Skip to contentSkip to conversation

archs: add ioremap_change_write_prot

PATCH In review Marco Stornelli 3 participants 6 messages Oct 24, 2010 – Oct 27, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 archs: add ioremap_change_write_prot

From: Marco Stornelli <[email protected]>

This patch adds a new function called ioremap_change_write_prot. Goal of this patch is to create a general function for all archs to be able to change a write protection of a memory mapped area. x86 has already got a similar framework, so this function is only a wrapper. Next target will be arm and powerpc. This function will be used by Pramfs (see its latest review).

---

arch/x86/mm/ioremap.c +21 -0
diff -Nurp linux-2.6.36-orig/arch/x86/mm/ioremap.c linux-2.6.36/arch/x86/mm/ioremap.c
@@ -50,6 +50,27 @@ int ioremap_change_attr(unsigned long va
 	return err;
 }
 
+int ioremap_change_write_prot(unsigned long vaddr, unsigned long size,
+			       unsigned int rw)
+{
+	unsigned long nrpages = size >> PAGE_SHIFT;
+	int ret;
+
+	/* Page aligned */
+	vaddr &= PAGE_MASK;
+
+	if (size & (PAGE_SIZE - 1))
+		nrpages++;
+
+	if (rw)
+		ret = set_memory_rw(vaddr, nrpages);
+	else
+		ret = set_memory_ro(vaddr, nrpages);
+
+	return ret;
+}
+EXPORT_SYMBOL(ioremap_change_write_prot);
+
 /*
  * Remap an arbitrary physical address space into the kernel virtual
  * address space. Needed when the kernel wants to access high addresses
arch/Kconfig +3 -0
@@ -83,6 +83,9 @@ config USER_RETURN_NOTIFIER
 config HAVE_IOREMAP_PROT
 	bool
 
+config HAVE_IOREMAP_CHANGE_WRITE_PROT
+	bool
+
 config HAVE_KPROBES
 	bool
 
arch/x86/Kconfig +1 -0
@@ -26,6 +26,7 @@ config X86
 	select HAVE_OPROFILE
 	select HAVE_PERF_EVENTS if (!M386 && !M486)
 	select HAVE_IOREMAP_PROT
+	select HAVE_IOREMAP_CHANGE_WRITE_PROT
 	select HAVE_KPROBES
 	select ARCH_WANT_OPTIONAL_GPIOLIB
 	select ARCH_WANT_FRAME_POINTERS
arch/x86/include/asm/io.h +2 -0
diff -Nurp linux-2.6.36-orig/arch/x86/include/asm/io.h linux-2.6.36/arch/x86/include/asm/io.h
@@ -333,6 +333,8 @@ extern void unxlate_dev_mem_ptr(unsigned
 
 extern int ioremap_change_attr(unsigned long vaddr, unsigned long size,
 				unsigned long prot_val);
+extern int ioremap_change_write_prot(unsigned long vaddr, unsigned long size,
+				unsigned int rw);
 extern void __iomem *ioremap_wc(resource_size_t offset, unsigned long size);
 
 /*
#1 lore raw

2010/10/24 Marco Stornelli <[email protected]>:

Quoted 10 lines

From: Marco Stornelli <[email protected]>

This patch adds a new function called ioremap_change_write_prot. Goal of this patch is to create a general function for all archs to be able to change a write protection of a memory mapped area. x86 has already got a similar framework, so this function is only a wrapper. Next target will be arm and powerpc. This function will be used by Pramfs (see its latest review).

Sorry to re-open the topic, but I'd like to receive any comments/feedback about this patch.

Marco

#2 lore raw
Quoted 2 lines

Sorry to re-open the topic, but I'd like to receive any comments/feedback about this patch.

The name of the wrapper seems wrong -- this has nothing to do with IO.

I must admit I'm also not a fan of wrappers, why not call the set_memory_* functions directly and implement them for the other archs too?

Other than that it seems reasonable.

-Andi

Signature · 1 line
[email protected] -- Speaking for myself only.
#3 lore raw
In reply to Marco Stornelli
Quoted 8 lines · Marco Stornelli

From: Marco Stornelli <[email protected]>

This patch adds a new function called ioremap_change_write_prot. Goal of this patch is to create a general function for all archs to be able to change a write protection of a memory mapped area. x86 has already got a similar framework, so this function is only a wrapper. Next target will be arm and powerpc. This function will be used by Pramfs (see its latest review).

Sorry for the delay. I'm traveling.

Quoted 20 lines

---

diff -Nurp linux-2.6.36-orig/arch/x86/mm/ioremap.c linux-2.6.36/arch/x86/mm/ioremap.c

arch/x86/mm/ioremap.c +10 -0
@@ -50,6 +50,27 @@ int ioremap_change_attr(unsigned long va
 	return err;
 }
 
+int ioremap_change_write_prot(unsigned long vaddr, unsigned long size,
+			       unsigned int rw)
+{
+	unsigned long nrpages = size >> PAGE_SHIFT;
+	int ret;
+
+	/* Page aligned */
+	vaddr &= PAGE_MASK;
+
+	if (size & (PAGE_SIZE - 1))

arch/x86/mm/ioremap.c int ioremap_change_attr(unsigned long va · line 72

Quoted 10 lines
arch/x86/mm/ioremap.c +10 -0
+		nrpages++;
+
+	if (rw)
+		ret = set_memory_rw(vaddr, nrpages);
+	else
+		ret = set_memory_ro(vaddr, nrpages);
+
+	return ret;
+}
+EXPORT_SYMBOL(ioremap_change_write_prot);

Why inventing a new interface if we have one already ?

set_memory_* which has a few more variants than ro/rw

Also this is not related to ioremap, these are functions which works on almost anything when there are other restrictions in place.

So implementing set_memory_ro(), set_memory_rw() for those archs which you need it for makes more sense than creating a misnomed and misplaced wrapper.

Thanks,

tglx

#4 lore raw

2010/10/26 Thomas Gleixner <[email protected]>:

Quoted 8 lines

Why inventing a new interface if we have one already ?

   set_memory_* which has a few more variants than ro/rw

Also this is not related to ioremap, these are functions which works on almost anything when there are other restrictions in place.

Quoted 8 lines

So implementing set_memory_ro(), set_memory_rw() for those archs which you need it for makes more sense than creating a misnomed and misplaced wrapper.

Thanks,

       tglx

Andi and Thomas, thanks for your response. I was trying to use an homogeneous interface, I see that only x86 has got set_memory_*, but ok I can use these ones. About the name I used, I was trying to use a way to change the protection after an ioremap, so for x86 means a call to these functions, but other archs could mean other checks.

However it needs to adds a define or something similar to know if the archs implements or not this interface. Suggestions about this point?

Thanks.

Marco

#5 lore raw
Quoted 2 lines

However it needs to adds a define or something similar to know if the archs implements or not this interface. Suggestions about this point?

That's fine, you can do that in Kconfig. I think you already had that in the earlier patch: just give it a better name.

-Andi

Signature · 1 line
[email protected] -- Speaking for myself only.