OUTPUT_LAST: INPUT_LAST);
 		/* add the stop command to the end of the list */
 		memset(table, 0, sizeof(struct dbdma_cmd));
-		st_le16(&table->command, DBDMA_STOP);
+		table->command = cpu_to_le16(DBDMA_STOP);
 		mb();
 		writel(hwif->dmatable_dma, &dma->cmdptr);
 		return 1;
diff --git a/drivers/macintosh/rack-meter.c b/drivers/macintosh/rack-meter.c
index 4192901..048901a 100644
--- a/drivers/macintosh/rack-meter.c
+++ b/drivers/macintosh/rack-meter.c
@@ -182,31 +182,31 @@ static void rackmeter_setup_dbdma(struct rackmeter *rm)
 
 	/* Prepare 4 dbdma commands for the 2 buffers */
 	memset(cmd, 0, 4 * sizeof(struct dbdma_cmd));
-	st_le16(&cmd->req_count, 4);
-	st_le16(&cmd->command, STORE_WORD | INTR_ALWAYS | KEY_SYSTEM);
-	st_le32(&cmd->phy_addr, rm->dma_buf_p +
+	cmd->req_count = cpu_to_le16(4);
+	cmd->command = cpu_to_le16(STORE_WORD | INTR_ALWAYS | KEY_SYSTEM);
+	cmd->phy_addr = cpu_to_le32(rm->dma_buf_p +
 		offsetof(struct rackmeter_dma, mark));
-	st_le32(&cmd->cmd_dep, 0x02000000);
+	cmd->cmd_dep = cpu_to_le32(0x02000000);
 	cmd++;
 
-	st_le16(&cmd->req_count, SAMPLE_COUNT * 4);
-	st_le16(&cmd->command, OUTPUT_MORE);
-	st_le32(&cmd->phy_addr, rm->dma_buf_p +
+	cmd->req_count = cpu_to_le16(SAMPLE_COUNT * 4);
+	cmd->command = cpu_to_le16(OUTPUT_MORE);
+	cmd->phy_addr = cpu_to_le32(rm->dma_buf_p +
 		offsetof(struct rackmeter_dma, buf1));
 	cmd++;
 
-	st_le16(&cmd->req_count, 4);
-	st_le16(&cmd->command, STORE_WORD | INTR_ALWAYS | KEY_SYSTEM);
-	st_le32(&cmd->phy_addr, rm->dma_buf_p +
+	cmd->req_count = cpu_to_le16(4);
+	cmd->command = cpu_to_le16(STORE_WORD | INTR_ALWAYS | KEY_SYSTEM);
+	cmd->phy_addr = cpu_to_le32(rm->dma_buf_p +
 		offsetof(struct rackmeter_dma, mark));
-	st_le32(&cmd->cmd_dep, 0x01000000);
+	cmd->cmd_dep = cpu_to_le32(0x01000000);
 	cmd++;
 
-	st_le16(&cmd->req_count, SAMPLE_COUNT * 4);
-	st_le16(&cmd->command, OUTPUT_MORE | BR_ALWAYS);
-	st_le32(&cmd->phy_addr, rm->dma_buf_p +
+	cmd->req_count = cpu_to_le16(SAMPLE_COUNT * 4);
+	cmd->command = cpu_to_le16(OUTPUT_MORE | BR_ALWAYS);
+	cmd->phy_addr = cpu_to_le32(rm->dma_buf_p +
 		offsetof(struct rackmeter_dma, buf2));
-	st_le32(&cmd->cmd_dep, rm->dma_buf_p);
+	cmd->cmd_dep = cpu_to_le32(rm->dma_buf_p);
 
 	rackmeter_do_pause(rm, 0);
 }
diff --git a/drivers/net/ethernet/apple/bmac.c b/drivers/net/ethernet/apple/bmac.c
index daae0e0..c0bd638 100644
--- a/drivers/net/ethernet/apple/bmac.c
+++ b/drivers/net/ethernet/apple/bmac.c
@@ -483,8 +483,8 @@ static int bmac_suspend(struct macio_dev *mdev, pm_message_t state)
        		bmwrite(dev, TXCFG, (config & ~TxMACEnable));
 		bmwrite(dev, INTDISABLE, DisableAll); /* disable all intrs */
        		/* disable rx and tx dma */
-       		st_le32(&rd->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
-       		st_le32(&td->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
+		rd->control = cpu_to_le32(DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
+		td->control = cpu_to_le32(DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
        		/* free some skb's */
        		for (i=0; i<N_RX_RING; i++) {
        			if (bp->rx_bufs[i] != NULL) {
@@ -699,8 +699,8 @@ static irqreturn_t bmac_rxdma_intr(int irq, void *dev_id)
 
 	while (1) {
 		cp = &bp->rx_cmds[i];
-		stat = ld_le16(&cp->xfer_status);
-		residual = ld_le16(&cp->res_count);
+		stat = le16_to_cpu(cp->xfer_status);
+		residual = le16_to_cpu(cp->res_count);
 		if ((stat & ACTIVE) == 0)
 			break;
 		nb = RX_BUFLEN - residual - 2;
@@ -728,8 +728,8 @@ static irqreturn_t bmac_rxdma_intr(int irq, void *dev_id)
 				skb_reserve(bp->rx_bufs[i], 2);
 		}
 		bmac_construct_rxbuff(skb, &bp->rx_cmds[i]);
-		st_le16(&cp->res_count, 0);
-		st_le16(&cp->xfer_status, 0);
+		cp->res_count = cpu_to_le16(0);
+		cp->xfer_status = cpu_to_le16(0);
 		last = i;
 		if (++i >= N_RX_RING) i = 0;
 	}
@@ -769,7 +769,7 @@ static irqreturn_t bmac_txdma_intr(int irq, void *dev_id)
 
 	while (1) {
 		cp = &bp->tx_cmds[bp->tx_empty];
-		stat = ld_le16(&cp->xfer_status);
+		stat = le16_to_cpu(cp->xfer_status);
 		if (txintcount < 10) {
 			XXDEBUG(("bmac_txdma_xfer_stat=%#0x\n", stat));
 		}
@@ -1411,8 +1411,8 @@ static int bmac_close(struct net_device *dev)
 	bmwrite(dev, INTDISABLE, DisableAll); /* disable all intrs */
 
