XIP fails on kernels 2.6.12 and higher

XIP fails on kernels 2.6.12 and higher

[Korolev, Alexey]

Hi all,

I can't bootup XIP linux kernel on Mainstone platform (based on PXA270
processor). I tried several different kernel versions 
2.6.9, 2.6.11 - bootup with enabled XIP kernel feature fine.
But 2.6.12 and 2.6.13 - faults at the start of booting. 
I found very big changes between 2.6.12 and 2.6.11 in memory management
code. I suspect it brokes XIP. 

Do you know this problem?
Has XIP been broken on all platforms or just on Mainstone platform only?



Here is kernel failure log. (It's rather short :-) )

Linux version 2.6.13.2 (root@localhost.localdomain) (gcc version 3.4.3)
#7 Thu S
ep 22 19:51:25 MSD 2005
CPU: XScale-PXA270 [69054114] revision 4 (ARMv5TE)
Machine: Intel HCDDBBVA0 Development Platform (aka Mainstone)
Memory policy: ECC disabled, Data cache writeback
Run Mode clock: 208.00MHz (*16)
Turbo Mode clock: 416.00MHz (*2.0, active)
Memory clock: 104.00MHz (/2)
System bus clock: 104.00MHz
CPU0: D VIVT undefined 5 cache
CPU0: I cache: 32768 bytes, associativity 32, 32 byte lines, 32 sets
CPU0: D cache: 32768 bytes, associativity 32, 32 byte lines, 32 sets
Built 1 zonelists
Kernel command line: console=ttyS1,115200 mem=64M
mtdparts=phys_mapped_flash:512
k(blob)ro,2m(kernel)ro,16m(root),10m(vol1) rootfstype=jffs2 root=31:02
ro
PID hash table entries: 512 (order: 9, 8192 bytes)
Console: colour dummy device 80x30
Dentry cache hash table entries: 16384 (order: 4, 65536 bytes)
Inode-cache hash table entries: 8192 (order: 3, 32768 bytes)
Unhandled fault: imprecise external abort (0x406) at 0x00000000

Re: XIP fails on kernels 2.6.12 and higher

[Nicolas Pitre]

On Thu, 22 Sep 2005, Korolev, Alexey wrote:

> Hi all,
> 
> I can't bootup XIP linux kernel on Mainstone platform (based on PXA270
> processor). I tried several different kernel versions 
> 2.6.9, 2.6.11 - bootup with enabled XIP kernel feature fine.
> But 2.6.12 and 2.6.13 - faults at the start of booting. 
> I found very big changes between 2.6.12 and 2.6.11 in memory management
> code. I suspect it brokes XIP. 

If you can find the exact intermediate release between 2.6.11 and 2.6.12 
that broke it that would be extrelely helpful.

> Do you know this problem?

It has been reported in the past.

> Has XIP been broken on all platforms or just on Mainstone platform only?

This is generic code so probably on all platforms.


Nicolas

Re: XIP fails on kernels 2.6.12 and higher

[Jared Hulbert]

> If you can find the exact intermediate release between 2.6.11 and 2.6.12
> that broke it that would be extrelely helpful.

2.6.11.12 works and 2.6.12 does not.  I'll see if I can't find an
intermediate release.

> This is generic code so probably on all platforms.

In mm/bootmem.c free_all_bootmem_core() does not return.   It seems
after i hits 288 if I step I get a imprecise data abort.  It seems to
occur at the __ClearPageReserved() which is an inlined __clear_bit()
from include/asm/bitops.h.  It does this at somewhere around
page=0xC0042400 which seems perfectly normal and is readable/writeable
with the JTAG debugger.  Its really hard to replicate but it seems
like a LDR op code is always at fault.  I've tried 2 different
toolchains and compiling with -O0, while the LDR command was always a
little different it was basically a read to the same address range.  
It is really baffling me.  Nothing seems wrong with what it is doing. 
I almost suspect hardware.  I'm also testing on a Mainstone2 like
Alexey.  My OSK got destroyed.  Can somebody confirm xipImage of
2.6.12 fails to boot where 2.6.11 runs on a different hardware
platform?

