git-block patches broken on 2.6.23-mm1 ia64 SN

git-block patches broken on 2.6.23-mm1 ia64 SN

[Paul Jackson]

[Try again, with a Subject this time ...]

The git-block patches in 2.6.23-mm1 freeze my ia64 SN Altix hard on boot.
Not good (tm).

Something broke between the git-block patch of Sept 26 in 2.6.23-rc8-mm2
and the git-block patches of Oct 10-11 in 2.6.23-mm1 on my ia64 SN Altix
system using sn2_defconfig.

Bisecting 2.6.23-mm1 narrows the problem down to the git-block patches:

    git-block.patch
    git-block-fixup-1.patch
    git-block-fixup.patch
    git-block-fixup-fix.patch
    git-block-borkages.patch
    git-block-s390-fix.patch

I can boot 2.6.23-rc8-mm2 fine, but I freeze early in boot
on 2.6.23-mm1, after the following prints on the console:

    McKinley Errata 9 workaround not needed; disabling it
    SLUB: Genslabs=26, HWalign=128, Order=0-2, MinObjects=8, CPUs=8, Nodes=1024
    Dentry cache hash table entries: 1048576 (order: 9, 8388608 bytes)
    Inode-cache hash table entries: 524288 (order: 8, 4194304 bytes)
    Mount-cache hash table entries: 1024
    ACPI: Core revision 20070126
    Boot processor id 0x0/0x0
    Brought up 8 CPUs
    Total of 8 processors activated (15564.80 BogoMIPS).

The next output that I -would- have expected, based on successful boots
without these patches, but never get, is:

    net_namespace: 120 bytes
    DMI not present or invalid.
    xor: measuring software checksum speed
       ia64      :  2692.000 MB/sec
    xor: using function: ia64 (2692.000 MB/sec)
    NET: Registered protocol family 16
    ACPI  DSDT OEM Rev 0x20101


-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401

Re: git-block patches broken on 2.6.23-mm1 ia64 SN

[Jens Axboe]

On Sat, Oct 13 2007, Paul Jackson wrote:
> [Try again, with a Subject this time ...]
> 
> The git-block patches in 2.6.23-mm1 freeze my ia64 SN Altix hard on boot.
> Not good (tm).
> 
> Something broke between the git-block patch of Sept 26 in 2.6.23-rc8-mm2
> and the git-block patches of Oct 10-11 in 2.6.23-mm1 on my ia64 SN Altix
> system using sn2_defconfig.
> 
> Bisecting 2.6.23-mm1 narrows the problem down to the git-block patches:
> 
>     git-block.patch
>     git-block-fixup-1.patch
>     git-block-fixup.patch
>     git-block-fixup-fix.patch
>     git-block-borkages.patch
>     git-block-s390-fix.patch
> 
> I can boot 2.6.23-rc8-mm2 fine, but I freeze early in boot
> on 2.6.23-mm1, after the following prints on the console:
> 
>     McKinley Errata 9 workaround not needed; disabling it
>     SLUB: Genslabs=26, HWalign=128, Order=0-2, MinObjects=8, CPUs=8, Nodes=1024
>     Dentry cache hash table entries: 1048576 (order: 9, 8388608 bytes)
>     Inode-cache hash table entries: 524288 (order: 8, 4194304 bytes)
>     Mount-cache hash table entries: 1024
>     ACPI: Core revision 20070126
>     Boot processor id 0x0/0x0
>     Brought up 8 CPUs
>     Total of 8 processors activated (15564.80 BogoMIPS).
> 
> The next output that I -would- have expected, based on successful boots
> without these patches, but never get, is:
> 
>     net_namespace: 120 bytes
>     DMI not present or invalid.
>     xor: measuring software checksum speed
>        ia64      :  2692.000 MB/sec
>     xor: using function: ia64 (2692.000 MB/sec)
>     NET: Registered protocol family 16
>     ACPI  DSDT OEM Rev 0x20101

Can you try with init call debugging, perhaps that'll show where it
stalls? Sounds a bit odd that the block changes would cause a failure so
early in the boot, it hasn't done anything yet. But it could be some
error in the init functions.

-- 
Jens Axboe

Re: git-block patches broken on 2.6.23-mm1 ia64 SN

[Paul Jackson]

Jens wrote:
> Can you try with init call debugging,

Sure - if I knew what that was or how to do it ;).

If you want to guide me, I'll give it try.

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401

Re: git-block patches broken on 2.6.23-mm1 ia64 SN

[Andrew Morton]

On Sat, 13 Oct 2007 00:20:03 -0700 Paul Jackson <pj@sgi.com> wrote:

> Jens wrote:
> > Can you try with init call debugging,
> 
> Sure - if I knew what that was or how to do it ;).
> 
> If you want to guide me, I'll give it try.
> 

Add initcall_debug to the boot command line.

Re: git-block patches broken on 2.6.23-mm1 ia64 SN

[Jens Axboe]

On Sat, Oct 13 2007, Paul Jackson wrote:
> Jens wrote:
> > Can you try with init call debugging,
> 
> Sure - if I knew what that was or how to do it ;).
> 
> If you want to guide me, I'll give it try.

Just pass 'initcall_debug' as a boot parameter.

-- 
Jens Axboe

Re: git-block patches broken on 2.6.23-mm1 ia64 SN

[Paul Jackson]

Booting with initcall_debug doesn't change the output at all.  If I
boot a working kernel with initcall_debug, the added output starts
showing up right after the point that my broken kernel is freezing.

    Mount-cache hash table entries: 1024
    ACPI: Core revision 20070126
    Boot processor id 0x0/0x0
    Brought up 8 CPUs
    Total of 8 processors activated (15564.80 BogoMIPS).
<< Broken kernel freezes here >>
<< Good kernel continues with ...>>
    Calling initcall 0xa0000001009d6310: net_ns_init+0x0/0x3e0()
    net_namespace: 120 bytes
    initcall 0xa0000001009d6310: net_ns_init+0x0/0x3e0() returned 0.
    initcall 0xa0000001009d6310 ran for 0 msecs: net_ns_init+0x0/0x3e0()
    Calling initcall 0xa0000001009d6630: run_dmi_scan+0x0/0x40()

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401

Re: git-block patches broken on 2.6.23-mm1 ia64 SN

[Paul Jackson]

Dang it -- I suspect I borked my bisection.

Looks like the problem is in the few patches
prior to git-block.

Sorry, Jens.

I'll start all over again, on some other thread,
once I have something useful (to someone, likely
not you ;) to say.

Sorry.

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401