Re: [PATCH v3 03/28] x86/sgx: Add 'struct sgx_epc_lru_lists' to encapsulate lru list(s)

[Haitao Huang]

Hi Kai
On Mon, 24 Jul 2023 05:04:48 -0500, Huang, Kai <kai.huang@xxxxxxxxx> wrote:

> On Mon, 2023-07-17 at 08:23 -0500, Haitao Huang wrote:
> > On Mon, 17 Jul 2023 07:45:36 -0500, Jarkko Sakkinen <jarkko@xxxxxxxxxx>
> > wrote:
> >
> > > On Wed Jul 12, 2023 at 11:01 PM UTC, Haitao Huang wrote:
> > > > From: Kristen Carlson Accardi <kristen@xxxxxxxxxxxxxxx>
> > > >
> > > > Introduce a data structure to wrap the existing reclaimable list
> > > > and its spinlock in a struct to minimize the code changes needed
> > > > to handle multiple LRUs as well as reclaimable and non-reclaimable
> > > > lists. The new structure will be used in a following set of patches  
> > to
> > > > implement SGX EPC cgroups.
>
> Although briefly mentioned in the first patch, it would be better to put  
> more
> background about the "reclaimable" and "non-reclaimable" thing here,  
> focusing on
> _why_ we need multiple LRUs (presumably you mean two lists: reclaimable  
> and non-
> reclaimable).
Sure I can add a little more background to introduce the  
reclaimable/unreclaimable concept. But why we need multiple LRUs would be  
self-evident in later patches, not sure I will add details here.

> > > >
> > > > The changes to the structure needed for unreclaimable lists will be
> > > > added in later patches.
> > > >
> > > > Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> > > > Signed-off-by: Kristen Carlson Accardi <kristen@xxxxxxxxxxxxxxx>
> > > > Signed-off-by: Haitao Huang <haitao.huang@xxxxxxxxxxxxxxx>
> > > > Cc: Sean Christopherson <seanjc@xxxxxxxxxx>
> > > >
> > > > V3:
> > > > Removed the helper functions and revised commit messages
>
> Please put change history into:
>
> ---
>   change history
> ---
>
> So it can be stripped away when applying the patch.
Will do that.

> > > > ---
> > > >  arch/x86/kernel/cpu/sgx/sgx.h | 17 +++++++++++++++++
> > > >  1 file changed, 17 insertions(+)
> > > >
> > > > diff --git a/arch/x86/kernel/cpu/sgx/sgx.h
> > > > b/arch/x86/kernel/cpu/sgx/sgx.h
> > > > index f6e3c5810eef..77fceba73a25 100644
> > > > --- a/arch/x86/kernel/cpu/sgx/sgx.h
> > > > +++ b/arch/x86/kernel/cpu/sgx/sgx.h
> > > > @@ -92,6 +92,23 @@ static inline void *sgx_get_epc_virt_addr(struct
> > > > sgx_epc_page *page)
> > > >  	return section->virt_addr + index * PAGE_SIZE;
> > > >  }
> > > >
> > > > +/*
> > > > + * This data structure wraps a list of reclaimable EPC pages, and a
> > > > list of
> > > > + * non-reclaimable EPC pages and is used to implement a LRU policy
> > > > during
> > > > + * reclamation.
> > > > + */
>
> I'd prefer to not mention the "non-reclaimable" thing in this patch, but  
> defer
> to the one actually introduces the "non-reclaimable" list.  Actually, I  
> don't
> think we even need this comment, given you have this in the structure:
>
> 	struct list_head reclaimable;

Agreed.

> Which already explains the "reclaimable" list.  I suppose the  
> non-reclaimable
> list would be named similarly thus need no comment either.
>
> Also, I am wondering why you need to split this out as a separate  
> patch.  It
> basically does nothing.  To me you should just merge this to the next  
> patch,

I think Kristen splitted the original patch based on Dave's comments:

https://lore.kernel.org/all/e71d76b2-4368-4627-abd4-2163e6786a20@xxxxxxxxx/

> which actually does what you claimed in the changelog:
>
> 	Introduce a data structure to wrap the existing reclaimable list and 
> 	its spinlock ...
>
> Then this can be an infrastructure change patch, which doesn't bring any
> functional change, to support the non-reclaimable list.
>
>
> > > > +struct sgx_epc_lru_lists {
> > > > +	/* Must acquire this lock to access */
> > > > +	spinlock_t lock;
> > >
> > > Isn't this self-explanatory, why the inline comment?
> >
> > I got a warning from the checkpatch script complaining this lock needs
> > comments.
>
> I suspected this, so I applied this patch, removed the comment,  
> generated a new
> patch, and run checkpatch.pl for it.  It didn't report any warning/error  
> in my
> testing.
>
> Are you sure you got a warning?

I did a reran and it's actually a "CHECK" I got:

$ ./scripts/checkpatch.pl --strict  
0001-x86-sgx-Add-struct-sgx_epc_lru_lists-to-encapsulate-.patch
CHECK: spinlock_t definition without comment
#41: FILE: arch/x86/kernel/cpu/sgx/sgx.h:101:
+       spinlock_t lock;

total: 0 errors, 0 warnings, 1 checks, 22 lines checked

Thanks
Haitao