Re: XIP fails on kernels 2.6.12 and higher

[Konstantin Kletschke]

* Jared Hulbert <jaredeh@gmail.com> [Thu, Sep 29, 2005 at 03:51:50PM -0700]:

> Alexey.  My OSK got destroyed.  Can somebody confirm xipImage of
> 2.6.12 fails to boot where 2.6.11 runs on a different hardware
> platform?

http://lists.infradead.org/pipermail/linux-mtd/2005-September/013536.html

Is about the same. I can't get it running on 2.6.13.
I dont know anymore how I got 2.6.12 working or digged around the 
problems but IIRC there where several regarding the XIP stuff.

Konsti

Re: XIP fails on kernels 2.6.12 and higher

[Jared Hulbert]

On 9/29/05, Jared Hulbert <jaredeh@gmail.com> wrote:
> > If you can find the exact intermediate release between 2.6.11 and 2.6.12
> > that broke it that would be extrelely helpful.

I found it.  This is the change that was causing my problem
-----------------------------------------------------------------------------------
commit 3871b11e8c07c9e6652f81ae5afc1ae3bb120c40
Author: mpm <mpm>
Date:   Tue Mar 8 18:05:40 2005 +0000

    [PATCH] lib/sort: Replace insertion sort in exception tables

    Replace exception table insertion sort with lib/sort

    Signed-off-by: Matt Mackall <mpm@selenic.com>
    Signed-off-by: Andrew Morton <akpm@osdl.org>
    Signed-off-by: Linus Torvalds <torvalds@osdl.org>

    BKrev: 422de974WUZpIt5eM36-PMJe_h6Nfg
-----------------------------------------------------------------------------------

The following is the diff from the old git repository.  Unpatching
this in the 2.6.14-rc4 kernel made the xipImage work.  What is going
on???

Index: lib/extable.c
===================================================================
--- lib/extable.c	(revision 29)
+++ lib/extable.c	(revision 28)
@@ -13,6 +13,7 @@
 #include <linux/config.h>
 #include <linux/module.h>
 #include <linux/init.h>
+#include <linux/sort.h>
 #include <asm/uaccess.h>

 extern struct exception_table_entry __start___ex_table[];
@@ -25,26 +26,17 @@
  * This is used both for the kernel exception table and for
  * the exception tables of modules that get loaded.
  */
+static int cmp_ex(const void *a, const void *b)
+{
+	const struct exception_table_entry *x = a, *y = b;
+	return x->insn - y->insn;
+}
+
 void sort_extable(struct exception_table_entry *start,
 		  struct exception_table_entry *finish)
 {
-	struct exception_table_entry el, *p, *q;
-
-	/* insertion sort */
-	for (p = start + 1; p < finish; ++p) {
-		/* start .. p-1 is sorted */
-		if (p[0].insn < p[-1].insn) {
-			/* move element p down to its right place */
-			el = *p;
-			q = p;
-			do {
-				/* el comes before q[-1], move q[-1] up one */
-				q[0] = q[-1];
-				--q;
-			} while (q > start && el.insn < q[-1].insn);
-			*q = el;
-		}
-	}
+	sort(start, finish - start, sizeof(struct exception_table_entry),
+	     cmp_ex, NULL);
 }
 #endif

Re: XIP fails on kernels 2.6.12 and higher

[Nicolas Pitre]

On Wed, 12 Oct 2005, Jared Hulbert wrote:

> On 9/29/05, Jared Hulbert <jaredeh@gmail.com> wrote:
> > > If you can find the exact intermediate release between 2.6.11 and 2.6.12
> > > that broke it that would be extrelely helpful.
> 
> I found it.  This is the change that was causing my problem
> -----------------------------------------------------------------------------------
> commit 3871b11e8c07c9e6652f81ae5afc1ae3bb120c40
> Author: mpm <mpm>
> Date:   Tue Mar 8 18:05:40 2005 +0000
> 
>     [PATCH] lib/sort: Replace insertion sort in exception tables
> 
>     Replace exception table insertion sort with lib/sort

Are you using JFFS2?

Does it still crash if you don't?


