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
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20200512002815.2708-3-sean.j.christopherson@intel.com>
Date:   Mon, 11 May 2020 17:28:15 -0700
From:   Sean Christopherson <sean.j.christopherson@...el.com>
To:     stable@...r.kernel.org,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Ben Hutchings <ben.hutchings@...ethink.co.uk>,
        Sasha Levin <sashal@...nel.org>
Cc:     Paolo Bonzini <pbonzini@...hat.com>, linux-kernel@...r.kernel.org,
        Tobias Urdin <tobias.urdin@...ero.com>
Subject: [PATCH 4.19 STABLE v2 2/2] KVM: VMX: Mark RCX, RDX and RSI as clobbered in vmx_vcpu_run()'s asm blob

Based on upstream commit f3689e3f17f064fd4cd5f0cb01ae2395c94f39d9.

Save RCX, RDX and RSI to fake outputs to coerce the compiler into
treating them as clobbered.  RCX in particular is likely to be reused by
the compiler to dereference the 'struct vcpu_vmx' pointer, which will
result in a null pointer dereference now that RCX is zeroed by the asm
blob.

Tag the asm() blob as volatile to prevent GCC from dropping the blob,
which is possible now that the blob has output values, all of which are
unused.

Upstream commit f3689e3f17f06 ("KVM: VMX: Save RSI to an unused output
in the vCPU-run asm blob") is not a direct equivalent of this patch. As
its shortlog states, it only tagged RSI as clobbered, whereas here RCX
and RDX are also clobbered.

In upstream at the time of the offending commit (b4be98039a92 in 4.19,
0e0ab73c9a024 upstream), the inline asm blob had previously been moved
to a dedicated helper, __vmx_vcpu_run().  For unrelated reasons,
__vmx_vcpu_run() was put into its own optimization unit, which for all
intents and purposes made it impossible to consume clobbered registers
because RCX, RDX and RSI are volatile and __vmx_vcpu_run() couldn't
itself be inlined.  In other words, the bug existed but couldn't be hit.

Similarly, the lack of "volatile" was also a bug in upstream that was
hidden by an unrelated change that exists in upstream but not in 4.19.
In this case, the asm blob also uses ASM_CALL_CONSTRAINT (marks RSP as
being an input/output constraint) in upstream to play nice with objtool
due the blob making a CALL.  In 4.19, there is no CALL and thus no
ASM_CALL_CONSTRAINT.

Furthermore, both of the lurking bugs were blasted away in upstream by
commits 5e0781df1899 ("KVM: VMX: Move vCPU-run code to a proper assembly
routine") and fc2ba5a27a1a ("KVM: VMX: Call vCPU-run asm sub-routine
from C and remove clobbering"), i.e. these bugs will never be directly
fixed in upstream.

Reported-by: Tobias Urdin <tobias.urdin@...ero.com>
Fixes: b4be98039a92 ("KVM: VMX: Zero out *all* general purpose registers after VM-Exit")
Signed-off-by: Sean Christopherson <sean.j.christopherson@...el.com>
---
 arch/x86/kvm/vmx.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 5b06a98ffd4cb..f08c287b62426 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -10771,7 +10771,7 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
 	else if (static_branch_unlikely(&mds_user_clear))
 		mds_clear_cpu_buffers();
 
-	asm(
+	asm volatile (
 		/* Store host registers */
 		"push %%" _ASM_DX "; push %%" _ASM_BP ";"
 		"push %%" _ASM_CX " \n\t" /* placeholder for guest rcx */
@@ -10882,7 +10882,8 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
 		".global vmx_return \n\t"
 		"vmx_return: " _ASM_PTR " 2b \n\t"
 		".popsection"
-	      : : "c"(vmx), "d"((unsigned long)HOST_RSP), "S"(evmcs_rsp),
+	      : "=c"((int){0}), "=d"((int){0}), "=S"((int){0})
+	      : "c"(vmx), "d"((unsigned long)HOST_RSP), "S"(evmcs_rsp),
 		[launched]"i"(offsetof(struct vcpu_vmx, __launched)),
 		[fail]"i"(offsetof(struct vcpu_vmx, fail)),
 		[host_rsp]"i"(offsetof(struct vcpu_vmx, host_rsp)),
-- 
2.26.0

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