Skip to contentSkip to conversation

uaccess: add probe_kernel_write()

PATCH In review Ingo Molnar 4 participants 5 messages Feb 10, 2008 – Feb 11, 2008

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

PATCH 2/6 uaccess: add probe_kernel_write()

From: Ingo Molnar <[email protected]>

add probe_kernel_write() - copy & paste of the existing probe_kernel_access(), extended to writes.

---
 include/linux/uaccess.h |   22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
include/linux/uaccess.h +22 -0
Index: linux-kgdb.q/include/linux/uaccess.h
===================================================================
@@ -84,4 +84,26 @@ static inline unsigned long __copy_from_
 		ret;					\
 	})
 
+/**
+ * probe_kernel_write(): safely attempt to write to a location
+ * @addr: address to write to - its type is type typeof(rdval)*
+ * @rdval: write to this variable
+ *
+ * Safely write to address @addr from variable @rdval.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */
+#define probe_kernel_write(addr, rdval)			\
+	({						\
+		long ret;				\
+		mm_segment_t old_fs = get_fs();		\
+							\
+		set_fs(KERNEL_DS);			\
+		pagefault_disable();			\
+		ret = __put_user(rdval,			\
+	 (__force typeof(rdval) __user *)(addr));	\
+		pagefault_enable();			\
+		set_fs(old_fs);				\
+		ret;					\
+	})
+
 #endif		/* __LINUX_UACCESS_H__ */
#1 lore raw
Quoted 4 lines · Ingo Molnar

From: Ingo Molnar <[email protected]>

add probe_kernel_write() - copy & paste of the existing probe_kernel_access(), extended to writes.

Along Linus' suggestion to work on larger chunks in kgdb, here are improved probe_kernel_read/write helpers that take a size argument.

I'm leaving the original probe_kernel_read as is for now, but maybe it can be converted to probe_kernel_read (I don't want to dive into this as well :) ).

---
 uaccess.h |   38 +++++++++++++++++++++++++++++++-------
 1 file changed, 31 insertions(+), 7 deletions(-)
include/linux/uaccess.h +31 -7
index 98cfe02..ef75869 100644
@@ -85,25 +85,49 @@ static inline unsigned long __copy_from_user_nocache(void *to,
 	})
 
 /**
+ * probe_kernel_read(): safely attempt to read from a location
+ * @dst: pointer to the buffer that shall take the data
+ * @src: address to read from
+ * @size: size of the data chunk
+ *
+ * Safely read from address @src to the buffer at @dst.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */
+#define probe_kernel_read(dst, src, size)		\
+	({						\
+		long ret;				\
+		mm_segment_t old_fs = get_fs();		\
+							\
+		set_fs(KERNEL_DS);			\
+		pagefault_disable();			\
+		ret = __copy_from_user_inatomic((dst),	\
+	 (__force const void __user *)(src), (size));	\
+		pagefault_enable();			\
+		set_fs(old_fs);				\
+		ret ? -EFAULT : 0;			\
+	})
+
+/**
  * probe_kernel_write(): safely attempt to write to a location
- * @addr: address to write to - its type is type typeof(rdval)*
- * @rdval: write to this variable
+ * @dst: address to write to
+ * @src: pointer to the data that shall be written
+ * @size: size of the data chunk
  *
- * Safely write to address @addr from variable @rdval.  If a kernel fault
+ * Safely write to address @dst from the buffer at @src.  If a kernel fault
  * happens, handle that and return -EFAULT.
  */
-#define probe_kernel_write(addr, rdval)			\
+#define probe_kernel_write(dst, src, size)		\
 	({						\
 		long ret;				\
 		mm_segment_t old_fs = get_fs();		\
 							\
 		set_fs(KERNEL_DS);			\
 		pagefault_disable();			\
-		ret = __put_user(rdval,			\
-	 (__force typeof(rdval) __user *)(addr));	\
+		ret = __copy_to_user_inatomic(		\
+	 (__force void __user *)(dst), (src), (size));	\
 		pagefault_enable();			\
 		set_fs(old_fs);				\
-		ret;					\
+		ret ? -EFAULT : 0;			\
 	})
 
 #endif		/* __LINUX_UACCESS_H__ */
#2 lore raw
Quoted 3 lines · Jan Kiszka

Along Linus' suggestion to work on larger chunks in kgdb, here are improved probe_kernel_read/write helpers that take a size argument.