 	/* disable rx and tx dma */
-	st_le32(&rd->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
-	st_le32(&td->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
+	rd->control = cpu_to_le32(DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
+	td->control = cpu_to_le32(DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE));	/* clear run bit */
 
 	/* free some skb's */
 	XXDEBUG(("bmac: free rx bufs\n"));
@@ -1493,7 +1493,7 @@ static void bmac_tx_timeout(unsigned long data)
 
 	cp = &bp->tx_cmds[bp->tx_empty];
 /*	XXDEBUG((KERN_DEBUG "bmac: tx dmastat=%x %x runt=%d pr=%x fs=%x fc=%x\n", */
-/* 	   ld_le32(&td->status), ld_le16(&cp->xfer_status), bp->tx_bad_runt, */
+/* 	   le32_to_cpu(td->status), le16_to_cpu(cp->xfer_status), bp->tx_bad_runt, */
 /* 	   mb->pr, mb->xmtfs, mb->fifofc)); */
 
 	/* turn off both tx and rx and reset the chip */
@@ -1506,7 +1506,7 @@ static void bmac_tx_timeout(unsigned long data)
 	bmac_enable_and_reset_chip(dev);
 
 	/* restart rx dma */
-	cp = bus_to_virt(ld_le32(&rd->cmdptr));
+	cp = bus_to_virt(le32_to_cpu(rd->cmdptr));
 	out_le32(&rd->control, DBDMA_CLEAR(RUN|PAUSE|FLUSH|WAKE|ACTIVE|DEAD));
 	out_le16(&cp->xfer_status, 0);
 	out_le32(&rd->cmdptr, virt_to_bus(cp));
@@ -1553,10 +1553,10 @@ static void dump_dbdma(volatile struct dbdma_cmd *cp,int count)
 		ip = (int*)(cp+i);
 
 		printk("dbdma req 0x%x addr 0x%x baddr 0x%x xfer/res 0x%x\n",
-		       ld_le32(ip+0),
-		       ld_le32(ip+1),
-		       ld_le32(ip+2),
-		       ld_le32(ip+3));
+		       le32_to_cpup(ip+0),
+		       le32_to_cpup(ip+1),
+		       le32_to_cpup(ip+2),
+		       le32_to_cpup(ip+3));
 	}
 
 }
diff --git a/drivers/net/ethernet/apple/mace.c b/drivers/net/ethernet/apple/mace.c
index 842fe76..73afe49 100644
--- a/drivers/net/ethernet/apple/mace.c
+++ b/drivers/net/ethernet/apple/mace.c
@@ -310,7 +310,7 @@ static void dbdma_reset(volatile struct dbdma_regs __iomem *dma)
      * way on some machines.
      */
     for (i = 200; i > 0; --i)
-	if (ld_le32(&dma->control) & RUN)
+	if (le32_to_cpu(dma->control) & RUN)
 	    udelay(1);
 }
 
@@ -452,21 +452,21 @@ static int mace_open(struct net_device *dev)
 	    data = skb->data;
 	}
 	mp->rx_bufs[i] = skb;
-	st_le16(&cp->req_count, RX_BUFLEN);
-	st_le16(&cp->command, INPUT_LAST + INTR_ALWAYS);
-	st_le32(&cp->phy_addr, virt_to_bus(data));
+	cp->req_count = cpu_to_le16(RX_BUFLEN);
+	cp->command = cpu_to_le16(INPUT_LAST + INTR_ALWAYS);
+	cp->phy_addr = cpu_to_le32(virt_to_bus(data));
 	cp->xfer_status = 0;
 	++cp;
     }
     mp->rx_bufs[i] = NULL;
-    st_le16(&cp->command, DBDMA_STOP);
+    cp->command = cpu_to_le16(DBDMA_STOP);
     mp->rx_fill = i;
     mp->rx_empty = 0;
 
     /* Put a branch back to the beginning of the receive command list */
     ++cp;
-    st_le16(&cp->command, DBDMA_NOP + BR_ALWAYS);
-    st_le32(&cp->cmd_dep, virt_to_bus(mp->rx_cmds));
+    cp->command = cpu_to_le16(DBDMA_NOP + BR_ALWAYS);
+    cp->cmd_dep = cpu_to_le32(virt_to_bus(mp->rx_cmds));
 
     /* start rx dma */
     out_le32(&rd->control, (RUN|PAUSE|FLUSH|WAKE) << 16); /* clear run bit */
@@ -475,8 +475,8 @@ static int mace_open(struct net_device *dev)
 
     /* put a branch at the end of the tx command list */
     cp = mp->tx_cmds + NCMDS_TX * N_TX_RING;
-    st_le16(&cp->command, DBDMA_NOP + BR_ALWAYS);
-    st_le32(&cp->cmd_dep, virt_to_bus(mp->tx_cmds));
+    cp->command = cpu_to_le16(DBDMA_NOP + BR_ALWAYS);
+    cp->cmd_dep = cpu_to_le32(virt_to_bus(mp->tx_cmds));
 
     /* reset tx dma */
     out_le32(&td->control, (RUN|PAUSE|FLUSH|WAKE) << 16);
@@ -507,8 +507,8 @@ static int mace_close(struct net_device *dev)
     out_8(&mb->imr, 0xff);		/* disable all intrs */
 
