Re: [PATCH v6 06/11] intel_sgx: driver for Intel Software Guard Extensions

[Jarkko Sakkinen]

On Thu, Nov 30, 2017 at 09:32:01AM -0800, Sean Christopherson wrote:
> On Sat, Nov 25, 2017 at 09:29:24PM +0200, Jarkko Sakkinen wrote:
> > +static void *sgx_try_alloc_page(void)
> > +{
> > +       struct sgx_epc_bank *bank;
> > +       void *page = NULL;
> > +       int i;
> > +
> > +       for (i = 0; i < sgx_nr_epc_banks; i++) {
> > +               bank = &sgx_epc_banks[i];
> > +
> > +               down_write(&bank->lock);
> 
> Is a R/W semaphore actually preferable to a spinlock?  Concurrent
> free calls don't seem that interesting/beneficial because freeing
> an enclave's pages isn't multiplexed across multiple CPUs, unlike
> the allocation of EPC pages.

I get about ~10-15% performance increase on high stress. It is benefical
to spinlock.

> As a whole, I'm not a fan of packing the EPC page pointers into an
> array rather than encapsulating them in a struct+list.  The primary
> benefit I see for the array approach is that it saves ~8 bytes per
> free EPC page, but at a cost of increased memory usage for in-use
> pages and severely restricting the ability to enhance/modify how
> EPC pages are tracked, reclaimed, etc...

This is not true. You can put EPC page descriptor to any structure you
want.

list_head can be dropped from struct sgx_enc_page too.

> The main issue is that the array approach relies on the caller to
> handle reclaim.  This effectively makes it impossible to reclaim
> pages from multiple processes, requires other consumers e.g. KVM
> to implement their own reclaim logic and kthread, and prevents
> cgroup accounting because the cgroup can't initiate reclaim.

Not really following here.

/Jarkko