[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v4 1/2] memory-hotplug: add automatic onlining policy for the newly added memory

David Rientjes <rientjes@xxxxxxxxxx> writes:

> On Tue, 12 Jan 2016, Vitaly Kuznetsov wrote:
>> diff --git a/Documentation/memory-hotplug.txt 
>> b/Documentation/memory-hotplug.txt
>> index ce2cfcf..ceaf40c 100644
>> --- a/Documentation/memory-hotplug.txt
>> +++ b/Documentation/memory-hotplug.txt
>> @@ -254,12 +254,23 @@ If the memory block is online, you'll read "online".
>>  If the memory block is offline, you'll read "offline".
>> -5.2. How to online memory
>> +5.2. Memory onlining
> Idk why you're changing this title since you didn't change it in the table 
> of contents and it already pairs with "6.2. How to offline memory".
> This makes it seem like you're covering all memory onlining operations in 
> the kernel (including xen onlining) rather than just memory onlined by 
> root.  It doesn't cover the fact that xen onlining can be done without 
> automatic onlining, so I would leave this section's title as it is and 
> only cover aspects of memory onlining that users are triggering 
> themselves.

Ok, I changed the title to reflect the fact that a special action to
online memory is not always required any more but as the global policy
stays 'offline' by default for now let's keep the original title.


>> +    /* online pages if requested */
>> +    if (online)
>> +            online_pages(start >> PAGE_SHIFT, size >> PAGE_SHIFT,
>> +                         MMOP_ONLINE_KEEP);
>> +
>>      goto out;
>>  error:
> Well, shucks, what happens if online_pages() fails, such as if a memory 
> hot-add notifier returns an errno for MEMORY_GOING_ONLINE?  The memory was 
> added but not subsequently onlined, although auto onlining was set, so how 
> does userspace know the state it is in?

Bad ... we could have checked the return value but I don't see a proper
way to handling it here: if we managed to online some blocks we can't
revert back. We'll probably have to online pages block-by-block (e.g. by
utilizing memory_block_change_state()) handling possible failures.

>> @@ -1315,7 +1327,7 @@ int __ref add_memory(int nid, u64 start, u64 size)
>>      if (!res)
>>              return -EEXIST;
>> -    ret = add_memory_resource(nid, res);
>> +    ret = add_memory_resource(nid, res, memhp_auto_online);
>>      if (ret < 0)
>>              release_memory_resource(res);
>>      return ret;


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.