     /* disable rx and tx dma */
-    st_le32(&rd->control, (RUN|PAUSE|FLUSH|WAKE) << 16); /* clear run bit */
-    st_le32(&td->control, (RUN|PAUSE|FLUSH|WAKE) << 16); /* clear run bit */
+    rd->control = cpu_to_le32((RUN|PAUSE|FLUSH|WAKE) << 16); /* clear run bit */
+    td->control = cpu_to_le32((RUN|PAUSE|FLUSH|WAKE) << 16); /* clear run bit */
 
     mace_clean_rings(mp);
 
@@ -558,8 +558,8 @@ static int mace_xmit_start(struct sk_buff *skb, struct net_device *dev)
     }
     mp->tx_bufs[fill] = skb;
     cp = mp->tx_cmds + NCMDS_TX * fill;
-    st_le16(&cp->req_count, len);
-    st_le32(&cp->phy_addr, virt_to_bus(skb->data));
+    cp->req_count = cpu_to_le16(len);
+    cp->phy_addr = cpu_to_le32(virt_to_bus(skb->data));
 
     np = mp->tx_cmds + NCMDS_TX * next;
     out_le16(&np->command, DBDMA_STOP);
@@ -691,7 +691,7 @@ static irqreturn_t mace_interrupt(int irq, void *dev_id)
 	    out_8(&mb->xmtfc, AUTO_PAD_XMIT);
 	    continue;
 	}
-	dstat = ld_le32(&td->status);
+	dstat = le32_to_cpu(td->status);
 	/* stop DMA controller */
 	out_le32(&td->control, RUN << 16);
 	/*
@@ -724,7 +724,7 @@ static irqreturn_t mace_interrupt(int irq, void *dev_id)
 		 */
 	}
 	cp = mp->tx_cmds + NCMDS_TX * i;
-	stat = ld_le16(&cp->xfer_status);
+	stat = le16_to_cpu(cp->xfer_status);
 	if ((fs & (UFLO|LCOL|LCAR|RTRY)) || (dstat & DEAD) || xcount == 0) {
 	    /*
 	     * Check whether there were in fact 2 bytes written to
@@ -830,7 +830,7 @@ static void mace_tx_timeout(unsigned long data)
     mace_reset(dev);
 
     /* restart rx dma */
-    cp = bus_to_virt(ld_le32(&rd->cmdptr));
+    cp = bus_to_virt(le32_to_cpu(rd->cmdptr));
     dbdma_reset(rd);
     out_le16(&cp->xfer_status, 0);
     out_le32(&rd->cmdptr, virt_to_bus(cp));
@@ -889,20 +889,20 @@ static irqreturn_t mace_rxdma_intr(int irq, void *dev_id)
     spin_lock_irqsave(&mp->lock, flags);
     for (i = mp->rx_empty; i != mp->rx_fill; ) {
 	cp = mp->rx_cmds + i;
-	stat = ld_le16(&cp->xfer_status);
+	stat = le16_to_cpu(cp->xfer_status);
 	if ((stat & ACTIVE) == 0) {
 	    next = i + 1;
 	    if (next >= N_RX_RING)
 		next = 0;
 	    np = mp->rx_cmds + next;
 	    if (next != mp->rx_fill &&
-		(ld_le16(&np->xfer_status) & ACTIVE) != 0) {
+		(le16_to_cpu(np->xfer_status) & ACTIVE) != 0) {
 		printk(KERN_DEBUG "mace: lost a status word\n");
 		++mace_lost_status;
 	    } else
 		break;
 	}
-	nb = ld_le16(&cp->req_count) - ld_le16(&cp->res_count);
+	nb = le16_to_cpu(cp->req_count) - le16_to_cpu(cp->res_count);
 	out_le16(&cp->command, DBDMA_STOP);
 	/* got a packet, have a look at it */
 	skb = mp->rx_bufs[i];
@@ -962,13 +962,13 @@ static irqreturn_t mace_rxdma_intr(int irq, void *dev_id)
 		mp->rx_bufs[i] = skb;
 	    }
 	}
-	st_le16(&cp->req_count, RX_BUFLEN);
+	cp->req_count = cpu_to_le16(RX_BUFLEN);
 	data = skb? skb->data: dummy_buf;
-	st_le32(&cp->phy_addr, virt_to_bus(data));
+	cp->phy_addr = cpu_to_le32(virt_to_bus(data));
 	out_le16(&cp->xfer_status, 0);
 	out_le16(&cp->command, INPUT_LAST + INTR_ALWAYS);
 #if 0
-	if ((ld_le32(&rd->status) & ACTIVE) != 0) {
+	if ((le32_to_cpu(rd->status) & ACTIVE) != 0) {
 	    out_le32(&rd->control, (PAUSE << 16) | PAUSE);
 	    while ((in_le32(&rd->status) & ACTIVE) != 0)
 		;
diff --git a/drivers/scsi/mac53c94.c b/drivers/scsi/mac53c94.c
index e5cd8d8..0adb2e0 100644
--- a/drivers/scsi/mac53c94.c
+++ b/drivers/scsi/mac53c94.c
@@ -382,16 +382,16 @@ static void set_dma_cmds(struct fsc_state *state, struct scsi_cmnd *cmd)
 		if (dma_len > 0xffff)
 			panic("mac53c94: scatterlist element >= 64k");
 		total += dma_len;
-		st_le16(&dcmds->req_count, dma_len);
-		st_le16(&dcmds->command, dma_cmd);
-		st_le32(&dcmds->phy_addr, dma_addr);
+		dcmds->req_count = cpu_to_le16(dma_len);
+		dcmds->command = cpu_to_le16(dma_cmd);
+		dcmds->phy_addr = cpu_to_le32(dma_addr);
 		dcmds->xfer_status = 0;
 		++dcmds;
 	}
 
 	dma_cmd += OUTPUT_LAST - OUTPUT_MORE;
-	st_le16(&dcmds[-1].command, dma_cmd);
-	st_le16(&dcmds->command, DBDMA_STOP);
+	dcmds[-1].command = cpu_to_le16(dma_cmd);
+	dcmds->command = cpu_to_le16(DBDMA_STOP);
 	cmd->SCp.this_residual = total;
 }
 
