[PATCH] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[PATCH] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Yu-cheng Yu]

The compacted-format XSAVES area is determined at boot time and
never changed after.  The field xsave.header.xcomp_bv indicates
which components are in the fixed XSAVES format.

In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
format since at the time there is no valid data.

However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
as in commit: b22cbe404a9cc3c7949e380fa1861e31934c8978, and when
__fpu_restore_sig() does fpu__restore() for a COMPAT-mode app,
a #GP occurs.  This can be easily triggered by doing valgrind on
a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
others:

	https://bugzilla.kernel.org/show_bug.cgi?id=190061

Fix it by setting xcomp_bv correctly.

Signed-off-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
Reported-by: Joakim Tjernlund <Joakim.Tjernlund@infinera.com>
---
 arch/x86/kernel/fpu/core.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/fpu/core.c b/arch/x86/kernel/fpu/core.c
index c289e2f..e540dc1 100644
--- a/arch/x86/kernel/fpu/core.c
+++ b/arch/x86/kernel/fpu/core.c
@@ -9,6 +9,7 @@
 #include <asm/fpu/regset.h>
 #include <asm/fpu/signal.h>
 #include <asm/fpu/types.h>
+#include <asm/fpu/xstate.h>
 #include <asm/traps.h>
 
 #include <linux/hardirq.h>
@@ -235,7 +236,8 @@ void fpstate_init(union fpregs_state *state)
 	 * it will #GP. Make sure it is replaced after the memset().
 	 */
 	if (static_cpu_has(X86_FEATURE_XSAVES))
-		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT;
+		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT |
+					       xfeatures_mask;
 
 	if (static_cpu_has(X86_FEATURE_FXSR))
 		fpstate_init_fxstate(&state->fxsave);
-- 
1.9.1

Re: [PATCH] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Ingo Molnar]

* Yu-cheng Yu <yu-cheng.yu@intel.com> wrote:

> The compacted-format XSAVES area is determined at boot time and
> never changed after.  The field xsave.header.xcomp_bv indicates
> which components are in the fixed XSAVES format.
> 
> In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
> format since at the time there is no valid data.
> 
> However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
> as in commit: b22cbe404a9cc3c7949e380fa1861e31934c8978, and when
> __fpu_restore_sig() does fpu__restore() for a COMPAT-mode app,
> a #GP occurs.  This can be easily triggered by doing valgrind on
> a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
> others:
> 
> 	https://bugzilla.kernel.org/show_bug.cgi?id=190061
> 
> Fix it by setting xcomp_bv correctly.
> 
> Signed-off-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
> Reported-by: Joakim Tjernlund <Joakim.Tjernlund@infinera.com>
> ---
>  arch/x86/kernel/fpu/core.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kernel/fpu/core.c b/arch/x86/kernel/fpu/core.c
> index c289e2f..e540dc1 100644
> --- a/arch/x86/kernel/fpu/core.c
> +++ b/arch/x86/kernel/fpu/core.c
> @@ -9,6 +9,7 @@
>  #include <asm/fpu/regset.h>
>  #include <asm/fpu/signal.h>
>  #include <asm/fpu/types.h>
> +#include <asm/fpu/xstate.h>
>  #include <asm/traps.h>
>  
>  #include <linux/hardirq.h>
> @@ -235,7 +236,8 @@ void fpstate_init(union fpregs_state *state)
>  	 * it will #GP. Make sure it is replaced after the memset().
>  	 */
>  	if (static_cpu_has(X86_FEATURE_XSAVES))
> -		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT;
> +		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT |
> +					       xfeatures_mask;

Ok, I have applied this - but it would be cleaner to go one step further and add a 
fpstate_init_xstate() method that does this in xstate.c and hides the details from 
arch/x86/kernel/fpu/core.c.

Similar to how the FX-state initialization is done today:

>  	if (static_cpu_has(X86_FEATURE_FXSR))
>  		fpstate_init_fxstate(&state->fxsave);

Thanks,

	Ingo

Re: [PATCH] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Joakim Tjernlund]

On Tue, 2017-01-24 at 09:09 +0100, Ingo Molnar wrote:
> * Yu-cheng Yu <yu-cheng.yu@intel.com> wrote:

