kexec fail

kexec fail

[Yinghai Lu]

just found that when kexec 2.6.24-rc7 from RHEL 5.1 kernel got

Your BIOS doesn't leave a aperture memory hole
Please enable the IOMMU option in the BIOS setup
This costs you 64 MB of RAM
Mapping aperture over 65536 KB of RAM @ 20000000
Bad page state in process 'swapper'
page:ffffe20004000018 flags:0x0000000000000000
mapping:0000000000000000 mapcount:1 count:0
Trying to fix it up, but a reboot is needed
Backtrace:
Pid: 0, comm: swapper Not tainted 2.6.24-rc7-smp #1

Call Trace:
 [<ffffffff80263fa3>] bad_page+0x63/0x8d
 [<ffffffff802640ed>] __free_pages_ok+0x7c/0x2a5
 [<ffffffff80ba7161>] free_all_bootmem_core+0xd0/0x198
 [<ffffffff80ba3a22>] numa_free_all_bootmem+0x3b/0x76
 [<ffffffff80ba30a2>] mem_init+0x4e/0x165
 [<ffffffff80b959dc>] start_kernel+0x236/0x2c2
 [<ffffffff80b95123>] _sinittext+0x123/0x12a

Bad page state in process 'swapper'
page:ffffe20004000050 flags:0x0000000000000000
mapping:0000000000000000 mapcount:1 count:0
Trying to fix it up, but a reboot is needed
Backtrace:
Pid: 0, comm: swapper Tainted: G    B   2.6.24-rc7-smp #1

Call Trace:
 [<ffffffff80263fa3>] bad_page+0x63/0x8d
 [<ffffffff802640ed>] __free_pages_ok+0x7c/0x2a5
 [<ffffffff80ba7161>] free_all_bootmem_core+0xd0/0x198
 [<ffffffff80ba3a22>] numa_free_all_bootmem+0x3b/0x76
 [<ffffffff80ba30a2>] mem_init+0x4e/0x165
 [<ffffffff80b959dc>] start_kernel+0x236/0x2c2
 [<ffffffff80b95123>] _sinittext+0x123/0x12a
...

somewhere from 64M

start_kernel
        ==> mem_init
                ==> numa_free_all_bootmem
                        ==> free_all_bootmem_node (NODE_DATA(i))
                                ==> free_all_bootmem_core ()
                                        ==> __free_pages_boot_mem
                                                ==> __free_pages(page, oder)
                                                        ==>
__free_pages_ok(page, order)
                                                                ==>
free_pages_check

 ==> bad_pages...


try to kexec 2.6.23 from RHEL 5.1, will get
Your BIOS doesn't leave a aperture memory hole
Please enable the IOMMU option in the BIOS setup
This costs you 64 MB of RAM ======> reboot

but 2.6.24-rc7 kexec 2.6.24-rc7 is ok.

YH

Re: kexec fail

[Christoph Lameter]

On Mon, 7 Jan 2008, Yinghai Lu wrote:

> try to kexec 2.6.23 from RHEL 5.1, will get
> Your BIOS doesn't leave a aperture memory hole
> Please enable the IOMMU option in the BIOS setup
> This costs you 64 MB of RAM ======> reboot
> 
> but 2.6.24-rc7 kexec 2.6.24-rc7 is ok.

BUG in 2.6.23 kexec?

Re: kexec fail

[Yinghai Lu]

On Jan 7, 2008 9:50 PM, Christoph Lameter <clameter@sgi.com> wrote:
> On Mon, 7 Jan 2008, Yinghai Lu wrote:
>
> > try to kexec 2.6.23 from RHEL 5.1, will get
> > Your BIOS doesn't leave a aperture memory hole
> > Please enable the IOMMU option in the BIOS setup
> > This costs you 64 MB of RAM ======> reboot
> >
> > but 2.6.24-rc7 kexec 2.6.24-rc7 is ok.
>
> BUG in 2.6.23 kexec?
>
>

yeah, that is more worse. it will cause reboot.

YH

Re: kexec fail

[Yinghai Lu]

On Jan 8, 2008 12:15 AM, Yinghai Lu <yhlu.kernel@gmail.com> wrote:
>
> On Jan 7, 2008 9:50 PM, Christoph Lameter <clameter@sgi.com> wrote:
> > On Mon, 7 Jan 2008, Yinghai Lu wrote:
> >
> > > try to kexec 2.6.23 from RHEL 5.1, will get
> > > Your BIOS doesn't leave a aperture memory hole
> > > Please enable the IOMMU option in the BIOS setup
> > > This costs you 64 MB of RAM ======> reboot
> > >
> > > but 2.6.24-rc7 kexec 2.6.24-rc7 is ok.
> >
> > BUG in 2.6.23 kexec?
> >
> >
>
> yeah, that is more worse. it will cause reboot.
>

change mem model from CONFIG_SPARSEMEM_MANUL to
CONFIG_DISCONTIGMEM_MANUAL it will work.

YH

Re: kexec fail

[Yinghai Lu]

On Jan 8, 2008 2:36 PM, Yinghai Lu <yhlu.kernel@gmail.com> wrote:
>
> On Jan 8, 2008 12:15 AM, Yinghai Lu <yhlu.kernel@gmail.com> wrote:
> >
> > On Jan 7, 2008 9:50 PM, Christoph Lameter <clameter@sgi.com> wrote:
> > > On Mon, 7 Jan 2008, Yinghai Lu wrote:
> > >
> > > > try to kexec 2.6.23 from RHEL 5.1, will get
> > > > Your BIOS doesn't leave a aperture memory hole
> > > > Please enable the IOMMU option in the BIOS setup
> > > > This costs you 64 MB of RAM ======> reboot
> > > >
> > > > but 2.6.24-rc7 kexec 2.6.24-rc7 is ok.
> > >
> > > BUG in 2.6.23 kexec?
> > >
> > >
> >
> > yeah, that is more worse. it will cause reboot.
> >
>
> change mem model from CONFIG_SPARSEMEM_MANUL to
> CONFIG_DISCONTIGMEM_MANUAL it will work.

find the cause. because two kernel has different addr for GART aper.
and second kernel will put vmemmap in the GART set by first kernel.

http://lkml.org/lkml/2008/1/10/517

YH