Re: [PATCH V3 19/30] x86/sgx: Free up EPC pages directly to support large page ranges

[Jarkko Sakkinen <jarkko@kernel.org>]

On Tue, 2022-04-05 at 10:13 -0700, Reinette Chatre wrote:
> Hi Jarkko,
> 
> On 4/5/2022 12:11 AM, Jarkko Sakkinen wrote:
> > On Mon, Apr 04, 2022 at 09:49:27AM -0700, Reinette Chatre wrote:
> > > The page reclaimer ensures availability of EPC pages across all
> > > enclaves. In support of this it runs independently from the
> > > individual enclaves in order to take locks from the different
> > > enclaves as it writes pages to swap.
> > > 
> > > When needing to load a page from swap an EPC page needs to be
> > > available for its contents to be loaded into. Loading an existing
> > > enclave page from swap does not reclaim EPC pages directly if
> > > none are available, instead the reclaimer is woken when the
> > > available EPC pages are found to be below a watermark.
> > > 
> > > When iterating over a large number of pages in an oversubscribed
> > > environment there is a race between the reclaimer woken up and
> > > EPC pages reclaimed fast enough for the page operations to proceed.
> > > 
> > > Ensure there are EPC pages available before attempting to load
> > > a page that may potentially be pulled from swap into an available
> > > EPC page.
> > > 
> > > Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
> > > ---
> > > No changes since V2
> > > 
> > > Changes since v1:
> > > - Reword commit message.
> > > 
> > >  arch/x86/kernel/cpu/sgx/ioctl.c | 6 ++++++
> > >  arch/x86/kernel/cpu/sgx/main.c  | 6 ++++++
> > >  arch/x86/kernel/cpu/sgx/sgx.h   | 1 +
> > >  3 files changed, 13 insertions(+)
> > > 
> > > diff --git a/arch/x86/kernel/cpu/sgx/ioctl.c b/arch/x86/kernel/cpu/sgx/ioctl.c
> > > index 515e1961cc02..f88bc1236276 100644
> > > --- a/arch/x86/kernel/cpu/sgx/ioctl.c
> > > +++ b/arch/x86/kernel/cpu/sgx/ioctl.c
> > > @@ -777,6 +777,8 @@ sgx_enclave_restrict_permissions(struct sgx_encl *encl,
> > >         for (c = 0 ; c < modp->length; c += PAGE_SIZE) {
> > >                 addr = encl->base + modp->offset + c;
> > >  
> > > +               sgx_direct_reclaim();
> > > +
> > >                 mutex_lock(&encl->lock);
> > >  
> > >                 entry = sgx_encl_load_page(encl, addr);
> > > @@ -934,6 +936,8 @@ static long sgx_enclave_modify_type(struct sgx_encl *encl,
> > >         for (c = 0 ; c < modt->length; c += PAGE_SIZE) {
> > >                 addr = encl->base + modt->offset + c;
> > >  
> > > +               sgx_direct_reclaim();
> > > +
> > >                 mutex_lock(&encl->lock);
> > >  
> > >                 entry = sgx_encl_load_page(encl, addr);
> > > @@ -1129,6 +1133,8 @@ static long sgx_encl_remove_pages(struct sgx_encl *encl,
> > >         for (c = 0 ; c < params->length; c += PAGE_SIZE) {
> > >                 addr = encl->base + params->offset + c;
> > >  
> > > +               sgx_direct_reclaim();
> > > +
> > >                 mutex_lock(&encl->lock);
> > >  
> > >                 entry = sgx_encl_load_page(encl, addr);
> > > diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c
> > > index 6e2cb7564080..545da16bb3ea 100644
> > > --- a/arch/x86/kernel/cpu/sgx/main.c
> > > +++ b/arch/x86/kernel/cpu/sgx/main.c
> > > @@ -370,6 +370,12 @@ static bool sgx_should_reclaim(unsigned long watermark)
> > >                !list_empty(&sgx_active_page_list);
> > >  }
> > >  
> > > +void sgx_direct_reclaim(void)
> > > +{
> > > +       if (sgx_should_reclaim(SGX_NR_LOW_PAGES))
> > > +               sgx_reclaim_pages();
> > > +}
> > 
> > Please, instead open code this to both locations - not enough redundancy
> > to be worth of new function. Causes only unnecessary cross-referencing
> > when maintaining. Otherwise, I agree with the idea.
> > 
> 
> hmmm, that means the heart of the reclaimer (sgx_reclaim_pages()) would be
> made available for direct use from everywhere in the driver. I will look into this.
> 
> Reinette
> 

It's a valid enough point. Let's keep it as it is :-)

Acked-by: Jarkko Sakkinen <jarkko@kernel.org>

BR, Jarkko