IDE woos in BE mode 2.6 kernel

All of lore.kernel.org
 help / color / mirror / Atom feed

* IDE woos in BE mode 2.6 kernel
@ 2004-09-17 19:18 Manish Lachwani
  2004-09-18 14:19 ` Atsushi Nemoto
  0 siblings, 1 reply; 6+ messages in thread
From: Manish Lachwani @ 2004-09-17 19:18 UTC (permalink / raw)
  To: linux-mips

[-- Attachment #1: Type: text/plain, Size: 953 bytes --]

Hello !

In response to Jun Suns mail sent on 06/24/2004

Anybody has tried IDE disks in big endian mode with 2.6 kernel?
I seem to have troubles with Malta board.

Current malta board has CONFIG_SWAP_IO_SPACE defined and therefore
all inw, inl and their friends are byte-swapped in BE mode.  As a
results all IDE IO ops (such as ide_inw, etc) do swapping too.

A quick experiement shows those IDE IO ops should not do swapping.
Anybody knows why?

Apparently fixing the above is not enough.  I either encountered
failure to read partition table or having DMA error.  Any clues
here?

I suppose this problem really should exist for other arches
with BE support.  Anybody knows how other arches deal with this?

Thanks.

Jun

---

The following patch gets the Malta to work well. However, this patch introduces board specific changes in the IDE subsystem.
This is not a final patch but maybe there can be a better approach to this issue

Thanks
Manish



[-- Attachment #2: patch-26-ide-malta --]
[-- Type: text/plain, Size: 1937 bytes --]

--- drivers/ide/ide-iops.c.orig	2004-09-16 19:20:52.000000000 -0700
+++ drivers/ide/ide-iops.c	2004-09-16 18:55:37.000000000 -0700
@@ -95,13 +95,23 @@
 	hwif->OUTBSYNC	= ide_outbsync;
 	hwif->OUTW	= ide_outw;
 	hwif->OUTL	= ide_outl;
+#if defined(CONFIG_MIPS_MALTA) && !defined(CONFIG_CPU_LITTLE_ENDIAN)
+	hwif->OUTSW	= malta_ide_outsw;
+	hwif->OUTSL	= malta_ide_outsl;
+#else
 	hwif->OUTSW	= ide_outsw;
 	hwif->OUTSL	= ide_outsl;
+#endif
 	hwif->INB	= ide_inb;
 	hwif->INW	= ide_inw;
 	hwif->INL	= ide_inl;
+#if defined(CONFIG_MIPS_MALTA) && !defined(CONFIG_CPU_LITTLE_ENDIAN)
+	hwif->INSW	= malta_ide_insw;
+	hwif->INSL	= malta_ide_insl;
+#else
 	hwif->INSW	= ide_insw;
 	hwif->INSL	= ide_insl;
+#endif
 }
 
 EXPORT_SYMBOL(default_hwif_iops);
--- include/asm-mips/ide.h.orig	2004-09-16 15:41:00.000000000 -0700
+++ include/asm-mips/ide.h	2004-09-16 18:12:26.000000000 -0700
@@ -20,6 +20,42 @@
 #define __ide_mm_outsw  ide_outsw
 #define __ide_mm_outsl  ide_outsl
 
+#if defined(CONFIG_MIPS_MALTA) && !defined(CONFIG_CPU_LITTLE_ENDIAN)
+extern const unsigned long mips_io_port_base;
+
+static inline void malta_ide_insw(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(u16 *)addr = *(volatile u16 *)(mips_io_port_base + port);
+		addr += 2;
+	}
+}
+
+static inline void malta_ide_outsw(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(volatile u16 *)(mips_io_port_base + (port)) = *(u16 *)addr;
+		addr += 2;
+	}
+}
+
+static inline void malta_ide_insl(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(u32 *)addr = *(volatile u32 *)(mips_io_port_base + port);
+		addr += 4;
+	}
+}
+
+static inline void malta_ide_outsl(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(volatile u32 *)(mips_io_port_base + (port)) = *(u32 *)addr;
+		addr += 4;
+	}
+}
+#endif
+
 #endif /* __KERNEL__ */
 