diff --git a/drivers/scsi/mesh.c b/drivers/scsi/mesh.c
index 57a95e2..555367f 100644
--- a/drivers/scsi/mesh.c
+++ b/drivers/scsi/mesh.c
@@ -1287,9 +1287,9 @@ static void set_dma_cmds(struct mesh_state *ms, struct scsi_cmnd *cmd)
 				}
 				if (dma_len > 0xffff)
 					panic("mesh: scatterlist element >= 64k");
-				st_le16(&dcmds->req_count, dma_len - off);
-				st_le16(&dcmds->command, dma_cmd);
-				st_le32(&dcmds->phy_addr, dma_addr + off);
+				dcmds->req_count = cpu_to_le16(dma_len - off);
+				dcmds->command = cpu_to_le16(dma_cmd);
+				dcmds->phy_addr = cpu_to_le32(dma_addr + off);
 				dcmds->xfer_status = 0;
 				++dcmds;
 				dtot += dma_len - off;
@@ -1303,15 +1303,15 @@ static void set_dma_cmds(struct mesh_state *ms, struct scsi_cmnd *cmd)
 		static char mesh_extra_buf[64];
 
 		dtot = sizeof(mesh_extra_buf);
-		st_le16(&dcmds->req_count, dtot);
-		st_le32(&dcmds->phy_addr, virt_to_phys(mesh_extra_buf));
+		dcmds->req_count = cpu_to_le16(dtot);
+		dcmds->phy_addr = cpu_to_le32(virt_to_phys(mesh_extra_buf));
 		dcmds->xfer_status = 0;
 		++dcmds;
 	}
 	dma_cmd += OUTPUT_LAST - OUTPUT_MORE;
-	st_le16(&dcmds[-1].command, dma_cmd);
+	dcmds[-1].command = cpu_to_le16(dma_cmd);
 	memset(dcmds, 0, sizeof(*dcmds));
-	st_le16(&dcmds->command, DBDMA_STOP);
+	dcmds->command = cpu_to_le16(DBDMA_STOP);
 	ms->dma_count = dtot;
 }
 
diff --git a/drivers/video/fbdev/controlfb.c b/drivers/video/fbdev/controlfb.c
index 080fdd2..8d14b29 100644
--- a/drivers/video/fbdev/controlfb.c
+++ b/drivers/video/fbdev/controlfb.c
@@ -315,7 +315,7 @@ static int controlfb_blank(int blank_mode, struct fb_info *info)
 		container_of(info, struct fb_info_control, info);
 	unsigned ctrl;
 
-	ctrl = ld_le32(CNTRL_REG(p,ctrl));
+	ctrl = le32_to_cpup(CNTRL_REG(p,ctrl));
 	if (blank_mode > 0)
 		switch (blank_mode) {
 		case FB_BLANK_VSYNC_SUSPEND:
diff --git a/drivers/video/fbdev/platinumfb.c b/drivers/video/fbdev/platinumfb.c
index 518d1fd..377d339 100644
--- a/drivers/video/fbdev/platinumfb.c
+++ b/drivers/video/fbdev/platinumfb.c
@@ -168,7 +168,7 @@ static int platinumfb_blank(int blank,  struct fb_info *fb)
 	struct fb_info_platinum *info = (struct fb_info_platinum *) fb;
 	int	ctrl;
 
-	ctrl = ld_le32(&info->platinum_regs->ctrl.r) | 0x33;
+	ctrl = le32_to_cpup(&info->platinum_regs->ctrl.r) | 0x33;
 	if (blank)
 		--blank_mode;
 	if (blank & VESA_VSYNC_SUSPEND)
diff --git a/sound/ppc/pmac.c b/sound/ppc/pmac.c
index 5a13b22..1b90be5 100644
--- a/sound/ppc/pmac.c
+++ b/sound/ppc/pmac.c
@@ -240,7 +240,7 @@ static int snd_pmac_pcm_prepare(struct snd_pmac *chip, struct pmac_stream *rec,
 	 */
 	spin_lock_irq(&chip->reg_lock);
 	snd_pmac_dma_stop(rec);
-	st_le16(&chip->extra_dma.cmds->command, DBDMA_STOP);
+	chip->extra_dma.cmds->command = cpu_to_le16(DBDMA_STOP);
 	snd_pmac_dma_set_command(rec, &chip->extra_dma);
 	snd_pmac_dma_run(rec, RUN);
 	spin_unlock_irq(&chip->reg_lock);
@@ -251,15 +251,15 @@ static int snd_pmac_pcm_prepare(struct snd_pmac *chip, struct pmac_stream *rec,
 	 */
 	offset = runtime->dma_addr;
 	for (i = 0, cp = rec->cmd.cmds; i < rec->nperiods; i++, cp++) {
-		st_le32(&cp->phy_addr, offset);
-		st_le16(&cp->req_count, rec->period_size);
-		/*st_le16(&cp->res_count, 0);*/
-		st_le16(&cp->xfer_status, 0);
+		cp->phy_addr = cpu_to_le32(offset);
+		cp->req_count = cpu_to_le16(rec->period_size);
+		/*cp->res_count = cpu_to_le16(0);*/
+		cp->xfer_status = cpu_to_le16(0);
 		offset += rec->period_size;
 	}
 	/* make loop */
-	st_le16(&cp->command, DBDMA_NOP + BR_ALWAYS);
-	st_le32(&cp->cmd_dep, rec->cmd.addr);
+	cp->command = cpu_to_le16(DBDMA_NOP + BR_ALWAYS);
+	cp->cmd_dep = cpu_to_le32(rec->cmd.addr);
 
