Git head doesn't boot on A500

Git head doesn't boot on A500

[Thibaut VARENE]

Hi pa-ckers

Just a quick failure report, I tested git head + jejb's zonelist patch
on my A500 (gcc-4.2) and it hpmc'd the box (afaict) very early during
bootup. I'll dig that more ASAP.

Console log:

Linux version 2.6.26-rc2-dirty (varenet@mkhppa3) (gcc version 4.2.3
(Debian 4.2.3-5)) #1 SMP Sat May 17 15:08:54 CEST 2008
FP[0] enabled: Rev 1 Model 16
The 64-bit Kernel has started...
console [ttyB0] enabled
Initialized PDC Console for debugging.
Determining PDC firmware type: 64 bit PAT.
model 00005d50 00000491 00000000 00000001 5cbb53d8 100000f0 00000008
000000b2 000000b2
vers  00000301
CPUID vers 18 rev 11 (0x0000024b)
capabilities 0x1
model 9000/800/A500-5X
Memory Ranges:
 0) Start 0x0000000000000000 End 0x00000000efffffff Size   3840 MB
 1) Start 0x00000010f0000000 End 0x00000010ffffffff Size    256 MB
Total Memory: 4096 MB
SMP: bootstrap CPU ID is 0
Built 2 zonelists in Zone order, mobility grouping on.  Total pages: 1034240
Kernel command line: root=/dev/sda5 console=ttyS0 TERM=vt102
palo_kernel=2/vmlinux-git
PID hash table entries: 4096 (order: 12, 32768 bytes)
Console: colour dummy device 160x64



Console reset done.

Boot device reset done.


HTH


-- 
Thibaut VARENE
http://www.parisc-linux.org/~varenet/

Re: Git head doesn't boot on A500

[Grant Grundler]

On Sat, May 17, 2008 at 03:20:46PM +0200, Thibaut VARENE wrote:
> Hi pa-ckers
> 
> Just a quick failure report, I tested git head + jejb's zonelist patch
> on my A500 (gcc-4.2) and it hpmc'd the box (afaict) very early during
> bootup. I'll dig that more ASAP.

Two questions:
1) I suspect you missed the final email of the thread:
      http://thread.gmane.org/gmane.linux.ports.parisc/301/focus=309
   where jejb conceded the original proposed fix was correct.
   Try that instead?
2) PIM dump would help point where in the kernel it HPMC'd.

I'll try current TOB from
    git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6

unless someone has a better suggestion.

thanks,
grant

Re: Git head doesn't boot on A500

[Thibaut VARENE]

On Sat, May 17, 2008 at 5:26 PM, Grant Grundler
<grundler@parisc-linux.org> wrote:
> On Sat, May 17, 2008 at 03:20:46PM +0200, Thibaut VARENE wrote:
>> Hi pa-ckers
>>
>> Just a quick failure report, I tested git head + jejb's zonelist patch
>> on my A500 (gcc-4.2) and it hpmc'd the box (afaict) very early during
>> bootup. I'll dig that more ASAP.

Kyle eventually spotted the offending patch:
http://git.kernel.org/?p=linux/kernel/git/kyle/parisc-2.6.git;a=commitdiff;h=bd3bb8c15b9a80dbddfb7905b237a4a11a4725b4;hp=492c2e476eac010962850006c49df326919b284c

Out of this patch, the only two non-offending hunks are:
diff --git a/arch/parisc/kernel/head.S b/arch/parisc/kernel/head.S
index ec2482d..5680a2c 100644 (file)
--- a/arch/parisc/kernel/head.S
+++ b/arch/parisc/kernel/head.S
@@ -32,7 +32,7 @@ ENTRY(boot_args)
        .word 0 /* arg3 */
 END(boot_args)

-       .section .text.head
+       __HEAD
        .align  4
        .import init_thread_union,data
        .import fault_vector_20,code    /* IVA parisc 2.0 32 bit */

and

diff --git a/arch/parisc/kernel/syscall.S b/arch/parisc/kernel/syscall.S
index 69b6eeb..ae509d8 100644 (file)
--- a/arch/parisc/kernel/syscall.S
+++ b/arch/parisc/kernel/syscall.S
@@ -636,7 +637,7 @@ END(sys_call_table64)
                All light-weight-syscall atomic operations
                will use this set of locks
        */
-       .section .data
+       .section .data, "aw"
        .align  PAGE_SIZE
 ENTRY(lws_lock_start)
        /* lws locks */

I'm now debugging the timer_interrupt problems, will keep the list posted.

