| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <1050a84caa9036e3ddb580fb510e634a216e692c.camel@intel.com>
Date: Thu, 13 Mar 2025 23:17:23 +0000
From: "Edgecombe, Rick P" <rick.p.edgecombe@...el.com>
To: "tglx@...utronix.de" <tglx@...utronix.de>, "peterz@...radead.org"
<peterz@...radead.org>, "mingo@...hat.com" <mingo@...hat.com>, "Hansen, Dave"
<dave.hansen@...el.com>, "Huang, Kai" <kai.huang@...el.com>, "bp@...en8.de"
<bp@...en8.de>, "kirill.shutemov@...ux.intel.com"
<kirill.shutemov@...ux.intel.com>
CC: "ashish.kalra@....com" <ashish.kalra@....com>, "dyoung@...hat.com"
<dyoung@...hat.com>, "thomas.lendacky@....com" <thomas.lendacky@....com>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"seanjc@...gle.com" <seanjc@...gle.com>, "dwmw@...zon.co.uk"
<dwmw@...zon.co.uk>, "pbonzini@...hat.com" <pbonzini@...hat.com>,
"bhe@...hat.com" <bhe@...hat.com>, "Yamahata, Isaku"
<isaku.yamahata@...el.com>, "nik.borisov@...e.com" <nik.borisov@...e.com>,
"Chatre, Reinette" <reinette.chatre@...el.com>, "hpa@...or.com"
<hpa@...or.com>, "sagis@...gle.com" <sagis@...gle.com>,
"david.kaplan@....com" <david.kaplan@....com>, "x86@...nel.org"
<x86@...nel.org>, "Williams, Dan J" <dan.j.williams@...el.com>
Subject: Re: [RFC PATCH 2/5] x86/kexec: Do unconditional WBINVD for bare-metal
in relocate_kernel()
On Thu, 2025-03-13 at 00:34 +1300, Kai Huang wrote:
> For both SME and TDX, dirty cachelines with and without the encryption
> bit(s) of the same physical memory address can coexist and the CPU can
> flush them back to memory in random order.
>
A lot going on in this sentence, how about simplifying it:
For SME and TDX, multiple dirty cachelines for the same memory can co-exist, and
the CPU can flush them back to memory in a random order.
> During kexec, the caches
> must be flushed before jumping to the new kernel to avoid silent memory
> corruption to the new kernel.
During kexec, the caches must be flushed before jumping to the new kernel to
avoid silent memory corruption when a cacheline with a different encryption
property is written back over whatever encryption properties the new kernel is
using.
...it distributes some of the details from the first sentence into the second.
Easier to read or no? I'm not sure.
>
> The WBINVD in stop_this_cpu() flushes caches for all remote CPUs when
> they are being stopped. For SME, the WBINVD in relocate_kernel()
> flushes the cache for the last running CPU (which is doing kexec).
>
> Similarly, to support kexec for TDX host, after stopping all remote CPUs
> with cache flushed, the kernel needs to flush cache for the last running
> CPU.
I mentioned this in a previous version. I think you need to give some hint to
where you are going before you start listing facts. Like:
During kexec, WBINVD needs to be executed on each CPU for TDX and SME. On the
remote CPUs this is covered in stop_this_cpu() for both TDX and SME. For the
kexecing CPU, SME handles this in relocate_kernel(). This leaves the TDX case
for the kexec-ing CPU still to implement.
...it first says the overall problem to solve, then explains what is missing in
the current code to solve it. The reader is already thinking of what the
solutions should be and...
>
> Use the existing WBINVD in relocate_kernel() to cover TDX host as well.
...they read the solution just as they are wondering about it. The reader can
feel like the change is aligned with their thinking.
>
> Just do unconditional
>
"Unconditional". Now I'm starting to think about how unconditional wbinvd will
be.
> WBINVD to cover both SME and TDX instead of
> sprinkling additional vendor-specific checks. Kexec is a slow path, and
> the additional WBINVD is acceptable for the sake of simplicity and
> maintainability.
>
> But only do WBINVD for bare-metal
>
But wait, now I'm learning it's not unconditional. I need to re-think what I
just evaluated. And I'm doubting the earlier statements because I just got
surprised.
> because TDX guests and SEV-ES/SEV-SNP
> guests will get unexpected (and yet unnecessary) exception (#VE or #VC)
> which the kernel is unable to handle at the time of relocate_kernel()
> since the kernel has torn down the IDT.
>
> Remove the host_mem_enc_active local variable and directly use
> !cpu_feature_enabled(X86_FEATURE_HYPERVISOR) as an argument of calling
> relocate_kernel().
>
Start with the problem here. It just describes another change and I'm not sure
why when I start reading it.
By problem I mean that host_mem_enc_active doesn't fit the conditional anymore,
so it needs to be changed.
> cpu_feature_enabled() is always inline but not a
I was just noticing this on the other patch. Actually it could call into some
kasan stuff.
> function call, thus it is safe to use after load_segments() when call
> depth tracking is enabled.
This function call tracking stuff is a wild card at the end. What about
describing the rules this function needs to follow due to call depth tracking,
and explain why the change does that.