This needs to go to stable too, 4.8 and 4.9

 Jocke

> 
> > The compacted-format XSAVES area is determined at boot time and
> > never changed after.  The field xsave.header.xcomp_bv indicates
> > which components are in the fixed XSAVES format.
> > 
> > In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
> > format since at the time there is no valid data.
> > 
> > However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
> > as in commit: b22cbe404a9cc3c7949e380fa1861e31934c8978, and when
> > __fpu_restore_sig() does fpu__restore() for a COMPAT-mode app,
> > a #GP occurs.  This can be easily triggered by doing valgrind on
> > a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
> > others:
> > 
> > 	https://bugzilla.kernel.org/show_bug.cgi?id=190061
> > 
> > Fix it by setting xcomp_bv correctly.
> > 
> > Signed-off-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
> > Reported-by: Joakim Tjernlund <Joakim.Tjernlund@infinera.com>
> > ---
> >  arch/x86/kernel/fpu/core.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/x86/kernel/fpu/core.c b/arch/x86/kernel/fpu/core.c
> > index c289e2f..e540dc1 100644
> > --- a/arch/x86/kernel/fpu/core.c
> > +++ b/arch/x86/kernel/fpu/core.c
> > @@ -9,6 +9,7 @@
> >  #include <asm/fpu/regset.h>
> >  #include <asm/fpu/signal.h>
> >  #include <asm/fpu/types.h>
> > +#include <asm/fpu/xstate.h>
> >  #include <asm/traps.h>
> >  
> >  #include <linux/hardirq.h>
> > @@ -235,7 +236,8 @@ void fpstate_init(union fpregs_state *state)
> >  	 * it will #GP. Make sure it is replaced after the memset().
> >  	 */
> >  	if (static_cpu_has(X86_FEATURE_XSAVES))
> > -		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT;
> > +		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT |
> > +					       xfeatures_mask;
> 
> Ok, I have applied this - but it would be cleaner to go one step further and add a 
> fpstate_init_xstate() method that does this in xstate.c and hides the details from 
> arch/x86/kernel/fpu/core.c.
> 
> Similar to how the FX-state initialization is done today:
> 
> >  	if (static_cpu_has(X86_FEATURE_FXSR))
> >  		fpstate_init_fxstate(&state->fxsave);
> 
> Thanks,
> 
> 	Ingo

[tip:x86/urgent] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[tip-bot for Yu-cheng Yu]

Commit-ID:  dffba9a31c7769be3231c420d4b364c92ba3f1ac
Gitweb:     http://git.kernel.org/tip/dffba9a31c7769be3231c420d4b364c92ba3f1ac
Author:     Yu-cheng Yu <yu-cheng.yu@intel.com>
AuthorDate: Mon, 23 Jan 2017 14:54:44 -0800
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Tue, 24 Jan 2017 09:04:48 +0100

x86/fpu/xstate: Fix xcomp_bv in XSAVES header

The compacted-format XSAVES area is determined at boot time and
never changed after.  The field xsave.header.xcomp_bv indicates
which components are in the fixed XSAVES format.

In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
format since at the time there is no valid data.

However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
as in commit:

  b22cbe404a9c x86/fpu: Fix invalid FPU ptrace state after execve()

and when __fpu_restore_sig() does fpu__restore() for a COMPAT-mode
app, a #GP occurs.  This can be easily triggered by doing valgrind on
a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
others:

	https://bugzilla.kernel.org/show_bug.cgi?id=190061

Fix it by setting xcomp_bv correctly.

This patch also moves the xcomp_bv initialization to the proper
place, which was in copyin_to_xsaves() as of:

  4c833368f0bf x86/fpu: Set the xcomp_bv when we fake up a XSAVES area

which fixed the bug too, but it's more efficient and cleaner to
initialize things once per boot, not for every signal handling
operation.

Reported-by: Kevin Hao <haokexin@gmail.com>
Reported-by: Joakim Tjernlund <Joakim.Tjernlund@infinera.com>
Signed-off-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi V. Shankar <ravi.v.shankar@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: haokexin@gmail.com
Link: http://lkml.kernel.org/r/1485212084-4418-1-git-send-email-yu-cheng.yu@intel.com
[ Combined it with 4c833368f0bf. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/kernel/fpu/core.c   | 4 +++-
 arch/x86/kernel/fpu/xstate.c | 1 -
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/fpu/core.c b/arch/x86/kernel/fpu/core.c
index e4e97a5..de72344 100644
--- a/arch/x86/kernel/fpu/core.c
+++ b/arch/x86/kernel/fpu/core.c
@@ -9,6 +9,7 @@
 #include <asm/fpu/regset.h>
 #include <asm/fpu/signal.h>
 #include <asm/fpu/types.h>
+#include <asm/fpu/xstate.h>
 #include <asm/traps.h>
 
 #include <linux/hardirq.h>
@@ -183,7 +184,8 @@ void fpstate_init(union fpregs_state *state)
 	 * it will #GP. Make sure it is replaced after the memset().
 	 */
 	if (static_cpu_has(X86_FEATURE_XSAVES))
-		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT;
+		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT |
+					       xfeatures_mask;
 
 	if (static_cpu_has(X86_FEATURE_FXSR))
 		fpstate_init_fxstate(&state->fxsave);
diff --git a/arch/x86/kernel/fpu/xstate.c b/arch/x86/kernel/fpu/xstate.c
index e287b90..1d77704 100644
--- a/arch/x86/kernel/fpu/xstate.c
+++ b/arch/x86/kernel/fpu/xstate.c
@@ -1070,7 +1070,6 @@ int copyin_to_xsaves(const void *kbuf, const void __user *ubuf,
 	 * Add back in the features that came in from userspace:
 	 */
 	xsave->header.xfeatures |= xfeatures;
-	xsave->header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT | xsave->header.xfeatures;
 
 	return 0;
 }

Re: [tip:x86/urgent] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Yu-cheng Yu]

