bootmem code - reboots after 'uncompressing linux' on old computers

bootmem code - reboots after 'uncompressing linux' on old computers

[Piotr Gluszenia Slawinski]

Hello.
while bisecting old pcmcia bug i've noticed kernels ~2.6.36
and up do not boot on 586 machines with small amounts of ram (16M)

suprisingly 3.5 kernel booted fine.

i've bisected the problem and found fix :

solidstate linux # git bisect good
4e1c2b284461fd8aa8d7b295a1e911fc4390755b is the first bad commit
commit 4e1c2b284461fd8aa8d7b295a1e911fc4390755b
Author: David Miller <davem@davemloft.net>
Date:   Wed Apr 25 16:10:50 2012 -0400

     mm: nobootmem: Correct alloc_bootmem semantics.

     The comments above __alloc_bootmem_node() claim that the code will
     first try the allocation using 'goal' and if that fails it will
     try again but with the 'goal' requirement dropped.

     Unfortunately, this is not what the code does, so fix it to do so.

     This is important for nobootmem conversions to architectures such
     as sparc where MAX_DMA_ADDRESS is infinity.

     On such architectures all of the allocations done by generic spots,
     such as the sparse-vmemmap implementation, will pass in:

         __pa(MAX_DMA_ADDRESS)

     as the goal, and with the limit given as "-1" this will always fail
     unless we add the appropriate fallback logic here.

     Signed-off-by: David S. Miller <davem@davemloft.net>
     Acked-by: Yinghai Lu <yinghai@kernel.org>
     Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

:040000 040000 5c42bbd13a81426248901205b051968bab14e6ff 
3e7fad8afb42036c6bbb1fcf5fcf12c87bbba9e2 M      mm


kernels before 3.6.39 do not have nobootmem.c , but they still have same 
bug!

this patch fixes the problem for them (useful for bisecting, etc, imho 
should be merged into 2.6.35-stable branch) :


diff --git a/mm/bootmem.c b/mm/bootmem.c
index 13b0caa..b0ccada 100644
--- a/mm/bootmem.c
+++ b/mm/bootmem.c
@@ -848,6 +848,7 @@ void * __init __alloc_bootmem_node(pg_data_t *pgdat, 
unsigned long size,
                 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);

  #ifdef CONFIG_NO_BOOTMEM
+again:
         ptr = __alloc_memory_core_early(pgdat->node_id, size, align,
                                          goal, -1ULL);
         if (ptr)
@@ -859,6 +860,10 @@ void * __init __alloc_bootmem_node(pg_data_t *pgdat, 
unsigned long size,
         ptr = ___alloc_bootmem_node(pgdat->bdata, size, align, goal, 0);
  #endif

+       if (!ptr && goal) {
+               goal = 0;
+               goto again;
+       }
         return ptr;
  }


-- 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: bootmem code - reboots after 'uncompressing linux' on old computers

[Jan Ceuleers]

On 08/03/2012 03:37 AM, Piotr Gluszenia Slawinski wrote:
> Hello.
> while bisecting old pcmcia bug i've noticed kernels ~2.6.36
> and up do not boot on 586 machines with small amounts of ram (16M)

Copying David Miller.

Not trimming the rest of the message for David's benefit; no further comments from me.

> suprisingly 3.5 kernel booted fine.
> 
> i've bisected the problem and found fix :
> 
> solidstate linux # git bisect good
> 4e1c2b284461fd8aa8d7b295a1e911fc4390755b is the first bad commit
> commit 4e1c2b284461fd8aa8d7b295a1e911fc4390755b
> Author: David Miller <davem@davemloft.net>
> Date:   Wed Apr 25 16:10:50 2012 -0400
> 
>     mm: nobootmem: Correct alloc_bootmem semantics.
> 
>     The comments above __alloc_bootmem_node() claim that the code will
>     first try the allocation using 'goal' and if that fails it will
>     try again but with the 'goal' requirement dropped.
> 
>     Unfortunately, this is not what the code does, so fix it to do so.
> 
>     This is important for nobootmem conversions to architectures such
>     as sparc where MAX_DMA_ADDRESS is infinity.
> 
>     On such architectures all of the allocations done by generic spots,
>     such as the sparse-vmemmap implementation, will pass in:
> 
>         __pa(MAX_DMA_ADDRESS)
> 
>     as the goal, and with the limit given as "-1" this will always fail
>     unless we add the appropriate fallback logic here.
> 
>     Signed-off-by: David S. Miller <davem@davemloft.net>
>     Acked-by: Yinghai Lu <yinghai@kernel.org>
>     Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
> 
> :040000 040000 5c42bbd13a81426248901205b051968bab14e6ff 3e7fad8afb42036c6bbb1fcf5fcf12c87bbba9e2 M      mm
> 
> 
> kernels before 3.6.39 do not have nobootmem.c , but they still have same bug!
> 
> this patch fixes the problem for them (useful for bisecting, etc, imho should be merged into 2.6.35-stable branch) :
> 
> 
> diff --git a/mm/bootmem.c b/mm/bootmem.c
> index 13b0caa..b0ccada 100644
> --- a/mm/bootmem.c
> +++ b/mm/bootmem.c
> @@ -848,6 +848,7 @@ void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
>                 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
> 
>  #ifdef CONFIG_NO_BOOTMEM
> +again:
>         ptr = __alloc_memory_core_early(pgdat->node_id, size, align,
>                                          goal, -1ULL);
>         if (ptr)
> @@ -859,6 +860,10 @@ void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
>         ptr = ___alloc_bootmem_node(pgdat->bdata, size, align, goal, 0);
>  #endif
> 
> +       if (!ptr && goal) {
> +               goal = 0;
> +               goto again;
> +       }
>         return ptr;
>  }
> 
> 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>