strcpy returns NULL pointer and not destination pointer

strcpy returns NULL pointer and not destination pointer

[Bob Picco]

Hi Ken:

I'm assuming you can speak to this because your email address is in
memcpy_mck.S.

The removal of -ffreestanding compiler flag in 2.6.17-rcX (commit id
6edfba1b33c701108717f4e036320fc39abe1912) causes strcpy calls 
with a known size for the source string to be replaced with memcpy which
is faster.  ia64 memcpy has return values which are unlike stdlibc and
the other arch memcpy routines examined by me in the kernel.  The ia64 return
values are 0 for success and number of bytes copied for failure. Thus any
instance of pointer = strcpy(dest, "some string") has zero assigned to
pointer.  I detected this when testing kgdb on 2.6.17-rc3.  

The implementation in memcpy_mck.S doesn't leave a single scratch
register or predicate. I didn't pursue very long for a solution.

thanks,

bob

RE: strcpy returns NULL pointer and not destination pointer

[Chen, Kenneth W]

Bob Picco wrote on Wednesday, May 03, 2006 6:38 AM
> The removal of -ffreestanding compiler flag in 2.6.17-rcX (commit id
> 6edfba1b33c701108717f4e036320fc39abe1912) causes strcpy calls 
> with a known size for the source string to be replaced with memcpy which
> is faster.  ia64 memcpy has return values which are unlike stdlibc and
> the other arch memcpy routines examined by me in the kernel.  The ia64 return
> values are 0 for success and number of bytes copied for failure. Thus any
> instance of pointer = strcpy(dest, "some string") has zero assigned to
> pointer.  I detected this when testing kgdb on 2.6.17-rc3.  
> 
> The implementation in memcpy_mck.S doesn't leave a single scratch
> register or predicate. I didn't pursue very long for a solution.

I thought I fixed that about 3 years ago:

http://marc.theaimsgroup.com/?l=linux-ia64&m\x105590709805435&w=2

Did it fall through the crack?  I'm not sure what's going on.

- Ken

Re: strcpy returns NULL pointer and not destination pointer

[Bob Picco]

Chen, Kenneth W wrote:	[Wed May 03 2006, 02:08:47PM EDT]
> Bob Picco wrote on Wednesday, May 03, 2006 6:38 AM
> > The removal of -ffreestanding compiler flag in 2.6.17-rcX (commit id
> > 6edfba1b33c701108717f4e036320fc39abe1912) causes strcpy calls 
> > with a known size for the source string to be replaced with memcpy which
> > is faster.  ia64 memcpy has return values which are unlike stdlibc and
> > the other arch memcpy routines examined by me in the kernel.  The ia64 return
> > values are 0 for success and number of bytes copied for failure. Thus any
> > instance of pointer = strcpy(dest, "some string") has zero assigned to
> > pointer.  I detected this when testing kgdb on 2.6.17-rc3.  
> > 
> > The implementation in memcpy_mck.S doesn't leave a single scratch
> > register or predicate. I didn't pursue very long for a solution.
> 
> I thought I fixed that about 3 years ago:
> 
> http://marc.theaimsgroup.com/?l=linux-ia64&m\x105590709805435&w=2
> 
> Did it fall through the crack?  I'm not sure what's going on.
> 
> - Ken
It looks to have definitely fallen through a crack or two.

I applied to rc3 and boot tested for you. I assume you'll post patch again.

bob

RE: strcpy returns NULL pointer and not destination pointer

[Chen, Kenneth W]

Bob Picco wrote on Wednesday, May 03, 2006 11:31 AM
> Chen, Kenneth W wrote:	[Wed May 03 2006, 02:08:47PM EDT]
> > Bob Picco wrote on Wednesday, May 03, 2006 6:38 AM
> > > The removal of -ffreestanding compiler flag in 2.6.17-rcX (commit id
> > > 6edfba1b33c701108717f4e036320fc39abe1912) causes strcpy calls 
> > > with a known size for the source string to be replaced with memcpy which
> > > is faster.  ia64 memcpy has return values which are unlike stdlibc and
> > > the other arch memcpy routines examined by me in the kernel.  The ia64 return
> > > values are 0 for success and number of bytes copied for failure. Thus any
> > > instance of pointer = strcpy(dest, "some string") has zero assigned to
> > > pointer.  I detected this when testing kgdb on 2.6.17-rc3.  
> > > 
> > > The implementation in memcpy_mck.S doesn't leave a single scratch
> > > register or predicate. I didn't pursue very long for a solution.
> > 
> > I thought I fixed that about 3 years ago:
> > 
> > http://marc.theaimsgroup.com/?l=linux-ia64&m\x105590709805435&w=2
> > 
> > Did it fall through the crack?  I'm not sure what's going on.
> 
> It looks to have definitely fallen through a crack or two.
> 
> I applied to rc3 and boot tested for you. I assume you'll post patch again.