Nicolas

Re: XIP fails on kernels 2.6.12 and higher

[Konstantin Kletschke]

At Wed, 12 Oct 2005 16:23:37 -0400 (EDT),
Nicolas Pitre <nico@cam.org> wrote:

> Are you using JFFS2?
> 
> Does it still crash if you don't?

I am curious if both times yes, could this be related to

http://lists.infradead.org/pipermail/linux-mtd/2005-September/013882.html

?

I will try out tomorrow...

Konsti

-- 
GPG KeyID EF62FCEF
Fingerprint: 13C9 B16B 9844 EC15 CC2E  A080 1E69 3FDA EF62 FCEF

Re: XIP fails on kernels 2.6.12 and higher

[Jared Hulbert]

> Are you using JFFS2?
>
> Does it still crash if you don't?

No.

Opps, this is on the wrong list, I'm not even building with the MTD. 
I meant to post this on arm-linux.  Watch for it there, please.

Re: XIP fails on kernels 2.6.12 and higher

[Konstantin Kletschke]

* Korolev, Alexey <alexey.korolev@intel.com> [Thu, Sep 22, 2005 at 08:08:39PM +0400]:

> Linux version 2.6.13.2 (root@localhost.localdomain) (gcc version 3.4.3)
> #7 Thu S
> ep 22 19:51:25 MSD 2005
> CPU: XScale-PXA270 [69054114] revision 4 (ARMv5TE)
> Machine: Intel HCDDBBVA0 Development Platform (aka Mainstone)
> Memory policy: ECC disabled, Data cache writeback
> Run Mode clock: 208.00MHz (*16)
> Turbo Mode clock: 416.00MHz (*2.0, active)
> Memory clock: 104.00MHz (/2)
> System bus clock: 104.00MHz
> CPU0: D VIVT undefined 5 cache
> CPU0: I cache: 32768 bytes, associativity 32, 32 byte lines, 32 sets
> CPU0: D cache: 32768 bytes, associativity 32, 32 byte lines, 32 sets
> Built 1 zonelists
> Kernel command line: console=ttyS1,115200 mem=64M
> mtdparts=phys_mapped_flash:512
> k(blob)ro,2m(kernel)ro,16m(root),10m(vol1) rootfstype=jffs2 root=31:02
> ro
> PID hash table entries: 512 (order: 9, 8192 bytes)
> Console: colour dummy device 80x30
> Dentry cache hash table entries: 16384 (order: 4, 65536 bytes)
> Inode-cache hash table entries: 8192 (order: 3, 32768 bytes)
> Unhandled fault: imprecise external abort (0x406) at 0x00000000


My 2.6.13 breaks near the same location:

Linux version 2.6.13-imx1 (konsti@synertronixx3) (gcc version 3.3.5) #58 Wed Oct 5 11:38:26 CEST 2005
CPU: ARM920Tid(wb) [41129200] revision 0 (ARMv4T)
Machine: synertronixx scb9328
Memory policy: ECC disabled, Data cache writeback
CPU0: D VIVT write-back cache
CPU0: I cache: 16384 bytes, associativity 64, 32 byte lines, 8 sets
CPU0: D cache: 16384 bytes, associativity 64, 32 byte lines, 8 sets
Built 1 zonelists
Kernel command line: console=ttySMX0,115200n8 root=/dev/mtdblock3 rootfstype=jffs2 mtdparts=scb9328_flash:128k(U-boot)ro,128k(U-boot_env),2m(kernel),5m(root),-(fs) board=evb9328
evb9328 board found
scb9328_init_irq
PID hash table entries: 128 (order: 7, 2048 bytes)
Dentry cache hash table entries: 4096 (order: 2, 16384 bytes)
Inode-cache hash table entries: 2048 (order: 1, 8192 bytes)

K

Re: XIP fails on kernels 2.6.12 and higher

[Konstantin Kletschke]

* Konstantin Kletschke <lists@ku-gbr.de> [Wed, Oct 05, 2005 at 01:45:11PM +0200]:

> My 2.6.13 breaks near the same location:

In order to try debugging this I found out, that ther Kernel boots
if I disable jffs2 support ... weird...

K