 	snd_pmac_dma_stop(rec);
 	snd_pmac_dma_set_command(rec, &rec->cmd);
@@ -328,7 +328,7 @@ static snd_pcm_uframes_t snd_pmac_pcm_pointer(struct snd_pmac *chip,
 #if 1 /* hmm.. how can we get the current dma pointer?? */
 	int stat;
 	volatile struct dbdma_cmd __iomem *cp = &rec->cmd.cmds[rec->cur_period];
-	stat = ld_le16(&cp->xfer_status);
+	stat = le16_to_cpu(cp->xfer_status);
 	if (stat & (ACTIVE|DEAD)) {
 		count = in_le16(&cp->res_count);
 		if (count)
@@ -427,26 +427,26 @@ static inline void snd_pmac_pcm_dead_xfer(struct pmac_stream *rec,
 		memcpy((void *)emergency_dbdma.cmds, (void *)cp,
 		       sizeof(struct dbdma_cmd));
 		emergency_in_use = 1;
-		st_le16(&cp->xfer_status, 0);
-		st_le16(&cp->req_count, rec->period_size);
+		cp->xfer_status = cpu_to_le16(0);
+		cp->req_count = cpu_to_le16(rec->period_size);
 		cp = emergency_dbdma.cmds;
 	}
 
 	/* now bump the values to reflect the amount
 	   we haven't yet shifted */
-	req = ld_le16(&cp->req_count);
-	res = ld_le16(&cp->res_count);
-	phy = ld_le32(&cp->phy_addr);
+	req = le16_to_cpu(cp->req_count);
+	res = le16_to_cpu(cp->res_count);
+	phy = le32_to_cpu(cp->phy_addr);
 	phy += (req - res);
-	st_le16(&cp->req_count, res);
-	st_le16(&cp->res_count, 0);
-	st_le16(&cp->xfer_status, 0);
-	st_le32(&cp->phy_addr, phy);
+	cp->req_count = cpu_to_le16(res);
+	cp->res_count = cpu_to_le16(0);
+	cp->xfer_status = cpu_to_le16(0);
+	cp->phy_addr = cpu_to_le32(phy);
 
-	st_le32(&cp->cmd_dep, rec->cmd.addr
+	cp->cmd_dep = cpu_to_le32(rec->cmd.addr
 		+ sizeof(struct dbdma_cmd)*((rec->cur_period+1)%rec->nperiods));
 
-	st_le16(&cp->command, OUTPUT_MORE | BR_ALWAYS | INTR_ALWAYS);
+	cp->command = cpu_to_le16(OUTPUT_MORE | BR_ALWAYS | INTR_ALWAYS);
 
 	/* point at our patched up command block */
 	out_le32(&rec->dma->cmdptr, emergency_dbdma.addr);
@@ -475,7 +475,7 @@ static void snd_pmac_pcm_update(struct snd_pmac *chip, struct pmac_stream *rec)
 			else
 				cp = &rec->cmd.cmds[rec->cur_period];
 
-			stat = ld_le16(&cp->xfer_status);
+			stat = le16_to_cpu(cp->xfer_status);
 
 			if (stat & DEAD) {
 				snd_pmac_pcm_dead_xfer(rec, cp);
@@ -489,9 +489,9 @@ static void snd_pmac_pcm_update(struct snd_pmac *chip, struct pmac_stream *rec)
 				break;
 
 			/*printk(KERN_DEBUG "update frag %d\n", rec->cur_period);*/
-			st_le16(&cp->xfer_status, 0);
-			st_le16(&cp->req_count, rec->period_size);
-			/*st_le16(&cp->res_count, 0);*/
+			cp->xfer_status = cpu_to_le16(0);
+			cp->req_count = cpu_to_le16(rec->period_size);
+			/*cp->res_count = cpu_to_le16(0);*/
 			rec->cur_period++;
 			if (rec->cur_period >= rec->nperiods) {
 				rec->cur_period = 0;
@@ -760,11 +760,11 @@ void snd_pmac_beep_dma_start(struct snd_pmac *chip, int bytes, unsigned long add
 	struct pmac_stream *rec = &chip->playback;
 
 	snd_pmac_dma_stop(rec);
-	st_le16(&chip->extra_dma.cmds->req_count, bytes);
-	st_le16(&chip->extra_dma.cmds->xfer_status, 0);
-	st_le32(&chip->extra_dma.cmds->cmd_dep, chip->extra_dma.addr);
-	st_le32(&chip->extra_dma.cmds->phy_addr, addr);
-	st_le16(&chip->extra_dma.cmds->command, OUTPUT_MORE + BR_ALWAYS);
+	chip->extra_dma.cmds->req_count = cpu_to_le16(bytes);
+	chip->extra_dma.cmds->xfer_status = cpu_to_le16(0);
+	chip->extra_dma.cmds->cmd_dep = cpu_to_le32(chip->extra_dma.addr);
+	chip->extra_dma.cmds->phy_addr = cpu_to_le32(addr);
+	chip->extra_dma.cmds->command = cpu_to_le16(OUTPUT_MORE + BR_ALWAYS);
 	out_le32(&chip->awacs->control,
 		 (in_le32(&chip->awacs->control) & ~0x1f00)
 		 | (speed << 8));
@@ -776,7 +776,7 @@ void snd_pmac_beep_dma_start(struct snd_pmac *chip, int bytes, unsigned long add
 void snd_pmac_beep_dma_stop(struct snd_pmac *chip)
 {
 	snd_pmac_dma_stop(&chip->playback);
-	st_le16(&chip->extra_dma.cmds->command, DBDMA_STOP);
+	chip->extra_dma.cmds->command = cpu_to_le16(DBDMA_STOP);
 	snd_pmac_pcm_set_format(chip); /* reset format */
 }
 
-- 
2.1.0

[PATCH 2/5] powerpc: Remove powerpc specific byteswap from bt8xx DVB driver

[David Gibson]

The bt8xx PCI DVB driver includes a powerpc specific hack, using one of
the powerpc specific byteswapping functions in an IO helper macro.

