diff mbox

[V4,2/3] arm64: support initrd outside kernel linear map

Message ID 20150908113113.GA20562@leverpostej
State New
Headers show

Commit Message

Mark Rutland Sept. 8, 2015, 11:31 a.m. UTC 
Hi Mark,

On Mon, Aug 17, 2015 at 06:01:06PM +0100, Mark Salter wrote:
> The use of mem= could leave part or all of the initrd outside of
> the kernel linear map. This will lead to an error when unpacking
> the initrd and a probable failure to boot. This patch catches that
> situation and relocates the initrd to be fully within the linear
> map.

With next-20150908, this patch results in a confusing message at boot when not
using an initrd:

Moving initrd from [4080000000-407fffffff] to [9fff49000-9fff48fff]

I think that can be solved by folding in the diff below.

Thanks,
Mark.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Comments

Mark Rutland Oct. 6, 2015, 5:11 p.m. UTC | #1 
On Tue, Sep 08, 2015 at 12:31:13PM +0100, Mark Rutland wrote:
> Hi Mark,
> 
> On Mon, Aug 17, 2015 at 06:01:06PM +0100, Mark Salter wrote:
> > The use of mem= could leave part or all of the initrd outside of
> > the kernel linear map. This will lead to an error when unpacking
> > the initrd and a probable failure to boot. This patch catches that
> > situation and relocates the initrd to be fully within the linear
> > map.
> 
> With next-20150908, this patch results in a confusing message at boot when not
> using an initrd:
> 
> Moving initrd from [4080000000-407fffffff] to [9fff49000-9fff48fff]
> 
> I think that can be solved by folding in the diff below.

Mark, it looks like this fell by the wayside.

Do you have any objection to this? I'll promote this to it's own patch
if not.

Mark.

> 
> Thanks,
> Mark.
> 
> diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
> index 6bab21f..2322479 100644
> --- a/arch/arm64/kernel/setup.c
> +++ b/arch/arm64/kernel/setup.c
> @@ -364,6 +364,8 @@ static void __init relocate_initrd(void)
>                 to_free = ram_end - orig_start;
>  
>         size = orig_end - orig_start;
> +       if (!size)
> +               return;
>  
>         /* initrd needs to be relocated completely inside linear mapping */
>         new_start = memblock_find_in_range(0, PFN_PHYS(max_pfn),
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Mark Salter Oct. 6, 2015, 5:16 p.m. UTC | #2 
On Tue, 2015-10-06 at 18:11 +0100, Mark Rutland wrote:
> On Tue, Sep 08, 2015 at 12:31:13PM +0100, Mark Rutland wrote:
> > Hi Mark,
> > 
> > On Mon, Aug 17, 2015 at 06:01:06PM +0100, Mark Salter wrote:
> > > The use of mem= could leave part or all of the initrd outside of
> > > the kernel linear map. This will lead to an error when unpacking
> > > the initrd and a probable failure to boot. This patch catches that
> > > situation and relocates the initrd to be fully within the linear
> > > map.
> > 
> > With next-20150908, this patch results in a confusing message at boot when not
> > using an initrd:
> > 
> > Moving initrd from [4080000000-407fffffff] to [9fff49000-9fff48fff]
> > 
> > I think that can be solved by folding in the diff below.
> 
> Mark, it looks like this fell by the wayside.
> 
> Do you have any objection to this? I'll promote this to it's own patch
> if not.
> 
> Mark.
> 
> > 
> > Thanks,
> > Mark.
> > 
> > diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
> > index 6bab21f..2322479 100644
> > --- a/arch/arm64/kernel/setup.c
> > +++ b/arch/arm64/kernel/setup.c
> > @@ -364,6 +364,8 @@ static void __init relocate_initrd(void)
> >                 to_free = ram_end - orig_start;
> >  
> >         size = orig_end - orig_start;
> > +       if (!size)
> > +               return;
> >  
> >         /* initrd needs to be relocated completely inside linear mapping */
> >         new_start = memblock_find_in_range(0, PFN_PHYS(max_pfn),

Sorry, no. That looks perfectly good to me.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Christoffer Dall Oct. 8, 2015, 8:49 a.m. UTC | #3 
On Tue, Oct 06, 2015 at 01:16:52PM -0400, Mark Salter wrote:
> On Tue, 2015-10-06 at 18:11 +0100, Mark Rutland wrote:
> > On Tue, Sep 08, 2015 at 12:31:13PM +0100, Mark Rutland wrote:
> > > Hi Mark,
> > > 
> > > On Mon, Aug 17, 2015 at 06:01:06PM +0100, Mark Salter wrote:
> > > > The use of mem= could leave part or all of the initrd outside of
> > > > the kernel linear map. This will lead to an error when unpacking
> > > > the initrd and a probable failure to boot. This patch catches that
> > > > situation and relocates the initrd to be fully within the linear
> > > > map.
> > > 
> > > With next-20150908, this patch results in a confusing message at boot when not
> > > using an initrd:
> > > 
> > > Moving initrd from [4080000000-407fffffff] to [9fff49000-9fff48fff]
> > > 
> > > I think that can be solved by folding in the diff below.
> > 
> > Mark, it looks like this fell by the wayside.
> > 
> > Do you have any objection to this? I'll promote this to it's own patch
> > if not.
> > 
> > Mark.
> > 
> > > 
> > > Thanks,
> > > Mark.
> > > 
> > > diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
> > > index 6bab21f..2322479 100644
> > > --- a/arch/arm64/kernel/setup.c
> > > +++ b/arch/arm64/kernel/setup.c
> > > @@ -364,6 +364,8 @@ static void __init relocate_initrd(void)
> > >                 to_free = ram_end - orig_start;
> > >  
> > >         size = orig_end - orig_start;
> > > +       if (!size)
> > > +               return;
> > >  
> > >         /* initrd needs to be relocated completely inside linear mapping */
> > >         new_start = memblock_find_in_range(0, PFN_PHYS(max_pfn),
> 
> Sorry, no. That looks perfectly good to me.
> 
FYI: I applied these patches to 4.0 (only a trivial conflict on the x86
side) and this fixed an issue for me booting systems with mem=X,
reducing the amount of physical memory available on a system, which
would otherwise cause the system to just silently halt during boot.