I don't think this is good.

Make it
- a function, not a #define
- preferably uninlined (this does *not* look performance-critical)
- get rid of the get_fs/set_fs/set_fs dance

The last point is somewhat debatable, but the fact is, those functions are not defined to be safe to be called from interrupt context. And since we use the "__copy_xxxx()" functions with two underscores, those really are supposed to mean that we've checked the address earlier.

So it's possible that some architecture does need the explicit segment munging (S390 comes to mind), but in that case, we really should add special support for that, or we should really validate that get_fs/set_fs are safe in all contexts.

And regardless, if we *do* end up needing to munge some segment register (which on other architectures tends to be an address space identifier, not a segment, but whatever), that just makes it even more clear that this isn't some macro or inline function, because those things tend to explode the code-space.

Linus

#3 lore raw
Quoted 9 lines · Linus Torvalds
Quoted 2 lines

Along Linus' suggestion to work on larger chunks in kgdb, here are improved probe_kernel_read/write helpers that take a size argument.

I don't think this is good.

Make it
- a function, not a #define
- preferably uninlined (this does *not* look performance-critical)
- get rid of the get_fs/set_fs/set_fs dance

yeah. I've done that (see the patch below), and i've just tested that kgdb memory accesses work fine with that:

(gdb) disassemble 0xc0153ed9 0xc0153eff
Dump of assembler code from 0xc0153ed9 to 0xc0153eff:

 0xc0153ed9:     sfence
 0xc0153edc:     xchg   %ax,%ax
 0xc0153edf:     pop    %ebp
 0xc0153ee0:     movl   $0x0,0xc0a48088
 0xc0153eea:     ret
 0xc0153eeb:     push   %ebp
 0xc0153eec:     mov    %esp,%ebp
 0xc0153eee:     push   $0xc058f11d
 0xc0153ef3:     movl   $0x0,0xc0a4bf4c
 0xc0153efd:     call   0xc0126ec5

End of assembler dump.

i have to say, it's quite nice that via kgdb i can _see_ what the paravirt and alternatives stuff ends up patching into our binary image - see the 'sfence' instruction above. Unfortunately looking at the vmlinux is not as reliable as it used to be ;-)

( i've added a separate file for it under mm/maccess.c, because these
functions will be needed on NOMMU kernel too, so i couldnt move them into their natural place, mm/memory.c. )

Ingo

---------------->
Subject: uaccess: add probe_kernel_write()
From: Ingo Molnar <[email protected]>

add probe_kernel_read() and probe_kernel_write().

Uninlined and restricted to kernel range memory only, as suggested by Linus.

---
 include/linux/uaccess.h |   22 ++++++++++++++++++++++
 mm/Makefile             |    2 +-
 mm/maccess.c            |   46 ++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 69 insertions(+), 1 deletion(-)
include/linux/uaccess.h +22 -0
Index: linux-kgdb.q/include/linux/uaccess.h
===================================================================
@@ -84,4 +84,26 @@ static inline unsigned long __copy_from_
 		ret;					\
 	})
 
+/*
+ * probe_kernel_read(): safely attempt to read from a location
+ * @dst: pointer to the buffer that shall take the data
+ * @src: address to read from
+ * @size: size of the data chunk
+ *
+ * Safely read from address @src to the buffer at @dst.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */
+extern long probe_kernel_read(void *dst, void *src, size_t size);
+
+/*
+ * probe_kernel_write(): safely attempt to write to a location
+ * @dst: address to write to
+ * @src: pointer to the data that shall be written
+ * @size: size of the data chunk
+ *
+ * Safely write to address @dst from the buffer at @src.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */
+extern long probe_kernel_write(void *dst, void *src, size_t size);
+
 #endif		/* __LINUX_UACCESS_H__ */
mm/Makefile +1 -1
Index: linux-kgdb.q/mm/Makefile
===================================================================
@@ -8,7 +8,7 @@ mmu-$(CONFIG_MMU)	:= fremap.o highmem.o 
 			   vmalloc.o
 