There's no reason to use the powerpc specific function instead of a
generic byteswap, so this patch removes it.  I'm not sure if the powerpc
specific memory barrier is required, so I'm leaving that in.

Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: Peter Hettkamp <peter.hettkamp@htp-tel.de>

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
 drivers/media/pci/bt8xx/bt878.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/media/pci/bt8xx/bt878.h b/drivers/media/pci/bt8xx/bt878.h
index d19b592..bbb76bb 100644
--- a/drivers/media/pci/bt8xx/bt878.h
+++ b/drivers/media/pci/bt8xx/bt878.h
@@ -145,12 +145,12 @@ void bt878_stop(struct bt878 *bt);
 #if defined(__powerpc__)	/* big-endian */
 static inline void io_st_le32(volatile unsigned __iomem *addr, unsigned val)
 {
-	st_le32(addr, val);
+	*addr = cpu_to_le32(val);
 	eieio();
 }
 
 #define bmtwrite(dat,adr)  io_st_le32((adr),(dat))
-#define bmtread(adr)       ld_le32((adr))
+#define bmtread(adr)       le32_to_cpu(*((volatile __le32 *)(adr)))
 #else
 #define bmtwrite(dat,adr)  writel((dat), (adr))
 #define bmtread(adr)       readl(adr)
-- 
2.1.0

[PATCH 3/5] powerpc: Remove arch specific byteswappers from the MXC MMC driver

[David Gibson]

When the MXC MMUC driver is used on a Freescale MPC512x machine, it
contains some additional byteswapping code (I'm assuming this is a
workaround for a hardware defect).  This uses the ppc specific st_le32()
function, but there's no reason not to use the generic swab32() function
instead.  gcc is capable of generating the efficient ppc byte-reversing
load/store instructions without the arch-specific helper.

This patch, therefore, switches to the generic byteswap routine.

Cc: Shawn Guo <shawn.guo@linaro.org>
Cc: Sascha Hauer <kernel@pengutronix.de>

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
 drivers/mmc/host/mxcmmc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/mmc/host/mxcmmc.c b/drivers/mmc/host/mxcmmc.c
index 5316d9b..317d709 100644
--- a/drivers/mmc/host/mxcmmc.c
+++ b/drivers/mmc/host/mxcmmc.c
@@ -281,7 +281,7 @@ static inline void buffer_swap32(u32 *buf, int len)
 	int i;
 
 	for (i = 0; i < ((len + 3) / 4); i++) {
-		st_le32(buf, *buf);
+		*buf = swab32(*buf);
 		buf++;
 	}
 }
-- 
2.1.0

[PATCH 4/5] powerpc: Cleanup KVM emulated load/store endian handling

[David Gibson]

Sometimes the KVM code on powerpc needs to emulate load or store
instructions from the guest, which can include both normal and byte
reversed forms.

We currently (AFAICT) handle this correctly, but some variable names are
very misleading.  In particular we use "is_bigendian" in several places to
actually mean "is the IO the same endian as the host", but we now support
little-endian powerpc hosts.  This also ties into the misleadingly named
ld_le*() and st_le*() functions, which in fact always byteswap, even on
an LE host.

This patch cleans this up by renaming to more accurate "host_swabbed", and
uses the generic swab*() functions instead of the powerpc specific and
misleadingly named ld_le*() and st_le*() functions.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
 arch/powerpc/include/asm/kvm_host.h |  2 +-
 arch/powerpc/kvm/powerpc.c          | 38 ++++++++++++++++++-------------------
 2 files changed, 19 insertions(+), 21 deletions(-)

diff --git a/arch/powerpc/include/asm/kvm_host.h b/arch/powerpc/include/asm/kvm_host.h
index 7efd666a..9b18149 100644
--- a/arch/powerpc/include/asm/kvm_host.h
+++ b/arch/powerpc/include/asm/kvm_host.h
@@ -584,7 +584,7 @@ struct kvm_vcpu_arch {
 	pgd_t *pgdir;
 
 	u8 io_gpr; /* GPR used as IO source/target */
-	u8 mmio_is_bigendian;
+	u8 mmio_host_swabbed;
 	u8 mmio_sign_extend;
 	u8 osi_needed;
 	u8 osi_enabled;
diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
index c45eaab..e115793 100644
--- a/arch/powerpc/kvm/powerpc.c
+++ b/arch/powerpc/kvm/powerpc.c
@@ -721,7 +721,7 @@ static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
 		return;
 	}
 
-	if (vcpu->arch.mmio_is_bigendian) {
+	if (!vcpu->arch.mmio_host_swabbed) {
 		switch (run->mmio.len) {
 		case 8: gpr = *(u64 *)run->mmio.data; break;
 		case 4: gpr = *(u32 *)run->mmio.data; break;
@@ -729,10 +729,10 @@ static void kvmppc_complete_mmio_load(struct kvm_vcpu *vcpu,
 		case 1: gpr = *(u8 *)run->mmio.data; break;
 		}
 	} else {
-		/* Convert BE data from userland back to LE. */
 		switch (run->mmio.len) {
-		case 4: gpr = ld_le32((u32 *)run->mmio.data); break;
-		case 2: gpr = ld_le16((u16 *)run->mmio.data); break;
+		case 8: gpr = swab64(*(u64 *)run->mmio.data); break;
+		case 4: gpr = swab32(*(u32 *)run->mmio.data); break;
+		case 2: gpr = swab16(*(u16 *)run->mmio.data); break;
 		case 1: gpr = *(u8 *)run->mmio.data; break;
 		}
 	}
@@ -781,14 +781,13 @@ int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 		       int is_default_endian)
 {
 	int idx, ret;
-	int is_bigendian;
+	bool host_swabbed;
 
