[PATCHv2 1/9] staging: zsmalloc: add gfp flags to zs_create_pool

[Seth Jennings]

On 01/24/2013 07:33 PM, Minchan Kim wrote:
> Hi Seth, frontswap guys
> 
> On Tue, Jan 8, 2013 at 5:24 AM, Seth Jennings
> <sjenning at linux.vnet.ibm.com> wrote:
>> zs_create_pool() currently takes a gfp flags argument
>> that is used when growing the memory pool.  However
>> it is not used in allocating the metadata for the pool
>> itself.  That is currently hardcoded to GFP_KERNEL.
>>
>> zswap calls zs_create_pool() at swapon time which is done
>> in atomic context, resulting in a "might sleep" warning.
> 
> I didn't review this all series, really sorry but totday I saw Nitin
> added Acked-by so I'm afraid Greg might get it under my radar. I'm not
> strong against but I would like know why we should call frontswap_init
> under swap_lock? Is there special reason?

The call stack is:

SYSCALL_DEFINE2(swapon.. <-- swapon_mutex taken here
enable_swap_info() <-- swap_lock taken here
frontswap_init()
__frontswap_init()
zswap_frontswap_init()
zs_create_pool()

It isn't entirely clear to me why frontswap_init() is called under
lock.  Then again, I'm not entirely sure what the swap_lock protects.
 There are no comments near the swap_lock definition to tell me.

I would guess that the intent is to block any writes to the swap
device until frontswap_init() has completed.

Dan care to weigh in?

Seth