HTH

T-Bone

Re: Git head doesn't boot on A500

[Grant Grundler]

On Sun, May 18, 2008 at 02:52:00AM +0200, Thibaut VARENE wrote:
> On Sat, May 17, 2008 at 5:26 PM, Grant Grundler
> <grundler@parisc-linux.org> wrote:
> > On Sat, May 17, 2008 at 03:20:46PM +0200, Thibaut VARENE wrote:
> >> Hi pa-ckers
> >>
> >> Just a quick failure report, I tested git head + jejb's zonelist patch
> >> on my A500 (gcc-4.2) and it hpmc'd the box (afaict) very early during
> >> bootup. I'll dig that more ASAP.
> 
> Kyle eventually spotted the offending patch:
> http://git.kernel.org/?p=linux/kernel/git/kyle/parisc-2.6.git;a=commitdiff;h=bd3bb8c15b9a80dbddfb7905b237a4a11a4725b4;hp=492c2e476eac010962850006c49df326919b284c
> 
> Out of this patch, the only two non-offending hunks are:
> diff --git a/arch/parisc/kernel/head.S b/arch/parisc/kernel/head.S
> index ec2482d..5680a2c 100644 (file)
> --- a/arch/parisc/kernel/head.S
> +++ b/arch/parisc/kernel/head.S
> @@ -32,7 +32,7 @@ ENTRY(boot_args)
>         .word 0 /* arg3 */
>  END(boot_args)
> 
> -       .section .text.head
> +       __HEAD
>         .align  4
>         .import init_thread_union,data
>         .import fault_vector_20,code    /* IVA parisc 2.0 32 bit */

So I don't understand how __HEAD works...and all the other arches are
using ".text.head" (not .head.text) which has corresponding "code" in
their vmlinux.lds.  Could someone kindly point out how .head.text 
works for parisc?

> and
> 
> diff --git a/arch/parisc/kernel/syscall.S b/arch/parisc/kernel/syscall.S
> index 69b6eeb..ae509d8 100644 (file)
> --- a/arch/parisc/kernel/syscall.S
> +++ b/arch/parisc/kernel/syscall.S
> @@ -636,7 +637,7 @@ END(sys_call_table64)
>                 All light-weight-syscall atomic operations
>                 will use this set of locks
>         */
> -       .section .data
> +       .section .data, "aw"
>         .align  PAGE_SIZE
>  ENTRY(lws_lock_start)
>         /* lws locks */
> 
> I'm now debugging the timer_interrupt problems, will keep the list posted.

That shouldn't be too hard...I can help a bit later tomorrow.

thanks,
grant

> 
> HTH
> 
> T-Bone
> --
> To unsubscribe from this list: send the line "unsubscribe linux-parisc" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Git head doesn't boot on A500

[Kyle McMartin]

On Sun, May 18, 2008 at 01:13:42AM -0600, Grant Grundler wrote:
> So I don't understand how __HEAD works...and all the other arches are
> using ".text.head" (not .head.text) which has corresponding "code" in
> their vmlinux.lds.  Could someone kindly point out how .head.text 
> works for parisc?
> 

*(.text) will include .head.text...

>

Re: Git head doesn't boot on A500

[Helge Deller]

[-- Attachment #1: Type: text/plain, Size: 1677 bytes --]

Grant Grundler wrote:
> On Sun, May 18, 2008 at 02:52:00AM +0200, Thibaut VARENE wrote:
>> On Sat, May 17, 2008 at 5:26 PM, Grant Grundler
>> <grundler@parisc-linux.org> wrote:
>>> On Sat, May 17, 2008 at 03:20:46PM +0200, Thibaut VARENE wrote:
>>>> Hi pa-ckers
>>>>
>>>> Just a quick failure report, I tested git head + jejb's zonelist patch
>>>> on my A500 (gcc-4.2) and it hpmc'd the box (afaict) very early during
>>>> bootup. I'll dig that more ASAP.
>> Kyle eventually spotted the offending patch:
>> http://git.kernel.org/?p=linux/kernel/git/kyle/parisc-2.6.git;a=commitdiff;h=bd3bb8c15b9a80dbddfb7905b237a4a11a4725b4;hp=492c2e476eac010962850006c49df326919b284c
>>
>> Out of this patch, the only two non-offending hunks are:
>> diff --git a/arch/parisc/kernel/head.S b/arch/parisc/kernel/head.S
>> index ec2482d..5680a2c 100644 (file)
>> --- a/arch/parisc/kernel/head.S
>> +++ b/arch/parisc/kernel/head.S
>> @@ -32,7 +32,7 @@ ENTRY(boot_args)
>>         .word 0 /* arg3 */
>>  END(boot_args)
>>
>> -       .section .text.head
>> +       __HEAD
>>         .align  4
>>         .import init_thread_union,data
>>         .import fault_vector_20,code    /* IVA parisc 2.0 32 bit */
> 
> So I don't understand how __HEAD works...and all the other arches are
> using ".text.head" (not .head.text) which has corresponding "code" in
> their vmlinux.lds.  Could someone kindly point out how .head.text 
> works for parisc?