+	/* Pity C doesn't have a logical XOR operator */
 	if (kvmppc_need_byteswap(vcpu)) {
-		/* Default endianness is "little endian". */
-		is_bigendian = !is_default_endian;
+		host_swabbed = is_default_endian;
 	} else {
-		/* Default endianness is "big endian". */
-		is_bigendian = is_default_endian;
+		host_swabbed = !is_default_endian;
 	}
 
 	if (bytes > sizeof(run->mmio.data)) {
@@ -801,7 +800,7 @@ int kvmppc_handle_load(struct kvm_run *run, struct kvm_vcpu *vcpu,
 	run->mmio.is_write = 0;
 
 	vcpu->arch.io_gpr = rt;
-	vcpu->arch.mmio_is_bigendian = is_bigendian;
+	vcpu->arch.mmio_host_swabbed = host_swabbed;
 	vcpu->mmio_needed = 1;
 	vcpu->mmio_is_write = 0;
 	vcpu->arch.mmio_sign_extend = 0;
@@ -841,14 +840,13 @@ int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
 {
 	void *data = run->mmio.data;
 	int idx, ret;
-	int is_bigendian;
+	bool host_swabbed;
 
+	/* Pity C doesn't have a logical XOR operator */
 	if (kvmppc_need_byteswap(vcpu)) {
-		/* Default endianness is "little endian". */
-		is_bigendian = !is_default_endian;
+		host_swabbed = is_default_endian;
 	} else {
-		/* Default endianness is "big endian". */
-		is_bigendian = is_default_endian;
+		host_swabbed = !is_default_endian;
 	}
 
 	if (bytes > sizeof(run->mmio.data)) {
@@ -863,7 +861,7 @@ int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
 	vcpu->mmio_is_write = 1;
 
 	/* Store the value at the lowest bytes in 'data'. */
-	if (is_bigendian) {
+	if (!host_swabbed) {
 		switch (bytes) {
 		case 8: *(u64 *)data = val; break;
 		case 4: *(u32 *)data = val; break;
@@ -871,11 +869,11 @@ int kvmppc_handle_store(struct kvm_run *run, struct kvm_vcpu *vcpu,
 		case 1: *(u8  *)data = val; break;
 		}
 	} else {
-		/* Store LE value into 'data'. */
 		switch (bytes) {
-		case 4: st_le32(data, val); break;
-		case 2: st_le16(data, val); break;
-		case 1: *(u8 *)data = val; break;
+		case 8: *(u64 *)data = swab64(val); break;
+		case 4: *(u32 *)data = swab32(val); break;
+		case 2: *(u16 *)data = swab16(val); break;
+		case 1: *(u8  *)data = val; break;
 		}
 	}
 
-- 
2.1.0

[PATCH 5/5] powerpc: Remove unused st_le*() and ld_le* functions

[David Gibson]

The powerpc specific st_le*() and ld_le*() functions in
arch/powerpc/asm/swab.h no longer have any users.  They are also
misleadingly named, since they always byteswap, even on a little-endian
host.

This patch removes them.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
 arch/powerpc/include/asm/swab.h | 26 --------------------------
 1 file changed, 26 deletions(-)

diff --git a/arch/powerpc/include/asm/swab.h b/arch/powerpc/include/asm/swab.h
index 96f59de..487e090 100644
--- a/arch/powerpc/include/asm/swab.h
+++ b/arch/powerpc/include/asm/swab.h
@@ -9,30 +9,4 @@
 
 #include <uapi/asm/swab.h>
 
-static __inline__ __u16 ld_le16(const volatile __u16 *addr)
-{
-	__u16 val;
-
-	__asm__ __volatile__ ("lhbrx %0,0,%1" : "=r" (val) : "r" (addr), "m" (*addr));
-	return val;
-}
-
-static __inline__ void st_le16(volatile __u16 *addr, const __u16 val)
-{
-	__asm__ __volatile__ ("sthbrx %1,0,%2" : "=m" (*addr) : "r" (val), "r" (addr));
-}
-
-static __inline__ __u32 ld_le32(const volatile __u32 *addr)
-{
-	__u32 val;
-
-	__asm__ __volatile__ ("lwbrx %0,0,%1" : "=r" (val) : "r" (addr), "m" (*addr));
-	return val;
-}
-
-static __inline__ void st_le32(volatile __u32 *addr, const __u32 val)
-{
-	__asm__ __volatile__ ("stwbrx %1,0,%2" : "=m" (*addr) : "r" (val), "r" (addr));
-}
-
 #endif /* _ASM_POWERPC_SWAB_H */
-- 
2.1.0

RE: [PATCH 0/5] powerpc: Get rid of redundant arch specific swab functions

[David Laight]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="utf-8", Size: 869 bytes --]

From: David Gibson
> arch/powerpc/include/asm/swab.h includes some powerpc specific
> byteswapping functions, which are implemented in terms of powerpc's
> built in byte reversed load/store instructions.  There are two problems with this:
> 
> 1) They're not necessary - gcc is perfectly capable of generating the
>    byte-reversed load and store instructions when using the normal,
>    generic byteswapping functions (tested with gcc (GCC) 4.8.3
>    20140911 (Red Hat 4.8.3-9))

Should you be worrying about older versions of gcc?
IIRC the internal byteswap 'stuff' is relatively recent (like
the last couple of years) so people building current kernels
on older distributions might have issues.