 #endif /* __ASM_IDE_H */

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: IDE woos in BE mode 2.6 kernel
  2004-09-17 19:18 IDE woos in BE mode 2.6 kernel Manish Lachwani
@ 2004-09-18 14:19 ` Atsushi Nemoto
  2004-09-24 23:40   ` Manish Lachwani
  0 siblings, 1 reply; 6+ messages in thread
From: Atsushi Nemoto @ 2004-09-18 14:19 UTC (permalink / raw)
  To: mlachwani; +Cc: linux-mips

>>>>> On Fri, 17 Sep 2004 12:18:37 -0700, Manish Lachwani <mlachwani@mvista.com> said:

mlachwani> In response to Jun Suns mail sent on 06/24/2004

jsun> Anybody has tried IDE disks in big endian mode with 2.6
jsun> kernel?  I seem to have troubles with Malta board.
...
mlachwani> The following patch gets the Malta to work well. However,
mlachwani> this patch introduces board specific changes in the IDE
mlachwani> subsystem.  This is not a final patch but maybe there can
mlachwani> be a better approach to this issue

FYI, here is my approach.  Not so beautiful but less intrusive...

1. copy include/asm-mips/mach-generic/ide.h to my mach-xxx directory.
2. add following lines to include/asm-mips/mach-xxx/ide.h
--- begin ---
/* get rid of defs from io.h - ide has its private and conflicting versions */
#ifdef insb
#undef insb
#endif
#ifdef outsb
#undef outsb
#endif
#ifdef insw
#undef insw
#endif
#ifdef outsw
#undef outsw
#endif
#ifdef insl
#undef insl
#endif
#ifdef outsl
#undef outsl
#endif

#define insb(port, addr, count) ___ide_insb(port, addr, count)
#define insw(port, addr, count) ___ide_insw(port, addr, count)
#define insl(port, addr, count) ___ide_insl(port, addr, count)
#define outsb(port, addr, count) ___ide_outsb(port, addr, count)
#define outsw(port, addr, count) ___ide_outsw(port, addr, count)
#define outsl(port, addr, count) ___ide_outsl(port, addr, count)

static inline void ___ide_insb(unsigned long port, void *addr, unsigned int count)
{
	unsigned long start = (unsigned long)addr;
	unsigned long size = (unsigned long)count;
	while (count--) {
		*(u16 *)addr = *(volatile u8 *)(mips_io_port_base + port);
		addr++;
	}
	if (cpu_has_dc_aliases)
		dma_cache_wback((unsigned long)start, size);
}

static inline void ___ide_outsb(unsigned long port, void *addr, unsigned int count)
{
	while (count--) {
		*(volatile u8 *)(mips_io_port_base + port) = *(u8 *)addr;
		addr++;
	}
}

static inline void ___ide_insw(unsigned long port, void *addr, unsigned int count)
{
	unsigned long start = (unsigned long)addr;
	unsigned long size = (unsigned long)count * 2;
	while (count--) {
		*(u16 *)addr = *(volatile u16 *)(mips_io_port_base + port);
		addr += 2;
	}
	if (cpu_has_dc_aliases)
		dma_cache_wback((unsigned long)start, size);
}

static inline void ___ide_outsw(unsigned long port, void *addr, unsigned int count)
{
	while (count--) {
		*(volatile u16 *)(mips_io_port_base + port) = *(u16 *)addr;
		addr += 2;
	}
}

static inline void ___ide_insl(unsigned long port, void *addr, unsigned int count)
{
	unsigned long start = (unsigned long)addr;
	unsigned long size = (unsigned long)count * 4;
	while (count--) {
		*(u32 *)addr = *(volatile u32 *)(mips_io_port_base + port);
		addr += 4;
	}
	if (cpu_has_dc_aliases)
		dma_cache_wback((unsigned long)start, size);
}

static inline void ___ide_outsl(unsigned long port, void *addr, unsigned int count)
{
	while (count--) {
		*(volatile u32 *)(mips_io_port_base + port) = *(u32 *)addr;
		addr += 4;
	}
}
--- end ---

Note that above codes include workarounds for PIO IDE cache problem
(dma_cache_wback) though I'm not sure this workaround still needed.
Please refer this ML thread for the workaround.
<http://www.linux-mips.org/archives/linux-mips/2004-03/msg00185.html>


And as I wrote before, I think this IDE endian problem still exists in
current 2.4 tree too.  Please refer my 02/26 mail for this topic.
<http://www.linux-mips.org/archives/linux-mips/2004-02/msg00219.html>

Thank you.

