[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <a2570676d1b06a0227e733ff09d408567d1d615d.camel@intel.com>
Date: Thu, 31 Aug 2023 17:29:36 +0000
From: "Edgecombe, Rick P" <rick.p.edgecombe@...el.com>
To: "linux-hyperv@...r.kernel.org" <linux-hyperv@...r.kernel.org>,
"Lutomirski, Andy" <luto@...nel.org>,
"dave.hansen@...ux.intel.com" <dave.hansen@...ux.intel.com>,
"thomas.lendacky@....com" <thomas.lendacky@....com>,
"haiyangz@...rosoft.com" <haiyangz@...rosoft.com>,
"kirill.shutemov@...ux.intel.com" <kirill.shutemov@...ux.intel.com>,
"Christopherson,, Sean" <seanjc@...gle.com>,
"mingo@...hat.com" <mingo@...hat.com>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"kys@...rosoft.com" <kys@...rosoft.com>,
"Cui, Dexuan" <decui@...rosoft.com>,
"mikelley@...rosoft.com" <mikelley@...rosoft.com>,
"hpa@...or.com" <hpa@...or.com>,
"peterz@...radead.org" <peterz@...radead.org>,
"wei.liu@...nel.org" <wei.liu@...nel.org>,
"bp@...en8.de" <bp@...en8.de>,
"sathyanarayanan.kuppuswamy@...ux.intel.com"
<sathyanarayanan.kuppuswamy@...ux.intel.com>,
"tglx@...utronix.de" <tglx@...utronix.de>,
"x86@...nel.org" <x86@...nel.org>
Subject: Re: [RFC PATCH 1/1] x86/mm: Mark CoCo VM pages invalid while moving
between private and shared
On Wed, 2023-08-30 at 16:40 -0700, Rick Edgecombe wrote:
> This is a bit of an existing problem, but the failure cases of these
> set_memory_en/decrypted() operations does not look to be in great
> shape. It could fail halfway through if it needs to split the direct
> map under memory pressure, in which case some of the callers will see
> the error and free the unmapped pages to the direct map. (I was
> looking
> at dma_direct_alloc()) Other's just leak the pages.
>
> But the situation before the patch is not much better, since the
> direct
> map change or enc_status_change_prepare/finish() could fail and leave
> the pages in an inconsistent state, like this patch is trying to
> address.
>
> This lack of rollback on failure for CPA calls needs particular odd
> handling in all the set_memory() callers. The way is to make a CPA
> call
> to restore it to the previous permission, regardless of the error
> code
> returned in the initial call that failed. The callers depend on any
> PTE
> change successfully made having any needed splits already done for
> those PTEs, so the restore can succeed at least as far as the failed
> CPA call got.
Wait, since this does set_memory_np() as the first step for both
set_memory_encrypted() and set_memory_decrypted(), that pattern in the
callers wouldn't work. I wonder if it should try to rollback itself if
set_memory_np() fails (call set_memory_p() before returning the error).
At least that will handle failures that happen on the guest side.
>
> In this COCO case apparently the enc_status_change_prepare/finish()
> could fail too (and maybe not have the same forward progress
> behavior?). So I'm not sure what you can do in that case.
>
> I'm also not sure how bad it is to free encryption mismatched pages.
> Is
> it the same as freeing unmapped pages? (likely oops or panic)
Powered by blists - more mailing lists