 obj-y			:= bootmem.o filemap.o mempool.o oom_kill.o fadvise.o \
-			   page_alloc.o page-writeback.o pdflush.o \
+			   maccess.o page_alloc.o page-writeback.o pdflush.o \
 			   readahead.o swap.o truncate.o vmscan.o \
 			   prio_tree.o util.o mmzone.o vmstat.o backing-dev.o \
 			   page_isolation.o $(mmu-y)
mm/maccess.c new file +46 -0

46 lines not on this page · raw message

#4 lore raw
Quoted 20 lines · On Sun, 10 Feb 2008 21:05:40 +0100 Ingo Molnar wrote:

Subject: uaccess: add probe_kernel_write()
From: Ingo Molnar <[email protected]>

add probe_kernel_read() and probe_kernel_write().

Uninlined and restricted to kernel range memory only, as suggested by Linus.

---
 include/linux/uaccess.h |   22 ++++++++++++++++++++++
 mm/Makefile             |    2 +-
 mm/maccess.c            |   46 ++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 69 insertions(+), 1 deletion(-)

Index: linux-kgdb.q/include/linux/uaccess.h
===================================================================

include/linux/uaccess.h static inline unsigned long __copy_from_ · line 88

Quoted 5 lines
include/linux/uaccess.h +2 -0
@@ -84,4 +84,26 @@ static inline unsigned long __copy_from_
 		ret;					\
 	})
 
+/*
+ * probe_kernel_read(): safely attempt to read from a location

Please insert a hyphen/dash/'-' between function name and its short description on the line above and on other similar lines.

Documentation/kernel-doc-nano-HOWTO.txt .

Quoted 63 lines
include/linux/uaccess.h +20 -0
+ * @dst: pointer to the buffer that shall take the data
+ * @src: address to read from
+ * @size: size of the data chunk
+ *
+ * Safely read from address @src to the buffer at @dst.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */
+extern long probe_kernel_read(void *dst, void *src, size_t size);
+
+/*
+ * probe_kernel_write(): safely attempt to write to a location
+ * @dst: address to write to
+ * @src: pointer to the data that shall be written
+ * @size: size of the data chunk
+ *
+ * Safely write to address @dst from the buffer at @src.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */
+extern long probe_kernel_write(void *dst, void *src, size_t size);
+
 #endif		/* __LINUX_UACCESS_H__ */

Index: linux-kgdb.q/mm/maccess.c
===================================================================

mm/maccess.c new file +36 -0
@@ -0,0 +1,46 @@
+/*
+ * Access kernel memory without faulting.
+ */
+#include <linux/uaccess.h>
+#include <linux/mm.h>
+
+/**
+ * probe_kernel_read(): safely attempt to read from a location
+ * @dst: pointer to the buffer that shall take the data
+ * @src: address to read from
+ * @size: size of the data chunk
+ *
+ * Safely read from address @src to the buffer at @dst.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */
+long probe_kernel_read(void *dst, void *src, size_t size)
+{
+	long ret;
+
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(dst,
+			(__force const void __user *)src, size);
+	pagefault_enable();
+
+	return ret ? -EFAULT : 0;
+}
+
+/**
+ * probe_kernel_write(): safely attempt to write to a location
+ * @dst: address to write to
+ * @src: pointer to the data that shall be written
+ * @size: size of the data chunk
+ *
+ * Safely write to address @dst from the buffer at @src.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ */

mm/maccess.c line 46

Quoted 10 lines
mm/maccess.c new file +10 -0
+long probe_kernel_write(void *dst, void *src, size_t size)
+{
+	long ret;
+
+	pagefault_disable();
+	ret = __copy_to_user_inatomic((__force void __user *)dst, src, size);
+	pagefault_enable();
+
+	return ret ? -EFAULT : 0;
+}

---
~Randy