---
Atsushi Nemoto

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: IDE woos in BE mode 2.6 kernel
  2004-09-18 14:19 ` Atsushi Nemoto
@ 2004-09-24 23:40   ` Manish Lachwani
  2004-09-25 14:41     ` Atsushi Nemoto
  0 siblings, 1 reply; 6+ messages in thread
From: Manish Lachwani @ 2004-09-24 23:40 UTC (permalink / raw)
  To: Atsushi Nemoto; +Cc: linux-mips

[-- Attachment #1: Type: text/plain, Size: 3960 bytes --]

Hello !

These changes below work and attached is the patch based on your 
suggestion. However, I have made changes to include/asm-mips/ide.h since 
I think there may be problems with other MIPS boards to in BE mode. 
And,  I have not included the  (dma_cache_wback) in the patch.

Thanks
Manish


Atsushi Nemoto wrote:

>>>>>>On Fri, 17 Sep 2004 12:18:37 -0700, Manish Lachwani <mlachwani@mvista.com> said:
>>>>>>            
>>>>>>
>
>mlachwani> In response to Jun Suns mail sent on 06/24/2004
>
>jsun> Anybody has tried IDE disks in big endian mode with 2.6
>jsun> kernel?  I seem to have troubles with Malta board.
>...
>mlachwani> The following patch gets the Malta to work well. However,
>mlachwani> this patch introduces board specific changes in the IDE
>mlachwani> subsystem.  This is not a final patch but maybe there can
>mlachwani> be a better approach to this issue
>
>FYI, here is my approach.  Not so beautiful but less intrusive...
>
>1. copy include/asm-mips/mach-generic/ide.h to my mach-xxx directory.
>2. add following lines to include/asm-mips/mach-xxx/ide.h
>--- begin ---
>/* get rid of defs from io.h - ide has its private and conflicting versions */
>#ifdef insb
>#undef insb
>#endif
>#ifdef outsb
>#undef outsb
>#endif
>#ifdef insw
>#undef insw
>#endif
>#ifdef outsw
>#undef outsw
>#endif
>#ifdef insl
>#undef insl
>#endif
>#ifdef outsl
>#undef outsl
>#endif
>
>#define insb(port, addr, count) ___ide_insb(port, addr, count)
>#define insw(port, addr, count) ___ide_insw(port, addr, count)
>#define insl(port, addr, count) ___ide_insl(port, addr, count)
>#define outsb(port, addr, count) ___ide_outsb(port, addr, count)
>#define outsw(port, addr, count) ___ide_outsw(port, addr, count)
>#define outsl(port, addr, count) ___ide_outsl(port, addr, count)
>
>static inline void ___ide_insb(unsigned long port, void *addr, unsigned int count)
>{
>	unsigned long start = (unsigned long)addr;
>	unsigned long size = (unsigned long)count;
>	while (count--) {
>		*(u16 *)addr = *(volatile u8 *)(mips_io_port_base + port);
>		addr++;
>	}
>	if (cpu_has_dc_aliases)
>		dma_cache_wback((unsigned long)start, size);
>}
>
>static inline void ___ide_outsb(unsigned long port, void *addr, unsigned int count)
>{
>	while (count--) {
>		*(volatile u8 *)(mips_io_port_base + port) = *(u8 *)addr;
>		addr++;
>	}
>}
>
>static inline void ___ide_insw(unsigned long port, void *addr, unsigned int count)
>{
>	unsigned long start = (unsigned long)addr;
>	unsigned long size = (unsigned long)count * 2;
>	while (count--) {
>		*(u16 *)addr = *(volatile u16 *)(mips_io_port_base + port);
>		addr += 2;
>	}
>	if (cpu_has_dc_aliases)
>		dma_cache_wback((unsigned long)start, size);
>}
>
>static inline void ___ide_outsw(unsigned long port, void *addr, unsigned int count)
>{
>	while (count--) {
>		*(volatile u16 *)(mips_io_port_base + port) = *(u16 *)addr;
>		addr += 2;
>	}
>}
>
>static inline void ___ide_insl(unsigned long port, void *addr, unsigned int count)
>{
>	unsigned long start = (unsigned long)addr;
>	unsigned long size = (unsigned long)count * 4;
>	while (count--) {
>		*(u32 *)addr = *(volatile u32 *)(mips_io_port_base + port);
>		addr += 4;
>	}
>	if (cpu_has_dc_aliases)
>		dma_cache_wback((unsigned long)start, size);
>}
>
>static inline void ___ide_outsl(unsigned long port, void *addr, unsigned int count)
>{
>	while (count--) {
>		*(volatile u32 *)(mips_io_port_base + port) = *(u32 *)addr;
>		addr += 4;
>	}
>}
>--- end ---
>
>Note that above codes include workarounds for PIO IDE cache problem
>(dma_cache_wback) though I'm not sure this workaround still needed.
>Please refer this ML thread for the workaround.
><http://www.linux-mips.org/archives/linux-mips/2004-03/msg00185.html>
>
>
>And as I wrote before, I think this IDE endian problem still exists in
>current 2.4 tree too.  Please refer my 02/26 mail for this topic.
><http://www.linux-mips.org/archives/linux-mips/2004-02/msg00219.html>
>
>Thank you.
>
>---
>Atsushi Nemoto
>  
>