	David

ÿôèº{.nÇ+‰·Ÿ®‰­†+%ŠËÿ±éݶ\x17¥Šwÿº{.nÇ+‰·¥Š{±þG«éÿŠ{ayº\x1dʇڙë,j\a­¢f£¢·hšïêÿ‘êçz_è®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿÿ¾\a«þG«éÿ¢¸?™¨è­Ú&£ø§~á¶iO•æ¬z·švØ^\x14\x04\x1a¶^[m§ÿÿÃ\fÿ¶ìÿ¢¸?–I¥

Re: [PATCH 0/5] powerpc: Get rid of redundant arch specific swab functions

['David Gibson']

[-- Attachment #1: Type: text/plain, Size: 1260 bytes --]

On Wed, Feb 04, 2015 at 11:54:39AM +0000, David Laight wrote:
> From: David Gibson
> > arch/powerpc/include/asm/swab.h includes some powerpc specific
> > byteswapping functions, which are implemented in terms of powerpc's
> > built in byte reversed load/store instructions.  There are two problems with this:
> > 
> > 1) They're not necessary - gcc is perfectly capable of generating the
> >    byte-reversed load and store instructions when using the normal,
> >    generic byteswapping functions (tested with gcc (GCC) 4.8.3
> >    20140911 (Red Hat 4.8.3-9))
> 
> Should you be worrying about older versions of gcc?
> IIRC the internal byteswap 'stuff' is relatively recent (like
> the last couple of years) so people building current kernels
> on older distributions might have issues.

Well.. even then, surely the worst that will happen is that there will
be a few extra instructions to do the byteswap in registers.  Given
that these are mostly used for IO, I find it hard to imagine that
would make a measurable performance difference.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH 4/5] powerpc: Cleanup KVM emulated load/store endian handling

[Alexander Graf]

On 03.02.15 06:36, David Gibson wrote:
> Sometimes the KVM code on powerpc needs to emulate load or store
> instructions from the guest, which can include both normal and byte
> reversed forms.
> 
> We currently (AFAICT) handle this correctly, but some variable names are
> very misleading.  In particular we use "is_bigendian" in several places to
> actually mean "is the IO the same endian as the host", but we now support
> little-endian powerpc hosts.  This also ties into the misleadingly named
> ld_le*() and st_le*() functions, which in fact always byteswap, even on
> an LE host.
> 
> This patch cleans this up by renaming to more accurate "host_swabbed", and
> uses the generic swab*() functions instead of the powerpc specific and
> misleadingly named ld_le*() and st_le*() functions.
> 
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

Reviewed-by: Alexander Graf <agraf@suse.de>


Alex

Re: [PATCH 2/5] powerpc: Remove powerpc specific byteswap from bt8xx DVB driver

[Benjamin Herrenschmidt]

On Tue, 2015-02-03 at 16:36 +1100, David Gibson wrote:
> The bt8xx PCI DVB driver includes a powerpc specific hack, using one of
> the powerpc specific byteswapping functions in an IO helper macro.
> 
> There's no reason to use the powerpc specific function instead of a
> generic byteswap, so this patch removes it.  I'm not sure if the powerpc
> specific memory barrier is required, so I'm leaving that in.
> 
> Cc: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
> Cc: Peter Hettkamp <peter.hettkamp@htp-tel.de>

In fact there's no reason for the powerpc specific bits in this driver
(they might have been useful 15 years ago but they definitely aren't
anymore), I'll replace this with the patch with:

(Mauro, unless you holler ASAP, I'll just put this in my tree along
with David series due to various cross-dependencies, ie, he removes
those old crufty powerpc-specific leXX_to/from_cpu we had).


media/bt8xx: Remove old powerpc cruft

The powerpc arch code enables PCI_COMMAND_MEMORY (and has done so for
more than 10 years at least !) on pci_enable_device() and the hackery
on the MMIO accessor is useless as well, our writel does everything
this driver should need.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 drivers/media/pci/bt8xx/bt878.c | 12 ------------
 drivers/media/pci/bt8xx/bt878.h | 11 -----------
 2 files changed, 23 deletions(-)

diff --git a/drivers/media/pci/bt8xx/bt878.c b/drivers/media/pci/bt8xx/bt878.c
index 0939d39..8aa7266 100644
--- a/drivers/media/pci/bt8xx/bt878.c
+++ b/drivers/media/pci/bt8xx/bt878.c
@@ -416,9 +416,6 @@ static int bt878_probe(struct pci_dev *dev, const struct pci_device_id *pci_id)
 	int result = 0;
 	unsigned char lat;
 	struct bt878 *bt;
-#if defined(__powerpc__)
-	unsigned int cmd;
-#endif
 	unsigned int cardid;
 
 	printk(KERN_INFO "bt878: Bt878 AUDIO function found (%d).\n",
@@ -461,15 +458,6 @@ static int bt878_probe(struct pci_dev *dev, const struct pci_device_id *pci_id)
 	printk("irq: %d, latency: %d, memory: 0x%lx\n",
 	       bt->irq, lat, bt->bt878_adr);
 
-
-#if defined(__powerpc__)
-	/* on OpenFirmware machines (PowerMac at least), PCI memory cycle */
-	/* response on cards with no firmware is not enabled by OF */
-	pci_read_config_dword(dev, PCI_COMMAND, &cmd);
-	cmd = (cmd | PCI_COMMAND_MEMORY);
-	pci_write_config_dword(dev, PCI_COMMAND, cmd);
-#endif
-
 #ifdef __sparc__
 	bt->bt878_mem = (unsigned char *) bt->bt878_adr;
 #else
diff --git a/drivers/media/pci/bt8xx/bt878.h b/drivers/media/pci/bt8xx/bt878.h
index d19b592..49af240 100644
--- a/drivers/media/pci/bt8xx/bt878.h
+++ b/drivers/media/pci/bt8xx/bt878.h
@@ -142,18 +142,7 @@ void bt878_start(struct bt878 *bt, u32 controlreg, u32 op_sync_orin,
 		u32 irq_err_ignore);
 void bt878_stop(struct bt878 *bt);
 
-#if defined(__powerpc__)	/* big-endian */
-static inline void io_st_le32(volatile unsigned __iomem *addr, unsigned val)
-{
-	st_le32(addr, val);
-	eieio();
-}
-
-#define bmtwrite(dat,adr)  io_st_le32((adr),(dat))
-#define bmtread(adr)       ld_le32((adr))
-#else
 #define bmtwrite(dat,adr)  writel((dat), (adr))
 #define bmtread(adr)       readl(adr)
-#endif
 
 #endif