Re: [PATCH 1/4] printk: Allocate kernel log buffer earlier

[Mike Travis <travis@sgi.com>]

Ingo Molnar wrote:
> * Mike Travis <travis@sgi.com> wrote:
> 
>> On larger systems, because of the numerous ACPI, Bootmem and EFI
>> messages, the static log buffer overflows before the larger one
>> specified by the log_buf_len param is allocated.  Minimize the
>> potential for overflow by allocating the new log buffer as soon
>> as possible.
>>
>> We do this by changing the log_buf_len from an early_param to a
>> _setup param.  But _setup params are processed before the
>> alloc_bootmem is available, so this function will now just save
>> the requested log buf len.  The real work routine (setup_log_buf)
>> is called after bootmem is available.
>>
>> Signed-off-by: Mike Travis <travis@sgi.com>
>> Reviewed-by: Jack Steiner <steiner@sgi.com>
>> Reviewed-by: Robin Holt <holt@sgi.com>
>> ---
>>  arch/x86/kernel/setup.c |    5 +++
>>  include/linux/printk.h  |    4 ++
>>  init/main.c             |    1 
>>  kernel/printk.c         |   75 ++++++++++++++++++++++++++++--------------------
>>  4 files changed, 54 insertions(+), 31 deletions(-)
> 
> Well, the modern allocation method is memblock - available on all major 
> architectures.
> 
> You could avoid all this ugly workaround of bootmem limitations by moving the 
> allocation to memblock_alloc() and desupporting the log_buf_len= boot parameter
> on non-memblock architectures.

Is it really that ugly?  I thought in some ways it cleaned it up.

I'm also hesitant to change code for other arch's when I can't test them.  This
approach seemed to be the safest.

> kernel log buffer size can be configured via the .config so they will not be left 
> without larger buffers.

We have asked about this, but distros are reluctant to increase memory usage
for their entire installed base.  I think we're lucky they bumped it up to 256k
from the default 128k.

> 
> Doing this should also have the advantage of getting all the early x86 messages into 
> the larger buffer already, reducing the pressure to apply some of the other patches 
> in your series.

There are only two and both remove only redundant information.

> 
> Thanks,
> 
> 	Ingo