On Wed, Feb 04, 2015 at 04:08:27PM +0000, Mark Salter wrote:

On Wed, 2015-02-04 at 11:25 +0000, Catalin Marinas wrote:

quoted

On Tue, Feb 03, 2015 at 05:29:36PM +0000, Mark Salter wrote:

quoted

On Mon, 2015-02-02 at 23:14 +0100, Rafael J. Wysocki wrote:

quoted

On Monday, February 02, 2015 08:45:30 PM Hanjun Guo wrote:

quoted

From: Mark Salter <redacted>

The acpi_os_ioremap() function may be used to map normal RAM or IO
regions. The current implementation simply uses ioremap_cache(). This
will work for some architectures, but arm64 ioremap_cache() cannot be
used to map IO regions which don't support caching. So for arm64, use
ioremap() for non-RAM regions.

CC: Rafael J Wysocki <redacted>
Signed-off-by: Mark Salter <redacted>
Signed-off-by: Hanjun Guo <redacted>
---
 include/acpi/acpi_io.h | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/include/acpi/acpi_io.h b/include/acpi/acpi_io.h
index 444671e..9d573db 100644
--- a/include/acpi/acpi_io.h
+++ b/include/acpi/acpi_io.h

@@ -1,11 +1,17 @@
 #ifndef _ACPI_IO_H_
 #define _ACPI_IO_H_
 
+#include <linux/mm.h>
 #include <linux/io.h>
 
 static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys,
 					    acpi_size size)
 {
+#ifdef CONFIG_ARM64
+	if (!page_is_ram(phys >> PAGE_SHIFT))
+		return ioremap(phys, size);
+#endif

I don't want to see #ifdef CONFIG_ARM64 in this file.

How about something like:

From: Mark Salter <redacted>
Date: Tue, 3 Feb 2015 10:51:16 -0500
Subject: [PATCH] acpi: fix acpi_os_ioremap for arm64

The acpi_os_ioremap() function may be used to map normal RAM or IO
regions. The current implementation simply uses ioremap_cache(). This
will work for some architectures, but arm64 ioremap_cache() cannot be
used to map IO regions which don't support caching. So for arm64, use
ioremap() for non-RAM regions.

CC: Rafael J Wysocki <redacted>
Signed-off-by: Mark Salter <redacted>
---
 arch/arm64/include/asm/acpi.h | 14 ++++++++++++++
 include/acpi/acpi_io.h        |  3 +++
 2 files changed, 17 insertions(+)

diff --git a/arch/arm64/include/asm/acpi.h b/arch/arm64/include/asm/acpi.h
index ea4d2b3..db82bc3 100644
--- a/arch/arm64/include/asm/acpi.h
+++ b/arch/arm64/include/asm/acpi.h

@@ -14,6 +14,7 @@
 
 #include <linux/irqchip/arm-gic-acpi.h>
 
+#include <linux/mm.h>
 #include <asm/smp_plat.h>
 
 /* Basic configuration for ACPI */

@@ -100,4 +101,17 @@ static inline bool acpi_psci_use_hvc(void) { return false; }
 static inline void acpi_init_cpus(void) { }
 #endif /* CONFIG_ACPI */
 
+/*
+ * ACPI table mapping
+ */
+static inline void __iomem *acpi_os_ioremap(acpi_physical_address phys,
+					    acpi_size size)
+{
+	if (!page_is_ram(phys >> PAGE_SHIFT))
+		return ioremap(phys, size);
+
+       return ioremap_cache(phys, size);
+}
+#define acpi_os_ioremap acpi_os_ioremap

That's one way of doing this, I'm not too bothered with the approach
(define the function name, an ARCH_HAS macro or a Kconfig option, it's
up to Rafael).

But a question I already asked is what we need ioremap_cache() for? We
don't use NVS on arm64 yet, so is there anything else requiring
cacheable mapping?

acpi_os_remap() is used to map ACPI tables. These tables may be in ram
which are already included in the kernel's linear RAM mapping. So we
need ioremap_cache to avoid two mappings to the same physical page
having different caching attributes.

What's the call path to acpi_os_ioremap() on such tables already in the
linear mapping? I can see an acpi_map() function which already takes
care of the RAM mapping case but there are other cases where
acpi_os_ioremap() is called directly. For example,
acpi_os_read_memory(), can it be called on both RAM and I/O?

-- 
Catalin