[-- Attachment #2: patch-26-ide-malta --]
[-- Type: text/plain, Size: 2398 bytes --]

--- include/asm-mips/ide.h.orig	2004-09-16 15:41:00.000000000 -0700
+++ include/asm-mips/ide.h	2004-09-20 09:45:14.000000000 -0700
@@ -20,6 +20,91 @@
 #define __ide_mm_outsw  ide_outsw
 #define __ide_mm_outsl  ide_outsl
 
+#ifndef CONFIG_CPU_LITTLE_ENDIAN
+/*
+ * Only for the Big Endian systems, do not do the swapping.
+ * We cannot turn off the CONFIG_SWAP_IO_SPACE since the
+ * other subsystems need it. Hence we need this approach for
+ * IDE only - Manish Lachwani (mlachwani@mvista.com)
+ */
+extern const unsigned long mips_io_port_base;
+
+#ifdef insb
+#undef insb
+#endif
+#ifdef outsb
+#undef outsb
+#endif
+#ifdef insw
+#undef insw
+#endif
+#ifdef outsw
+#undef outsw
+#endif
+#ifdef insl
+#undef insl
+#endif
+#ifdef outsl
+#undef outsl
+#endif
+
+#define insb(port, addr, count)		___ide_insb(port, addr, count)
+#define insw(port, addr, count)		___ide_insw(port, addr, count)
+#define insl(port, addr, count)		___ide_insl(port, addr, count)
+#define outsb(port, addr, count)	___ide_outsb(port, addr, count)
+#define outsw(port, addr, count)	___ide_outsw(port, addr, count)
+#define outsl(port, addr, count)	___ide_outsl(port, addr, count)
+
+static inline void ___ide_insb(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(u16 *)addr = *(volatile u8 *)(mips_io_port_base + port);
+		addr++;
+	}
+}
+
+static inline void ___ide_outsb(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(volatile u8 *)(mips_io_port_base + port) = *(u8 *)addr;
+		addr++;
+	}
+}
+
+static inline void ___ide_insw(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(u16 *)addr = *(volatile u16 *)(mips_io_port_base + port);
+		addr += 2;
+	}
+}
+
+static inline void ___ide_outsw(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(volatile u16 *)(mips_io_port_base + port) = *(u16 *)addr;
+		addr += 2;
+	}
+}
+
+static inline void ___ide_insl(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(u32 *)addr = *(volatile u32 *)(mips_io_port_base + port);
+		addr += 4;
+	}
+}
+
+static inline void ___ide_outsl(unsigned long port, void *addr, unsigned int count)
+{
+	while (count--) {
+		*(volatile u32 *)(mips_io_port_base + port) = *(u32 *)addr;
+		addr += 4;
+	}
+}
+
+#endif /* CONFIG_LITTLE_ENDIAN */
+
 #endif /* __KERNEL__ */
 
 #endif /* __ASM_IDE_H */

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: IDE woos in BE mode 2.6 kernel
  2004-09-24 23:40   ` Manish Lachwani
@ 2004-09-25 14:41     ` Atsushi Nemoto
  0 siblings, 0 replies; 6+ messages in thread
From: Atsushi Nemoto @ 2004-09-25 14:41 UTC (permalink / raw)
  To: mlachwani; +Cc: linux-mips

>>>>> On Fri, 24 Sep 2004 16:40:25 -0700, Manish Lachwani <mlachwani@mvista.com> said:

mlachwani> These changes below work and attached is the patch based on
mlachwani> your suggestion. However, I have made changes to
mlachwani> include/asm-mips/ide.h since I think there may be problems
mlachwani> with other MIPS boards to in BE mode.  And, I have not
mlachwani> included the (dma_cache_wback) in the patch.

Hmm, I think declaration of mips_io_port_base is not needed since
asm/io.h must be included before asm/ide.h.  If asm/io.h was not
included first, ide.h can not override these functions anyway.

Also, moving the fixup block before "#include <ide.h>" will give a
chance to override these functions again in mach-dependent ide.h. (for
more weired hardware)

---
Atsushi Nemoto

^ permalink raw reply	[flat|nested] 6+ messages in thread

* IDE woos in BE mode 2.6 kernel
@ 2004-06-24 23:45 Jun Sun
  2004-06-25  8:45 ` Geert Uytterhoeven
  0 siblings, 1 reply; 6+ messages in thread
From: Jun Sun @ 2004-06-24 23:45 UTC (permalink / raw)
  To: linux-mips; +Cc: jsun

Anybody has tried IDE disks in big endian mode with 2.6 kernel?
I seem to have troubles with Malta board.

Current malta board has CONFIG_SWAP_IO_SPACE defined and therefore
all inw, inl and their friends are byte-swapped in BE mode.  As a
results all IDE IO ops (such as ide_inw, etc) do swapping too.

A quick experiement shows those IDE IO ops should not do swapping.
Anybody knows why?

Apparently fixing the above is not enough.  I either encountered
failure to read partition table or having DMA error.  Any clues
here?

I suppose this problem really should exist for other arches
with BE support.  Anybody knows how other arches deal with this?

Thanks.

Jun

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: IDE woos in BE mode 2.6 kernel
  2004-06-24 23:45 Jun Sun
@ 2004-06-25  8:45 ` Geert Uytterhoeven
  0 siblings, 0 replies; 6+ messages in thread
From: Geert Uytterhoeven @ 2004-06-25  8:45 UTC (permalink / raw)
  To: Jun Sun; +Cc: Linux/MIPS Development

On Thu, 24 Jun 2004, Jun Sun wrote:
> I suppose this problem really should exist for other arches
> with BE support.  Anybody knows how other arches deal with this?

My Amiga is happily running 2.6.7 and doesn't have problems with IDE, so m68k
is not affected.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2004-09-25 14:31 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-09-17 19:18 IDE woos in BE mode 2.6 kernel Manish Lachwani
2004-09-18 14:19 ` Atsushi Nemoto
2004-09-24 23:40   ` Manish Lachwani
2004-09-25 14:41     ` Atsushi Nemoto
  -- strict thread matches above, loose matches on Subject: below --
2004-06-24 23:45 Jun Sun
2004-06-25  8:45 ` Geert Uytterhoeven

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.