Hi, Herbert:

I've figured out a new patch for this issue reported by me previously, 
the basic idea is adding a cryptd_flush_queue function fixing it by 
being called from softirq to flush all previous queued elements before 
processing a new one, and it works very well so far per my test,  would 
you please review it?

the best,
thank you

On 08/14/2014 04:47 PM, Herbert Xu wrote:

On Sun, Aug 03, 2014 at 05:57:06PM +0800, Ming Liu wrote:

quoted

Please review this attached patch instead, the original one has a
problem causing the kernel crash.

Thanks for the patch.  I think it would better to enforce ordering
for all operations rather than treat encryptions separately from
decryptions.  We could conceivably have more complex operations made
up from both encryptions and decryptions that could then get
out-of-order.

It would also make the code simpler.

Cheers,