>
> Signed-off-by: Kai Huang <kai.huang@...el.com>
> Reviewed-by: Kirill A. Shutemov <kirill.shutemov@...ux.intel.com>
> Cc: Tom Lendacky <thomas.lendacky@....com>
> Cc: Dave Young <dyoung@...hat.com>
> Cc: David Kaplan <david.kaplan@....com>
> Reviewed-by: Tom Lendacky <thomas.lendacky@....com>
> Tested-by: David Kaplan <david.kaplan@....com>
> ---
> arch/x86/include/asm/kexec.h | 2 +-
> arch/x86/kernel/machine_kexec_64.c | 14 ++++++--------
> arch/x86/kernel/relocate_kernel_64.S | 15 ++++++++++-----
> 3 files changed, 17 insertions(+), 14 deletions(-)
>
> diff --git a/arch/x86/include/asm/kexec.h b/arch/x86/include/asm/kexec.h
> index 8ad187462b68..48c313575262 100644
> --- a/arch/x86/include/asm/kexec.h
> +++ b/arch/x86/include/asm/kexec.h
> @@ -123,7 +123,7 @@ relocate_kernel_fn(unsigned long indirection_page,
> unsigned long pa_control_page,
> unsigned long start_address,
> unsigned int preserve_context,
> - unsigned int host_mem_enc_active);
> + unsigned int bare_metal);
> #endif
> extern relocate_kernel_fn relocate_kernel;
> #define ARCH_HAS_KIMAGE_ARCH
> diff --git a/arch/x86/kernel/machine_kexec_64.c b/arch/x86/kernel/machine_kexec_64.c
> index a68f5a0a9f37..0e9808eeb63e 100644
> --- a/arch/x86/kernel/machine_kexec_64.c
> +++ b/arch/x86/kernel/machine_kexec_64.c
> @@ -346,16 +346,9 @@ void __nocfi machine_kexec(struct kimage *image)
> {
> unsigned long reloc_start = (unsigned long)__relocate_kernel_start;
> relocate_kernel_fn *relocate_kernel_ptr;
> - unsigned int host_mem_enc_active;
> int save_ftrace_enabled;
> void *control_page;
>
> - /*
> - * This must be done before load_segments() since if call depth tracking
> - * is used then GS must be valid to make any function calls.
> - */
> - host_mem_enc_active = cc_platform_has(CC_ATTR_HOST_MEM_ENCRYPT);
> -
> #ifdef CONFIG_KEXEC_JUMP
> if (image->preserve_context)
> save_processor_state();
> @@ -398,6 +391,11 @@ void __nocfi machine_kexec(struct kimage *image)
> *
> * I take advantage of this here by force loading the
> * segments, before I zap the gdt with an invalid value.
> + *
> + * load_segments() resets GS to 0. Don't make any function call
> + * after here since call depth tracking uses per-CPU variables to
> + * operate (relocate_kernel() is explicitly ignored by call depth
> + * tracking).
I think I suggested you should call out the opportunistic change here in the
log. Did you disagree?
> */
> load_segments();
> /*
> @@ -412,7 +410,7 @@ void __nocfi machine_kexec(struct kimage *image)
> virt_to_phys(control_page),
> image->start,
> image->preserve_context,
> - host_mem_enc_active);
> + !cpu_feature_enabled(X86_FEATURE_HYPERVISOR));
>
> #ifdef CONFIG_KEXEC_JUMP
> if (image->preserve_context)
> diff --git a/arch/x86/kernel/relocate_kernel_64.S b/arch/x86/kernel/relocate_kernel_64.S
> index b44d8863e57f..dc1a59cd8139 100644
> --- a/arch/x86/kernel/relocate_kernel_64.S
> +++ b/arch/x86/kernel/relocate_kernel_64.S
> @@ -50,7 +50,7 @@ SYM_CODE_START_NOALIGN(relocate_kernel)
> * %rsi pa_control_page
> * %rdx start address
> * %rcx preserve_context
> - * %r8 host_mem_enc_active
> + * %r8 bare_metal
> */
>
> /* Save the CPU context, used for jumping back */
> @@ -107,7 +107,7 @@ SYM_CODE_START_LOCAL_NOALIGN(identity_mapped)
> /*
> * %rdi indirection page
> * %rdx start address
> - * %r8 host_mem_enc_active
> + * %r8 bare_metal
> * %r9 page table page
> * %r11 preserve_context
> * %r13 original CR4 when relocate_kernel() was invoked
> @@ -156,14 +156,19 @@ SYM_CODE_START_LOCAL_NOALIGN(identity_mapped)
> movq %r9, %cr3
>
> /*
> - * If SME is active, there could be old encrypted cache line
> + * If SME/TDX is active, there could be old encrypted cache line
> * entries that will conflict with the now unencrypted memory
> * used by kexec. Flush the caches before copying the kernel.
> + *
> + * Do WBINVD for bare-metal only to cover both SME and TDX. Doing
> + * WBINVD in guest results in an unexpected exception (#VE or #VC)
> + * for TDX and SEV-ES/SNP guests which then crashes the guest (the
> + * kernel has torn down the IDT).
> */
> testq %r8, %r8
> - jz .Lsme_off
> + jz .Lno_wbinvd
> wbinvd
> -.Lsme_off:
> +.Lno_wbinvd:
>
> call swap_pages
>
Powered by blists - more mailing lists