Note that this seems to fix even more than it promises, because one of
those systems does not use an initrd, but I'm thinking maybe this fixes
issues with the DT as well?

In any case, I think this may be a good candidate for cc'ing to stable?

Thanks,
-Christoffer
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
yalin wang Oct. 8, 2015, 9:18 a.m. UTC | #4 
> On Oct 8, 2015, at 16:49, Christoffer Dall <christoffer.dall@linaro.org> wrote:
> 
> On Tue, Oct 06, 2015 at 01:16:52PM -0400, Mark Salter wrote:
>> On Tue, 2015-10-06 at 18:11 +0100, Mark Rutland wrote:
>>> On Tue, Sep 08, 2015 at 12:31:13PM +0100, Mark Rutland wrote:
>>>> Hi Mark,
>>>> 
>>>> On Mon, Aug 17, 2015 at 06:01:06PM +0100, Mark Salter wrote:
>>>>> The use of mem= could leave part or all of the initrd outside of
>>>>> the kernel linear map. This will lead to an error when unpacking
>>>>> the initrd and a probable failure to boot. This patch catches that
>>>>> situation and relocates the initrd to be fully within the linear
>>>>> map.
>>>> 
>>>> With next-20150908, this patch results in a confusing message at boot when not
>>>> using an initrd:
>>>> 
>>>> Moving initrd from [4080000000-407fffffff] to [9fff49000-9fff48fff]
>>>> 
>>>> I think that can be solved by folding in the diff below.
>>> 
>>> Mark, it looks like this fell by the wayside.
>>> 
>>> Do you have any objection to this? I'll promote this to it's own patch
>>> if not.
>>> 
>>> Mark.
>>> 
>>>> 
>>>> Thanks,
>>>> Mark.
>>>> 
>>>> diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
>>>> index 6bab21f..2322479 100644
>>>> --- a/arch/arm64/kernel/setup.c
>>>> +++ b/arch/arm64/kernel/setup.c
>>>> @@ -364,6 +364,8 @@ static void __init relocate_initrd(void)
>>>>                to_free = ram_end - orig_start;
>>>> 
>>>>        size = orig_end - orig_start;
>>>> +       if (!size)
>>>> +               return;
>>>> 
>>>>        /* initrd needs to be relocated completely inside linear mapping */
>>>>        new_start = memblock_find_in_range(0, PFN_PHYS(max_pfn),
>> 
>> Sorry, no. That looks perfectly good to me.
>> 
is it also possible to implement it on ARM platforms?
ARM64 platform don’t have HIGH_MEM zone .
but ARM platform have .
i remember boot loader must put init rd  into low memory region,
so if some boot loader put init rd into HIGH men zone
we can also relocate it to low men region ?
then boot loader don’t need care about this ,
and since vmalloc= boot option will change HIGH mem region size,
if we can relocate init rd , boot loader don’t need care about init rd load address,
when change vmalloc= boot options .


Thanks









--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Russell King - ARM Linux Oct. 8, 2015, 9:40 a.m. UTC | #5 
On Thu, Oct 08, 2015 at 05:18:14PM +0800, yalin wang wrote:
> is it also possible to implement it on ARM platforms?
> ARM64 platform don’t have HIGH_MEM zone .
> but ARM platform have .
> i remember boot loader must put init rd  into low memory region,
> so if some boot loader put init rd into HIGH men zone
> we can also relocate it to low men region ?
> then boot loader don’t need care about this ,
> and since vmalloc= boot option will change HIGH mem region size,
> if we can relocate init rd , boot loader don’t need care about init rd load address,
> when change vmalloc= boot options .

I'd be more inclined to say yes if the kernel wasn't buggering around
passing virtual addresses (initrd_start) of the initrd image around,
but instead used a physical address.  initrd_start must be a lowmem
address.
diff mbox

Patch

diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
index 6bab21f..2322479 100644
--- a/arch/arm64/kernel/setup.c
+++ b/arch/arm64/kernel/setup.c
@@ -364,6 +364,8 @@  static void __init relocate_initrd(void)
                to_free = ram_end - orig_start;
 
        size = orig_end - orig_start;
+       if (!size)
+               return;
 
        /* initrd needs to be relocated completely inside linear mapping */
        new_start = memblock_find_in_range(0, PFN_PHYS(max_pfn),