Yeah, my fault (no idea why it worked for me???).

I see two options:
1) revert my patch (gets to linker warnings and have to build up a new 
patch, ideally adding a section for .text.head to vmlinux.lds.S),
2) apply the attached patch.

Opinions?

Helge

[-- Attachment #2: t66 --]
[-- Type: text/plain, Size: 337 bytes --]

diff --git a/arch/parisc/kernel/vmlinux.lds.S b/arch/parisc/kernel/vmlinux.lds.S
index 50b4a3a..2e516b8 100644
--- a/arch/parisc/kernel/vmlinux.lds.S
+++ b/arch/parisc/kernel/vmlinux.lds.S
@@ -50,6 +50,7 @@ SECTIONS
 
 	_text = .;		/* Text and read-only data */
 	.text ALIGN(16) : {
+		HEAD_TEXT
 		TEXT_TEXT
 		SCHED_TEXT
 		LOCK_TEXT

Re: Git head doesn't boot on A500

[Grant Grundler]

On Sun, May 18, 2008 at 06:19:52PM +0200, Helge Deller wrote:
...
>> using ".text.head" (not .head.text) which has corresponding "code" in
>> their vmlinux.lds.  Could someone kindly point out how .head.text works 
>> for parisc?
>
> Yeah, my fault (no idea why it worked for me???).

Kyle's explanation sounds right or I expect the sections wouldn't
have been included at all (or something like that) - and maybe it
really only matters for SMP - ie getting secondary CPUs running.

> I see two options:
> 1) revert my patch (gets to linker warnings and have to build up a new 
> patch, ideally adding a section for .text.head to vmlinux.lds.S),
> 2) apply the attached patch.
>
> Opinions?

(2) works for me. I'm booting with 64-bit on a500...at least it's
now finding all the disks and attempting to mount NFS root.

>
> Helge

> diff --git a/arch/parisc/kernel/vmlinux.lds.S b/arch/parisc/kernel/vmlinux.lds.S
> index 50b4a3a..2e516b8 100644
> --- a/arch/parisc/kernel/vmlinux.lds.S
> +++ b/arch/parisc/kernel/vmlinux.lds.S
> @@ -50,6 +50,7 @@ SECTIONS
>  
>  	_text = .;		/* Text and read-only data */
>  	.text ALIGN(16) : {
> +		HEAD_TEXT

grundler <537>fgrep HEAD_TEXT include/*/*.h
include/asm-generic/vmlinux.lds.h:#define HEAD_TEXT  *(.head.text)

Yeah, I like that better than depending on *(.text) to include it.

thanks,
grant

>  		TEXT_TEXT
>  		SCHED_TEXT
>  		LOCK_TEXT

Re: Git head doesn't boot on A500

[Helge Deller]

Grant Grundler wrote:
> On Sun, May 18, 2008 at 06:19:52PM +0200, Helge Deller wrote:
...
> (2) works for me. I'm booting with 64-bit on a500...at least it's
> now finding all the disks and attempting to mount NFS root.
> 
>> diff --git a/arch/parisc/kernel/vmlinux.lds.S b/arch/parisc/kernel/vmlinux.lds.S
>> index 50b4a3a..2e516b8 100644
>> --- a/arch/parisc/kernel/vmlinux.lds.S
>> +++ b/arch/parisc/kernel/vmlinux.lds.S
>> @@ -50,6 +50,7 @@ SECTIONS
>>  
>>  	_text = .;		/* Text and read-only data */
>>  	.text ALIGN(16) : {
>> +		HEAD_TEXT
> 
> grundler <537>fgrep HEAD_TEXT include/*/*.h
> include/asm-generic/vmlinux.lds.h:#define HEAD_TEXT  *(.head.text)
> 
> Yeah, I like that better than depending on *(.text) to include it.

So, this is the way to go ?
Should I clean up the patch and send to Linus, or Kyle, will you do that 
anyway with other patches?
I think this really should go into 2.6.26.

Helge