Re: Linux 2.6.24-rc7 Build-Failure at __you_cannot_kmalloc_that_much

[Balbir Singh <balbir@linux.vnet.ibm.com>]

Kamalesh Babulal wrote:
> Andrew Morton wrote:
>> On Mon, 07 Jan 2008 16:06:20 +0530 Kamalesh Babulal <kamalesh@linux.vnet.ibm.com> wrote:
>>
>>> The defconfig make fails on x86_64 (AMD box) with following error
>>>
>>>   CHK     include/linux/utsrelease.h
>>>   CALL    scripts/checksyscalls.sh
>>>   CHK     include/linux/compile.h
>>>   GEN     .version
>>>   CHK     include/linux/compile.h
>>>   UPD     include/linux/compile.h
>>>   CC      init/version.o
>>>   LD      init/built-in.o
>>>   LD      .tmp_vmlinux1
>>> drivers/built-in.o(.init.text+0x8d76): In function `dmi_id_init':
>>> : undefined reference to `__you_cannot_kmalloc_that_much'
>>> make: *** [.tmp_vmlinux1] Error 1
>>>
>>>
>>> # gcc --version
>>> gcc (GCC) 3.2.3 20030502 (Red Hat Linux 3.2.3-59)
>>>
>>> This was reported by Adrian Bunk http://lkml.org/lkml/2007/12/1/39
>> That's odd.  afacit the only kmalloc in dmi_id_init() is
>>
>>         dmi_dev = kzalloc(sizeof(*dmi_dev), GFP_KERNEL);
>>
>> and even gcc-3.2.3 should be able to get that right.
>>
>> Could you please a) verify that simply removing that line fixes the build
>> error and then b) try to find some way of fixing it?
>>
>> Try replacing `sizeof(*dmi_dev)' with `sizeof(struct dmi_device_attribute)'
>> and any other tricks you can think of to try to make the compiler process
>> the code differently.
>>
> 
> removing the line fixes the issue, but changing the sizeof(*dmi_dev) to
> sizeof(struct device) is not helping.
> 

Hi, Andrew,

We tried the following, generated stabs information. The size of struct
device is 560 bytes. We found that dead code was not being eliminated
(__you_cannot_kmalloc_that_much), even though no one called that
function. I suspect builtin_constant_p() and dead code elimination as
the root causes of this error.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL