[Devel] Re: [PATCH 7/9] cr: checkpoint saved_auxv as u64s

[Oren Laadan]

Serge E. Hallyn wrote:
> Quoting Oren Laadan (orenl at cs.columbia.edu):
>>
>> serue at us.ibm.com wrote:
>>> From: Serge E. Hallyn <serue at us.ibm.com>
>>>
>>> unsigned longs are not a good value to checkpoint between
>>> x86-32 and x86-64 32-bit tasks :)
>>>
>>> Signed-off-by: Serge E. Hallyn <serue at us.ibm.com>
>>> ---
>>>  checkpoint/checkpoint.c        |    5 +--
>>>  checkpoint/memory.c            |   53 +++++++++++++++++++++++++++++++++++++--
>>>  checkpoint/restart.c           |    6 ++--
>>>  include/linux/checkpoint_hdr.h |    2 +-
>>>  4 files changed, 56 insertions(+), 10 deletions(-)
>> [...]
>>
>> Sketch for a sanity check:
>>
>>> +static int ckpt_read_auxv(struct ckpt_ctx *ctx, struct mm_struct *mm)
>>> +{
>>> +	int i, ret;
>>> +	u64 *buf = kmalloc(CKPT_AT_SZ, GFP_KERNEL);
>>> +
>>> +	if (!buf)
>>> +		return -ENOMEM;
>>> +	ret = _ckpt_read_buffer(ctx, buf, CKPT_AT_SZ);
>>> +	if (ret < 0) {
>>> +		kfree(buf);
>>> +		return ret;
>>> +	}
>>> +
>> 	ret = -E2BIG;
>>> +	for (i=0; i<AT_VECTOR_SIZE; i++)
>>> +		if (buf[i] > (u64) ULONG_MAX)
>> 			goto out;
>>
>> 	ret = -EINVAL;
>> 	for (i=0; i<AT_VECTOR_SIZE; i++)
>> 		if (mm->saved_auxv[i] == AT_NULL)
>> 			ret = 0;
>> 	if (ret < 0)
>> 		goto out;
> 
> Yup, that would work.
> 
> Alternatively, do you think it would be safe to just
> always set the last entry to AT_NULL?

A bit hacky, but should work.
... so I'll do just that :)

Oren.

> 
>>> +
>>> +	for (i=0; i<AT_VECTOR_SIZE; i++)
>>> +		mm->saved_auxv[i] = buf[i];
>>> +
>>  out:
>>> +	kfree(buf);
>>> +	return 0;
>> 	       ret;
>>
>> [...]
>>
>> Oren.
> 
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers