Should be CONFIG_GART_IOMMU instead of CONFIG_IOMMU here?
Hmm, looks like it would be nice to have more cleanups of iommu.h and
gart.h.
This is against tip/master.
=
From: FUJITA Tomonori <[email protected]>
Subject: [PATCH] replace CONFIG_IOMMU with CONFIG_GART_IOMMU in iommu.h
Signed-off-by: FUJITA Tomonori <[email protected]>
---
include/asm-x86/iommu.h | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/include/asm-x86/iommu.h b/include/asm-x86/iommu.h
index 07862fd..cb7e989 100644
--- a/include/asm-x86/iommu.h
+++ b/include/asm-x86/iommu.h
@@ -5,7 +5,7 @@ extern void pci_iommu_shutdown(void);
extern void no_iommu_init(void);
extern int force_iommu, no_iommu;
extern int iommu_detected;
-#ifdef CONFIG_IOMMU
+#ifdef CONFIG_GART_IOMMU
extern void gart_iommu_init(void);
extern void gart_iommu_shutdown(void);
extern void __init gart_parse_options(char *);
--
1.5.4.2
* FUJITA Tomonori <[email protected]> wrote:
> Should be CONFIG_GART_IOMMU instead of CONFIG_IOMMU here?
>
> Hmm, looks like it would be nice to have more cleanups of iommu.h and
> gart.h.
>
> This is against tip/master.
applied to tip/x86/core, thanks!
agreed about the cleanups. We are willing to test RFC patches :-)
Ingo
On Wed, 9 Jul 2008 10:52:30 +0200
Ingo Molnar <[email protected]> wrote:
>
> * FUJITA Tomonori <[email protected]> wrote:
>
> > Should be CONFIG_GART_IOMMU instead of CONFIG_IOMMU here?
> >
> > Hmm, looks like it would be nice to have more cleanups of iommu.h and
> > gart.h.
> >
> > This is against tip/master.
>
> applied to tip/x86/core, thanks!
>
> agreed about the cleanups. We are willing to test RFC patches :-)
Ok, how about this?
If you are fine with the intention of this, then I'll send the
follow-up cleanups.
Note that iommu.h includes some gart things that everyone needs. If
you like, I'll rename gart.h gart_priv.h (or something) and move the
gart stuff in iommu.h to gart.h.
=
From: FUJITA Tomonori <[email protected]>
Subject: [PATCH] x86: remove duplications in iommu.h and gart.h
This remove lots of duplications in iommu.h and gart.h.
The end result of this patch is:
- iommu.h is a header file for everyone related with IOMMUs.
- gart.h is the private header file. Only pci-gart_64.c and its
friends include it.
Signed-off-by: FUJITA Tomonori <[email protected]>
---
include/asm-x86/gart.h | 34 ++++------------------------------
include/asm-x86/iommu.h | 29 +++++++++++++++++------------
2 files changed, 21 insertions(+), 42 deletions(-)
diff --git a/include/asm-x86/gart.h b/include/asm-x86/gart.h
index 417f76e..33b9aee 100644
--- a/include/asm-x86/gart.h
+++ b/include/asm-x86/gart.h
@@ -1,40 +1,14 @@
-#ifndef _ASM_X8664_IOMMU_H
-#define _ASM_X8664_IOMMU_H 1
+#ifndef _ASM_X8664_GART_H
+#define _ASM_X8664_GART_H 1
#include <asm/e820.h>
+#include <asm/iommu.h>
-extern void pci_iommu_shutdown(void);
-extern void no_iommu_init(void);
-extern int force_iommu, no_iommu;
-extern int iommu_detected;
-extern int agp_amd64_init(void);
-#ifdef CONFIG_GART_IOMMU
-extern void gart_iommu_init(void);
-extern void gart_iommu_shutdown(void);
-extern void __init gart_parse_options(char *);
-extern void early_gart_iommu_check(void);
-extern void gart_iommu_hole_init(void);
extern void set_up_gart_resume(u32, u32);
+
extern int fallback_aper_order;
extern int fallback_aper_force;
-extern int gart_iommu_aperture;
-extern int gart_iommu_aperture_allowed;
-extern int gart_iommu_aperture_disabled;
extern int fix_aperture;
-#else