On Tue, Jan 24, 2017 at 12:46:17AM -0800, tip-bot for Yu-cheng Yu wrote:
> Commit-ID:  dffba9a31c7769be3231c420d4b364c92ba3f1ac
> Gitweb:     http://git.kernel.org/tip/dffba9a31c7769be3231c420d4b364c92ba3f1ac
> Author:     Yu-cheng Yu <yu-cheng.yu@intel.com>
> AuthorDate: Mon, 23 Jan 2017 14:54:44 -0800
> Committer:  Ingo Molnar <mingo@kernel.org>
> CommitDate: Tue, 24 Jan 2017 09:04:48 +0100
> 
> x86/fpu/xstate: Fix xcomp_bv in XSAVES header
> 
> The compacted-format XSAVES area is determined at boot time and
> never changed after.  The field xsave.header.xcomp_bv indicates
> which components are in the fixed XSAVES format.
> 
> In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
> format since at the time there is no valid data.
> 
> However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
> as in commit:
> 
>   b22cbe404a9c x86/fpu: Fix invalid FPU ptrace state after execve()
> 
> and when __fpu_restore_sig() does fpu__restore() for a COMPAT-mode
> app, a #GP occurs.  This can be easily triggered by doing valgrind on
> a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
> others:
> 
> 	https://bugzilla.kernel.org/show_bug.cgi?id=190061
> 
> Fix it by setting xcomp_bv correctly.
> 

Please consider pulling this into the stable tree because my previous
patch in commit: 

  b22cbe404a9c x86/fpu: Fix invalid FPU ptrace state after execve()

exposes the same issue in the stable tree.

Thanks,
Yu-cheng

Re: [tip:x86/urgent] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Greg Kroah-Hartman]

On Tue, Feb 14, 2017 at 01:26:41PM -0800, Yu-cheng Yu wrote:
> On Tue, Jan 24, 2017 at 12:46:17AM -0800, tip-bot for Yu-cheng Yu wrote:
> > Commit-ID:  dffba9a31c7769be3231c420d4b364c92ba3f1ac
> > Gitweb:     http://git.kernel.org/tip/dffba9a31c7769be3231c420d4b364c92ba3f1ac
> > Author:     Yu-cheng Yu <yu-cheng.yu@intel.com>
> > AuthorDate: Mon, 23 Jan 2017 14:54:44 -0800
> > Committer:  Ingo Molnar <mingo@kernel.org>
> > CommitDate: Tue, 24 Jan 2017 09:04:48 +0100
> > 
> > x86/fpu/xstate: Fix xcomp_bv in XSAVES header
> > 
> > The compacted-format XSAVES area is determined at boot time and
> > never changed after.  The field xsave.header.xcomp_bv indicates
> > which components are in the fixed XSAVES format.
> > 
> > In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
> > format since at the time there is no valid data.
> > 
> > However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
> > as in commit:
> > 
> >   b22cbe404a9c x86/fpu: Fix invalid FPU ptrace state after execve()
> > 
> > and when __fpu_restore_sig() does fpu__restore() for a COMPAT-mode
> > app, a #GP occurs.  This can be easily triggered by doing valgrind on
> > a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
> > others:
> > 
> > 	https://bugzilla.kernel.org/show_bug.cgi?id=190061
> > 
> > Fix it by setting xcomp_bv correctly.
> > 
> 
> Please consider pulling this into the stable tree because my previous
> patch in commit: 
> 
>   b22cbe404a9c x86/fpu: Fix invalid FPU ptrace state after execve()
> 
> exposes the same issue in the stable tree.

Ah, thanks, I'll do that for the next round of stable releases after the
ones that happen tomorrow.

greg k-h

Re: [tip:x86/urgent] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Greg Kroah-Hartman]

