On 07.05.2020 09:45, Paul Durrant wrote:
>> -----Original Message-----
>> From: Jan Beulich <jbeul...@suse.com>
>> Sent: 07 May 2020 08:40
>> To: p...@xen.org
>> Cc: xen-devel@lists.xenproject.org; 'Paul Durrant' <pdurr...@amazon.com>;
>> 'Andrew Cooper'
>> <andrew.coop...@citrix.com>; 'George Dunlap' <george.dun...@citrix.com>;
>> 'Ian Jackson'
>> <ian.jack...@eu.citrix.com>; 'Julien Grall' <jul...@xen.org>; 'Stefano
>> Stabellini'
>> <sstabell...@kernel.org>; 'Wei Liu' <w...@xen.org>; 'Volodymyr Babchuk'
>> <volodymyr_babc...@epam.com>;
>> 'Roger Pau Monné' <roger....@citrix.com>
>> Subject: Re: [PATCH v2 1/5] xen/common: introduce a new framework for
>> save/restore of 'domain' context
>>
>> On 07.05.2020 09:34, Paul Durrant wrote:
>>>> From: Jan Beulich <jbeul...@suse.com>
>>>> Sent: 07 May 2020 08:22
>>>>
>>>> On 06.05.2020 18:44, Paul Durrant wrote:
>>>>>> From: Jan Beulich <jbeul...@suse.com>
>>>>>> Sent: 29 April 2020 12:02
>>>>>>
>>>>>> On 07.04.2020 19:38, Paul Durrant wrote:
>>>>>>> +int domain_load_begin(struct domain_context *c, unsigned int tc,
>>>>>>> + const char *name, const struct vcpu *v, size_t
>>>>>>> len,
>>>>>>> + bool exact)
>>>>>>> +{
>>>>>>> + if ( c->log )
>>>>>>> + gdprintk(XENLOG_INFO, "%pv load: %s (%lu)\n", v, name,
>>>>>>> + (unsigned long)len);
>>>>>>> +
>>>>>>> + BUG_ON(tc != c->desc.typecode);
>>>>>>> + BUG_ON(v->vcpu_id != c->desc.vcpu_id);
>>>>>>> +
>>>>>>> + if ( (exact && (len != c->desc.length)) ||
>>>>>>> + (len < c->desc.length) )
>>>>>>> + return -EINVAL;
>>>>>>
>>>>>> How about
>>>>>>
>>>>>> if ( exact ? len != c->desc.length
>>>>>> : len < c->desc.length )
>>>>>>
>>>>>
>>>>> Yes, that doesn't look too bad.
>>>>>
>>>>>> ? I'm also unsure about the < - don't you mean > instead? Too
>>>>>> little data would be compensated by zero padding, but too
>>>>>> much data can't be dealt with. But maybe I'm getting the sense
>>>>>> of len wrong ...
>>>>>
>>>>> I think the < is correct. The caller needs to have at least enough
>>>>> space to accommodate the context record.
>>>>
>>>> But this is load, not save - the caller supplies the data. If
>>>> there's less data than can be fit, it'll be zero-extended. If
>>>> there's too much data, the excess you don't know what to do
>>>> with (it might be okay to tolerate it being all zero).
>>>>
>>>
>>> But this is a callback. The outer load function iterates over
>>> the records calling the appropriate hander for each one. Those
>>> handlers then call this function saying how much data they
>>> expect and whether they want exactly that amount, or whether
>>> they can tolerate less (i.e. zero-extend). Hence
>>> len < c->desc.length is an error, because it means the
>>> descriptor contains more data than the hander knows how to
>>> handle.
>>
>> Oh, I see - "But maybe I'm getting the sense of len wrong ..."
>> then indeed applies.
>>
>> Any thoughts on tolerating the excess data being zero?
>>
>
> Well the point of the check here is to not tolerate excess data...
> Are you suggesting that it might be a reasonable idea?
Well - it looks to be the obvious counterpart to zero-extending.
I'm not going to assert though that I've thought through all
possible consequences...
Jan