-#define gart_iommu_aperture 0
-#define gart_iommu_aperture_allowed 0
-#define gart_iommu_aperture_disabled 1
-
-static inline void early_gart_iommu_check(void)
-{
-}
-
-static inline void gart_iommu_shutdown(void)
-{
-}
-
-#endif
/* PTE bits. */
#define GPTE_VALID 1
diff --git a/include/asm-x86/iommu.h b/include/asm-x86/iommu.h
index cb7e989..068c9a4 100644
--- a/include/asm-x86/iommu.h
+++ b/include/asm-x86/iommu.h
@@ -1,29 +1,34 @@
-#ifndef _ASM_X8664_GART_H
-#define _ASM_X8664_GART_H 1
+#ifndef _ASM_X8664_IOMMU_H
+#define _ASM_X8664_IOMMU_H 1
extern void pci_iommu_shutdown(void);
extern void no_iommu_init(void);
extern int force_iommu, no_iommu;
extern int iommu_detected;
+
#ifdef CONFIG_GART_IOMMU
+extern int gart_iommu_aperture;
+extern int gart_iommu_aperture_allowed;
+extern int gart_iommu_aperture_disabled;
+
+extern void early_gart_iommu_check(void);
extern void gart_iommu_init(void);
extern void gart_iommu_shutdown(void);
extern void __init gart_parse_options(char *);
-extern void iommu_hole_init(void);
-extern int fallback_aper_order;
-extern int fallback_aper_force;
-extern int iommu_aperture;
-extern int iommu_aperture_allowed;
-extern int iommu_aperture_disabled;
-extern int fix_aperture;
+extern void gart_iommu_hole_init(void);
+
#else
-#define iommu_aperture 0
-#define iommu_aperture_allowed 0
+#define gart_iommu_aperture 0
+#define gart_iommu_aperture_allowed 0
+#define gart_iommu_aperture_disabled 1
-static inline void gart_iommu_shutdown(void)
+static inline void early_gart_iommu_check(void)
{
}
+static inline void gart_iommu_shutdown(void)
+{
+}
#endif
#endif
--
1.5.5.GIT
* FUJITA Tomonori <[email protected]> wrote:
> > agreed about the cleanups. We are willing to test RFC patches :-)
>
> Ok, how about this?
a perfect beginning - applied to tip/x86/core, thanks!
> If you are fine with the intention of this, then I'll send the
> follow-up cleanups.
please do!
> Note that iommu.h includes some gart things that everyone needs. If
> you like, I'll rename gart.h gart_priv.h (or something) and move the
> gart stuff in iommu.h to gart.h.
yeah. Or just move it into iommu.h and turn gart.h into the true
GART-specific include file? (and add iommu.h includes to files that are
affected)
Ingo
On Wed, Jul 09, 2008 at 04:29:41PM +0900, FUJITA Tomonori wrote:
> Should be CONFIG_GART_IOMMU instead of CONFIG_IOMMU here?
>
> Hmm, looks like it would be nice to have more cleanups of iommu.h and
> gart.h.
>
> This is against tip/master.
>
> =
> From: FUJITA Tomonori <[email protected]>
> Subject: [PATCH] replace CONFIG_IOMMU with CONFIG_GART_IOMMU in iommu.h
>
> Signed-off-by: FUJITA Tomonori <[email protected]>
> ---
> include/asm-x86/iommu.h | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/include/asm-x86/iommu.h b/include/asm-x86/iommu.h
> index 07862fd..cb7e989 100644
> --- a/include/asm-x86/iommu.h
> +++ b/include/asm-x86/iommu.h
> @@ -5,7 +5,7 @@ extern void pci_iommu_shutdown(void);
> extern void no_iommu_init(void);
> extern int force_iommu, no_iommu;
> extern int iommu_detected;
> -#ifdef CONFIG_IOMMU
> +#ifdef CONFIG_GART_IOMMU
> extern void gart_iommu_init(void);
> extern void gart_iommu_shutdown(void);
> extern void __init gart_parse_options(char *);
The asm-x86/iommu.h file is a leftover from some patches I submitted
last year. I renamed iommu.h to gart.h but somehow iommu.h still exists
in mainline. I think we should check again if somebody is using that
file and remove it if not.