On Tue, Feb 14, 2017 at 01:26:41PM -0800, Yu-cheng Yu wrote:
> On Tue, Jan 24, 2017 at 12:46:17AM -0800, tip-bot for Yu-cheng Yu wrote:
> > Commit-ID:  dffba9a31c7769be3231c420d4b364c92ba3f1ac
> > Gitweb:     http://git.kernel.org/tip/dffba9a31c7769be3231c420d4b364c92ba3f1ac
> > Author:     Yu-cheng Yu <yu-cheng.yu@intel.com>
> > AuthorDate: Mon, 23 Jan 2017 14:54:44 -0800
> > Committer:  Ingo Molnar <mingo@kernel.org>
> > CommitDate: Tue, 24 Jan 2017 09:04:48 +0100
> > 
> > x86/fpu/xstate: Fix xcomp_bv in XSAVES header
> > 
> > The compacted-format XSAVES area is determined at boot time and
> > never changed after.  The field xsave.header.xcomp_bv indicates
> > which components are in the fixed XSAVES format.
> > 
> > In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
> > format since at the time there is no valid data.
> > 
> > However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
> > as in commit:
> > 
> >   b22cbe404a9c x86/fpu: Fix invalid FPU ptrace state after execve()
> > 
> > and when __fpu_restore_sig() does fpu__restore() for a COMPAT-mode
> > app, a #GP occurs.  This can be easily triggered by doing valgrind on
> > a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
> > others:
> > 
> > 	https://bugzilla.kernel.org/show_bug.cgi?id=190061
> > 
> > Fix it by setting xcomp_bv correctly.
> > 
> 
> Please consider pulling this into the stable tree because my previous
> patch in commit: 
> 
>   b22cbe404a9c x86/fpu: Fix invalid FPU ptrace state after execve()
> 
> exposes the same issue in the stable tree.

I just tried to apply this, and it does not apply at all.  Can you
please provide a backported version that you have verified will work, so
that I can apply it?

thanks,

greg k-h

Re: [tip:x86/urgent] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Yu-cheng Yu]

On Thu, Feb 16, 2017 at 09:51:12AM -0800, Greg Kroah-Hartman wrote:
> On Tue, Feb 14, 2017 at 01:26:41PM -0800, Yu-cheng Yu wrote:
> > On Tue, Jan 24, 2017 at 12:46:17AM -0800, tip-bot for Yu-cheng Yu wrote:
> > > Commit-ID:  dffba9a31c7769be3231c420d4b364c92ba3f1ac
> > > Gitweb:     http://git.kernel.org/tip/dffba9a31c7769be3231c420d4b364c92ba3f1ac
> > > Author:     Yu-cheng Yu <yu-cheng.yu@intel.com>
> > > AuthorDate: Mon, 23 Jan 2017 14:54:44 -0800
> > > Committer:  Ingo Molnar <mingo@kernel.org>
> > > CommitDate: Tue, 24 Jan 2017 09:04:48 +0100
> > > 
> > > x86/fpu/xstate: Fix xcomp_bv in XSAVES header

> I just tried to apply this, and it does not apply at all.  Can you
> please provide a backported version that you have verified will work, so
> that I can apply it?

I will do that.

Thanks,
Yu-cheng

[PATCH] x86/fpu/xstate: Fix xcomp_bv in XSAVES header

[Yu-cheng Yu]

This is essentially the same patch as:

  commit dffba9a31c77 ("x86/fpu/xstate: Fix xcomp_bv in XSAVES header")

in the upstream tree, rebased and tested for the stable tree.
This can be applied to both linux-stable-rc/linux-4.9.y and
linux-stable-rc/linux-4.8.y if desired. 

The compacted-format XSAVES area is determined at boot time and
never changed after.  The field xsave.header.xcomp_bv indicates
which components are in the fixed XSAVES format.

In fpstate_init() we did not set xcomp_bv to reflect the XSAVES
format since at the time there is no valid data.

However, after we do copy_init_fpstate_to_fpregs() in fpu__clear(),
as in commit

  b22cbe404a9c ("x86/fpu: Fix invalid FPU ptrace state after execve()")

and when __fpu_restore_sig() does fpu__restore() for a COMPAT-mode
app, a #GP occurs.  This can be easily triggered by doing valgrind on
a COMPAT-mode "Hello World," as reported by Joakim Tjernlund and
others:

	https://bugzilla.kernel.org/show_bug.cgi?id=190061

Fix it by setting xcomp_bv correctly.

Signed-off-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
Reported-by: Joakim Tjernlund <Joakim.Tjernlund@infinera.com>
---
 arch/x86/kernel/fpu/core.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/fpu/core.c b/arch/x86/kernel/fpu/core.c
index ebb4e95..96d80df 100644
--- a/arch/x86/kernel/fpu/core.c
+++ b/arch/x86/kernel/fpu/core.c
@@ -236,7 +236,8 @@ void fpstate_init(union fpregs_state *state)
 	 * it will #GP. Make sure it is replaced after the memset().
 	 */
 	if (static_cpu_has(X86_FEATURE_XSAVES))
-		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT;
+		state->xsave.header.xcomp_bv = XCOMP_BV_COMPACTED_FORMAT |
+					       xfeatures_mask;
 
 	if (static_cpu_has(X86_FEATURE_FXSR))
 		fpstate_init_fxstate(&state->fxsave);
-- 
1.9.1