Re: [PATCH v12 5/5] x86/sgx: Enable automatic SVN updates for SGX enclaves

["Huang, Kai" <kai.huang@intel.com>]

> > 
> > > 
> > > +/* Mutex to ensure no concurrent EPC accesses during EUPDATESVN */
> > > +static DEFINE_MUTEX(sgx_svn_lock);
> > > +
> > >  int sgx_inc_usage_count(void)
> > >  {
> > > +	guard(mutex)(&sgx_svn_lock);
> > > +
> > > +	if (sgx_usage_count++ == 0)
> > > +		return sgx_update_svn();
> > > +
> > 
> > Hmm.. sorry for not noticing this before.. But I think we might have a
> > problem here since the sgx_usage_count is increased regardless of the
> > result of sgx_update_svn().
> > 
> > If sgx_update_svn() fails, it makes sgx_inc_usage_count() return error
> > too, so sgx_{vepc_}open() will fail and return immediately w/o calling
> > sgx_dec_usage_count().
> > 
> > But the sgx_usage_count has been increased.
> > 
> > AFAICT when sgx_{vepc_}_open() fails, the sgx_{vepc_}release() is not
> > called, so sgx_dec_usage_count() is never called and sgx_usage_count
> > remains increased.
> > 
> > So when sgx_{vepc_}open() calls sgx_inc_usage_count() again, it will skip
> > calling sgx_update_svn(), and allow enclave/vEPC to be created
> > successfully, which just defeats the purpose.
> > 
> > So if I am not missing anything, I think we should only increase the count
> > when sgx_update_svn() returns success?
> 
> Yes, you are right, thanks for catching this! In past the atomic version of
> this patch did exactly, but after I went into this simplified version of counting,
> this angle got broken.
> Will fix. 

Btw, I noticed this when I was looking at:

	WARN(sgx_usage_count != 1, "...");

in patch 4 and wondering why it's not "!= 0".

Please don't forget to update that when needed.