Skip to contentSkip to conversation

trm290: do hook dma_host_{on,off} methods (take 2)

PATCH Applied Sergei Shtylyov 2 participants 3 messages Dec 27, 2007 – Dec 28, 2007 libata

lore
Share & tools
lore

Sergei Shtylyov

author
#0 lore raw

PATCH 1/1 trm290: do hook dma_host_{on,off} methods (take 2)

Using default methods caused the chip's DMA PRD count registers, inadvertently starting DMA! While fixing it, also do:

- get rid of the 'ide_' prefixes in several functions for which the prefix in the method's name has been 'ide_' ectomized already;

- align the code hooking the IDE DMA methods in init_hwif_trm290()...

---
Fixed the header, added 'static' to the new method definitions. The patch is against Linus' kernel tree...

Signed-off-by Sergei Shtylyov <[email protected]>
 drivers/ide/pci/trm290.c |   29 ++++++++++++++++++++---------
 1 files changed, 20 insertions(+), 9 deletions(-)
drivers/ide/pci/trm290.c +20 -9
Index: linux-2.6/drivers/ide/pci/trm290.c
===================================================================
@@ -1,7 +1,8 @@
 /*
- *  linux/drivers/ide/pci/trm290.c		Version 1.02	Mar. 18, 2000
+ *  linux/drivers/ide/pci/trm290.c		Version 1.05	Dec. 26, 2007
  *
  *  Copyright (c) 1997-1998  Mark Lord
+ *  Copyright (c) 2007       MontaVista Software, Inc. <[email protected]>
  *  May be copied or modified under the terms of the GNU General Public License
  *
  *  June 22, 2004 - get rid of check_region

@@ -177,7 +178,7 @@ static void trm290_selectproc (ide_drive
 	trm290_prepare_drive(drive, drive->using_dma);
 }
 
-static void trm290_ide_dma_exec_cmd(ide_drive_t *drive, u8 command)
+static void trm290_dma_exec_cmd(ide_drive_t *drive, u8 command)
 {
 	BUG_ON(HWGROUP(drive)->handler != NULL);	/* paranoia check */
 	ide_set_handler(drive, &ide_dma_intr, WAIT_CMD, NULL);

@@ -185,7 +186,7 @@ static void trm290_ide_dma_exec_cmd(ide_
 	outb(command, IDE_COMMAND_REG);
 }
 
-static int trm290_ide_dma_setup(ide_drive_t *drive)
+static int trm290_dma_setup(ide_drive_t *drive)
 {
 	ide_hwif_t *hwif = drive->hwif;
 	struct request *rq = hwif->hwgroup->rq;

@@ -215,7 +216,7 @@ static int trm290_ide_dma_setup(ide_driv
 	return 0;
 }
 
-static void trm290_ide_dma_start(ide_drive_t *drive)
+static void trm290_dma_start(ide_drive_t *drive)
 {
 }
 

@@ -240,6 +241,14 @@ static int trm290_ide_dma_test_irq (ide_
 	return (status == 0x00ff);
 }
 
+static void trm290_dma_host_on(ide_drive_t *drive)
+{
+}
+
+static void trm290_dma_host_off(ide_drive_t *drive)
+{
+}
+
 static void __devinit init_hwif_trm290(ide_hwif_t *hwif)
 {
 	unsigned int cfgbase = 0;

@@ -280,11 +289,13 @@ static void __devinit init_hwif_trm290(i
 
 	ide_setup_dma(hwif, (hwif->config_data + 4) ^ (hwif->channel ? 0x0080 : 0x0000), 3);
 
-	hwif->dma_setup = &trm290_ide_dma_setup;
-	hwif->dma_exec_cmd = &trm290_ide_dma_exec_cmd;
-	hwif->dma_start = &trm290_ide_dma_start;
-	hwif->ide_dma_end = &trm290_ide_dma_end;
-	hwif->ide_dma_test_irq = &trm290_ide_dma_test_irq;
+	hwif->dma_host_off	= &trm290_dma_host_off;
+	hwif->dma_host_on	= &trm290_dma_host_on;
+	hwif->dma_setup 	= &trm290_dma_setup;
+	hwif->dma_exec_cmd	= &trm290_dma_exec_cmd;
+	hwif->dma_start 	= &trm290_dma_start;
+	hwif->ide_dma_end	= &trm290_ide_dma_end;
+	hwif->ide_dma_test_irq	= &trm290_ide_dma_test_irq;
 
 	hwif->selectproc = &trm290_selectproc;
 #if 1
Applied
#1 lore raw
Quoted 2 lines · Sergei Shtylyov

Using default methods caused the chip's DMA PRD count registers, inadvertently starting DMA! While fixing it, also do:

nasty, this could possibly explain the following trm290.c hack:

...
#if 0 /* play it safe for now */
IDE_HFLAG_TRUST_BIOS_FOR_DMA |
#endif
...

Quoted 10 lines

- get rid of the 'ide_' prefixes in several functions for which the prefix in the method's name has been 'ide_' ectomized already;

- align the code hooking the IDE DMA methods in init_hwif_trm290()...

---
Fixed the header, added 'static' to the new method definitions. The patch is against Linus' kernel tree...

Signed-off-by Sergei Shtylyov <[email protected]>

applied

author
#2 lore raw

Hello.

Quoted 4 lines · Bartlomiej Zolnierkiewicz
Quoted 2 lines

Using default methods caused the chip's DMA PRD count registers, inadvertently starting DMA! While fixing it, also do:

nasty, this could possibly explain the following trm290.c hack:

Frankly speaking, I'm not sure it's that destructive: the register in
question is accessed by inw()/outw() while the genric code would only "spoil" it by inb(). So, I don't know if this would be access to the same register as
outw() does (that write starts DMA, and it is a number of 32-bit words in PRD table - 1). Actually, the PRD address is loaded by 32-bit write at address 2 bytes below the one in question, and so overlaps our register -- the chip must be looking at byte enables when deciding what register is addressed).

Quoted 2 lines

...
#if 0 /* play it safe for now */

IDE_HFLAG_TRUST_BIOS_FOR_DMA |

Quoted 8 lines

#endif
...

Quoted 4 lines

- get rid of the 'ide_' prefixes in several functions for which the prefix in the method's name has been 'ide_' ectomized already;

- align the code hooking the IDE DMA methods in init_hwif_trm290()...

Quoted 7 lines
Quoted 5 lines

---
Fixed the header, added 'static' to the new method definitions. The patch is against Linus' kernel tree...

Signed-off-by Sergei Shtylyov <[email protected]>

applied

The next patch is cooking... :-)

WBR, Sergei