Skip to contentSkip to conversation

x86: use bitmap library functions

PATCH In review Akinobu Mita 2 participants 5 messages Feb 16, 2011 – Feb 17, 2011 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86: use bitmap library functions

Use bitmap_set()/bitmap_clear() to fill/zero a region of a bitmap instead of doing set_bit()/clear_bit() each bit.

Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc "H. Peter Anvin" <[email protected]>
---
 arch/x86/kernel/ioport.c |   20 +++++---------------
 1 files changed, 5 insertions(+), 15 deletions(-)
arch/x86/kernel/ioport.c +5 -15
index 8eec0ec..8c96897 100644
@@ -14,22 +14,9 @@
 #include <linux/slab.h>
 #include <linux/thread_info.h>
 #include <linux/syscalls.h>
+#include <linux/bitmap.h>
 #include <asm/syscalls.h>
 
-/* Set EXTENT bits starting at BASE in BITMAP to value TURN_ON. */
-static void set_bitmap(unsigned long *bitmap, unsigned int base,
-		       unsigned int extent, int new_value)
-{
-	unsigned int i;
-
-	for (i = base; i < base + extent; i++) {
-		if (new_value)
-			__set_bit(i, bitmap);
-		else
-			__clear_bit(i, bitmap);
-	}
-}
-
 /*
  * this changes the io permissions bitmap in the current task.
  */
@@ -69,7 +56,10 @@ asmlinkage long sys_ioperm(unsigned long from, unsigned long num, int turn_on)
 	 */
 	tss = &per_cpu(init_tss, get_cpu());
 
-	set_bitmap(t->io_bitmap_ptr, from, num, !turn_on);
+	if (turn_on)
+		bitmap_clear(t->io_bitmap_ptr, from, num);
+	else
+		bitmap_set(t->io_bitmap_ptr, from, num);
 
 	/*
 	 * Search for a (possibly new) maximum. This is simple and stupid,
Signature · 1 line
1.7.4
#1 lore raw
Quoted 11 lines · Akinobu Mita

Use bitmap_set()/bitmap_clear() to fill/zero a region of a bitmap instead of doing set_bit()/clear_bit() each bit.

Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc "H. Peter Anvin" <[email protected]>
---
 arch/x86/kernel/ioport.c |   20 +++++---------------
 1 files changed, 5 insertions(+), 15 deletions(-)

Before we can merge this it would be nice to tested this using sys_ioperm(), to make sure port IO still gets denied/allowed as expected. You can check /proc/ioperm to see which ports are (probably) safe to read: 0x80/0xed is usually safe to read/write.

Ingo

#2 lore raw

2011/2/17 Ingo Molnar <[email protected]>:

Quoted 15 lines · Akinobu Mita
Quoted 4 lines

Use bitmap_set()/bitmap_clear() to fill/zero a region of a bitmap instead of doing set_bit()/clear_bit() each bit.

Quoted 8 lines
Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc "H. Peter Anvin" <[email protected]>

---
 arch/x86/kernel/ioport.c |   20 +++++---------------

 1 files changed, 5 insertions(+), 15 deletions(-)
Quoted 4 lines

Before we can merge this it would be nice to tested this using sys_ioperm(), to make sure port IO still gets denied/allowed as expected. You can check /proc/ioperm to see which ports are (probably) safe to read: 0x80/0xed is usually safe to read/write.

I have tested with this program and there is no difference in behavior with this patch.

#include <sys/io.h>
#include <stdio.h>
#include <err.h>

int main(void)

{
	unsigned long from = 0xed;
	unsigned long num = 1;
	int ret;
	int i;

	ret = ioperm(from, num, 1);
	if (ret)
		err(1, "ioperm");

	for (i = 0; i < num; i++)
		ret = inb(from + i);

	printf("OK\n");

	ret = ioperm(from, num, 0);
	if (ret)
		err(1, "ioperm");

	printf("will receive a segmentation fault\n");

	ret = inb(from);
	errx(1, "Oops");
}
#3 lore raw
Quoted 16 lines · Akinobu Mita

2011/2/17 Ingo Molnar <[email protected]>:

Quoted 15 lines · Akinobu Mita
Quoted 12 lines

Use bitmap_set()/bitmap_clear() to fill/zero a region of a bitmap instead of doing set_bit()/clear_bit() each bit.

Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc "H. Peter Anvin" <[email protected]>

---
 arch/x86/kernel/ioport.c |   20 +++++---------------

 1 files changed, 5 insertions(+), 15 deletions(-)
Quoted 7 lines
Quoted 4 lines

Before we can merge this it would be nice to tested this using sys_ioperm(), to make sure port IO still gets denied/allowed as expected. You can check /proc/ioperm to see which ports are (probably) safe to read: 0x80/0xed is usually safe to read/write.

I have tested with this program and there is no difference in behavior with this patch.

Great - i've queued it up in tip:x86/asm.

Thanks,

Ingo

#4 lore raw

[tip:x86/asm] x86: Use bitmap library functions

In reply to Akinobu Mita
Commit-ID:  da1016df85ed67b6f7dbb765532c54bc35ba08d7
Gitweb:     http://git.kernel.org/tip/da1016df85ed67b6f7dbb765532c54bc35ba08d7
Author:     Akinobu Mita <[email protected]>
AuthorDate: Wed, 16 Feb 2011 23:48:35 +0900
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Feb 2011 14:59:22 +0100

x86: Use bitmap library functions

Use bitmap_set()/bitmap_clear() to fill/zero a region of a bitmap instead of doing set_bit()/clear_bit() each bit.

This change has been tested with ioperm() and there's no change in behavior.

LKML-Reference: <[email protected]>

---
 arch/x86/kernel/ioport.c |   20 +++++---------------
 1 files changed, 5 insertions(+), 15 deletions(-)
arch/x86/kernel/ioport.c +5 -15
index 8eec0ec..8c96897 100644
@@ -14,22 +14,9 @@
 #include <linux/slab.h>
 #include <linux/thread_info.h>
 #include <linux/syscalls.h>
+#include <linux/bitmap.h>
 #include <asm/syscalls.h>
 
-/* Set EXTENT bits starting at BASE in BITMAP to value TURN_ON. */
-static void set_bitmap(unsigned long *bitmap, unsigned int base,
-		       unsigned int extent, int new_value)
-{
-	unsigned int i;
-
-	for (i = base; i < base + extent; i++) {
-		if (new_value)
-			__set_bit(i, bitmap);
-		else
-			__clear_bit(i, bitmap);
-	}
-}
-
 /*
  * this changes the io permissions bitmap in the current task.
  */
@@ -69,7 +56,10 @@ asmlinkage long sys_ioperm(unsigned long from, unsigned long num, int turn_on)
 	 */
 	tss = &per_cpu(init_tss, get_cpu());
 
-	set_bitmap(t->io_bitmap_ptr, from, num, !turn_on);
+	if (turn_on)
+		bitmap_clear(t->io_bitmap_ptr, from, num);
+	else
+		bitmap_set(t->io_bitmap_ptr, from, num);
 
 	/*
 	 * Search for a (possibly new) maximum. This is simple and stupid,