Wow, 3 years ago patch still applies.  I'm impressed, that doesn't happen
very often ;-)

Here is an updated version, with comments updated (bcopy was removed).




[patch] fix return value of memcpy

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>


--- ./arch/ia64/lib/memcpy_mck.S.orig	2006-03-19 21:53:29.000000000 -0800
+++ ./arch/ia64/lib/memcpy_mck.S	2006-05-03 12:30:20.000000000 -0700
@@ -6,7 +6,9 @@
  *	in1:	source address
  *	in2:	number of bytes to copy
  * Output:
- * 	0 if success, or number of byte NOT copied if error occurred.
+ *	for memcpy:    retrun dest
+ * 	for copy_user: return 0 if success,
+ *		       or number of byte NOT copied if error occurred.
  *
  * Copyright (C) 2002 Intel Corp.
  * Copyright (C) 2002 Ken Chen <kenneth.w.chen@intel.com>
@@ -73,6 +75,7 @@ GLOBAL_ENTRY(memcpy)
 	and	r28=0x7,in0
 	and	r29=0x7,in1
 	mov	f6ð
+	mov	retval=in0
 	br.cond.sptk .common_code
 	;;
 END(memcpy)
@@ -84,7 +87,7 @@ GLOBAL_ENTRY(__copy_user)
 	mov	f6ñ
 	mov	saved_in0=in0	// save dest pointer
 	mov	saved_in1=in1	// save src pointer
-	mov	saved_in2=in2	// save len
+	mov	retval=r0	// initialize return value
 	;;
 .common_code:
 	cmp.gt	p15,p0=8,in2	// check for small size
@@ -92,7 +95,7 @@ GLOBAL_ENTRY(__copy_user)
 	cmp.ne	p14,p0=0,r29	// check src alignment
 	add	src0=0,in1
 	sub	r30=8,r28	// for .align_dest
-	mov	retval=r0	// initialize return value
+	mov	saved_in2=in2	// save len
 	;;
 	add	dst0=0,in0
 	add	dst1=1,in0	// dest odd index

Re: strcpy returns NULL pointer and not destination pointer

[Matthew Wilcox]

On Wed, May 03, 2006 at 11:46:07AM -0700, Chen, Kenneth W wrote:
> - * 	0 if success, or number of byte NOT copied if error occurred.
> + *	for memcpy:    retrun dest

typo

RE: strcpy returns NULL pointer and not destination pointer

[Chen, Kenneth W]

Matthew Wilcox wrote on Wednesday, May 03, 2006 11:49 AM
> On Wed, May 03, 2006 at 11:46:07AM -0700, Chen, Kenneth W wrote:
> > - * 	0 if success, or number of byte NOT copied if error occurred.
> > + *	for memcpy:    retrun dest
> 
> typo


Thank you.  I need a spell-o-matic checker :-)



[patch] fix return value of memcpy

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>


--- ./arch/ia64/lib/memcpy_mck.S.orig	2006-03-19 21:53:29.000000000 -0800
+++ ./arch/ia64/lib/memcpy_mck.S	2006-05-03 12:30:20.000000000 -0700
@@ -6,7 +6,9 @@
  *	in1:	source address
  *	in2:	number of bytes to copy
  * Output:
- * 	0 if success, or number of byte NOT copied if error occurred.
+ *	for memcpy:    return dest
+ * 	for copy_user: return 0 if success,
+ *		       or number of byte NOT copied if error occurred.
  *
  * Copyright (C) 2002 Intel Corp.
  * Copyright (C) 2002 Ken Chen <kenneth.w.chen@intel.com>
@@ -73,6 +75,7 @@ GLOBAL_ENTRY(memcpy)
 	and	r28=0x7,in0
 	and	r29=0x7,in1
 	mov	f6ð
+	mov	retval=in0
 	br.cond.sptk .common_code
 	;;
 END(memcpy)
@@ -84,7 +87,7 @@ GLOBAL_ENTRY(__copy_user)
 	mov	f6ñ
 	mov	saved_in0=in0	// save dest pointer
 	mov	saved_in1=in1	// save src pointer
-	mov	saved_in2=in2	// save len
+	mov	retval=r0	// initialize return value
 	;;
 .common_code:
 	cmp.gt	p15,p0=8,in2	// check for small size
@@ -92,7 +95,7 @@ GLOBAL_ENTRY(__copy_user)
 	cmp.ne	p14,p0=0,r29	// check src alignment
 	add	src0=0,in1
 	sub	r30=8,r28	// for .align_dest
-	mov	retval=r0	// initialize return value
+	mov	saved_in2=in2	// save len
 	;;
 	add	dst0=0,in0
 	add	dst1=1,in0	// dest odd index