Joerg
On Fri, 18 Jul 2008 13:31:05 +0200
Joerg Roedel <[email protected]> wrote:
> On Wed, Jul 09, 2008 at 04:29:41PM +0900, FUJITA Tomonori wrote:
> > Should be CONFIG_GART_IOMMU instead of CONFIG_IOMMU here?
> >
> > Hmm, looks like it would be nice to have more cleanups of iommu.h and
> > gart.h.
> >
> > This is against tip/master.
> >
> > =
> > From: FUJITA Tomonori <[email protected]>
> > Subject: [PATCH] replace CONFIG_IOMMU with CONFIG_GART_IOMMU in iommu.h
> >
> > Signed-off-by: FUJITA Tomonori <[email protected]>
> > ---
> > include/asm-x86/iommu.h | 2 +-
> > 1 files changed, 1 insertions(+), 1 deletions(-)
> >
> > diff --git a/include/asm-x86/iommu.h b/include/asm-x86/iommu.h
> > index 07862fd..cb7e989 100644
> > --- a/include/asm-x86/iommu.h
> > +++ b/include/asm-x86/iommu.h
> > @@ -5,7 +5,7 @@ extern void pci_iommu_shutdown(void);
> > extern void no_iommu_init(void);
> > extern int force_iommu, no_iommu;
> > extern int iommu_detected;
> > -#ifdef CONFIG_IOMMU
> > +#ifdef CONFIG_GART_IOMMU
> > extern void gart_iommu_init(void);
> > extern void gart_iommu_shutdown(void);
> > extern void __init gart_parse_options(char *);
>
> The asm-x86/iommu.h file is a leftover from some patches I submitted
> last year. I renamed iommu.h to gart.h but somehow iommu.h still exists
> in mainline. I think we should check again if somebody is using that
> file and remove it if not.
I think that we need a common header file (iommu.h or whatever) for
what everyone needs, pci_iommu_shutdown(), no_iommu_init(), etc.
Anyway, I did more cleanups:
http://lkml.org/lkml/2008/7/10/510
or
tip/x86/gart
Let me know if you think we still need cleanups.
Thanks,
On Sat, Jul 19, 2008 at 09:12:56PM +0900, FUJITA Tomonori wrote:
>
> I think that we need a common header file (iommu.h or whatever) for
> what everyone needs, pci_iommu_shutdown(), no_iommu_init(), etc.
>
> Anyway, I did more cleanups:
>
> http://lkml.org/lkml/2008/7/10/510
>
> or
>
> tip/x86/gart
>
> Let me know if you think we still need cleanups.
Now that you ask, I think the x86 DMA layer in general needs some rework
and cleanup. The implementation specific free_coherent function is never
called at the moment for example. For paravirtualized DMA we also need
some kind of stackable dma_ops according to Amit.
Joerg
On Sat, 19 Jul 2008 14:49:13 +0200
Joerg Roedel <[email protected]> wrote:
> On Sat, Jul 19, 2008 at 09:12:56PM +0900, FUJITA Tomonori wrote:
> >
> > I think that we need a common header file (iommu.h or whatever) for
> > what everyone needs, pci_iommu_shutdown(), no_iommu_init(), etc.
> >
> > Anyway, I did more cleanups:
> >
> > http://lkml.org/lkml/2008/7/10/510
> >
> > or
> >
> > tip/x86/gart
> >
> > Let me know if you think we still need cleanups.
>
> Now that you ask, I think the x86 DMA layer in general needs some rework
> and cleanup.
Agreed, I think that every IOMMU developer agrees :)
> The implementation specific free_coherent function is never
> called at the moment for example.
Right, x86 coherent functions need to be fixed. There are some other
areas to improve, I think.
> For paravirtualized DMA we also need
> some kind of stackable dma_ops according to Amit.
Yeah, we need that:
http://marc.info/?t=121117888500001&r=1&w=2
I think that we will need lots of IOMMU (and DMA stuff) work in the
near future.