[v2] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

[PATCH v2] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

Cc: <stable@vger.kernel.org>
Signed-off-by:  Michael Zhivich <mzhivich@akamai.com>
Fixes: 7a0395f6607a5 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v2:
- Rework patch per feedback
- Add Cc: stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index efd42ee9d1cc..289ff197b1b3 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -378,6 +378,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved    = 0;
 
 	if (cpu_has(c, X86_FEATURE_ZEN3) ||
 	    cpu_has(c, X86_FEATURE_ZEN4)) {
-- 
2.34.1

Re: [PATCH v2] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Borislav Petkov 2 months, 2 weeks ago

On Tue, Jul 22, 2025 at 08:28:44AM -0400, Michael Zhivich wrote:
> For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
> field in zen_patch_rev union on the stack may be garbage.  If so, it will
> prevent correct microcode check when consulting p.ucode_rev, resulting in
> incorrect mitigation selection.

"This is a stable-only fix." so that the AI is happy. :-P

> Cc: <stable@vger.kernel.org>
> Signed-off-by:  Michael Zhivich <mzhivich@akamai.com>

Acked-by: Borislav Petkov (AMD) <bp@alien8.de>

> Fixes: 7a0395f6607a5 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")

That commit in Fixes: is the 6.12 stable one.

The 6.6 one is:

Fixes: 90293047df18 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")

The 6.1 is:

Fixes: d12145e8454f ("x86/bugs: Add a Transient Scheduler Attacks mitigation")

The 5.15 one:

Fixes: f2b75f1368af ("x86/bugs: Add a Transient Scheduler Attacks mitigation")

and the 5.10 one is

Fixes: 78192f511f40 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")

and since all stable kernels above have INIT_STACK_NONE, that same
one-liner should be applied to all of them.

Greg, I'm thinking this one-liner should apply to all of the above with
some fuzz. Can you simply add it to each stable version with a different
Fixes: tag each?

Or do you prefer separate submissions?

Thx.

>  arch/x86/kernel/cpu/amd.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
> index efd42ee9d1cc..289ff197b1b3 100644
> --- a/arch/x86/kernel/cpu/amd.c
> +++ b/arch/x86/kernel/cpu/amd.c
> @@ -378,6 +378,8 @@ static bool amd_check_tsa_microcode(void)
>  	p.model		= c->x86_model;
>  	p.ext_model	= c->x86_model >> 4;
>  	p.stepping	= c->x86_stepping;
> +	/* reserved bits are expected to be 0 in test below */
> +	p.__reserved    = 0;
>  
>  	if (cpu_has(c, X86_FEATURE_ZEN3) ||
>  	    cpu_has(c, X86_FEATURE_ZEN4)) {
> -- 

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH v2] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Greg Kroah-Hartman 2 months, 2 weeks ago

On Tue, Jul 22, 2025 at 04:22:54PM +0200, Borislav Petkov wrote:
> On Tue, Jul 22, 2025 at 08:28:44AM -0400, Michael Zhivich wrote:
> > For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
> > field in zen_patch_rev union on the stack may be garbage.  If so, it will
> > prevent correct microcode check when consulting p.ucode_rev, resulting in
> > incorrect mitigation selection.
> 
> "This is a stable-only fix." so that the AI is happy. :-P
> 
> > Cc: <stable@vger.kernel.org>
> > Signed-off-by:  Michael Zhivich <mzhivich@akamai.com>
> 
> Acked-by: Borislav Petkov (AMD) <bp@alien8.de>
> 
> > Fixes: 7a0395f6607a5 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
> 
> That commit in Fixes: is the 6.12 stable one.
> 
> The 6.6 one is:
> 
> Fixes: 90293047df18 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
> 
> The 6.1 is:
> 
> Fixes: d12145e8454f ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
> 
> The 5.15 one:
> 
> Fixes: f2b75f1368af ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
> 
> and the 5.10 one is
> 
> Fixes: 78192f511f40 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
> 
> and since all stable kernels above have INIT_STACK_NONE, that same
> one-liner should be applied to all of them.
> 
> Greg, I'm thinking this one-liner should apply to all of the above with
> some fuzz. Can you simply add it to each stable version with a different
> Fixes: tag each?
> 
> Or do you prefer separate submissions?

Ideally, separate submissions, otherwise I have to do this all by hand
:(

thanks

greg k-h

[PATCH v3 6.6] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: d12145e8454f ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 4785d41558d6..2d71c329b347 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -563,6 +563,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (c->x86 == 0x19) {
 		switch (p.ucode_rev >> 8) {
-- 
2.34.1

[PATCH v3 6.1] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: d12145e8454f ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 4785d41558d6..2d71c329b347 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -563,6 +563,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (c->x86 == 0x19) {
 		switch (p.ucode_rev >> 8) {
-- 
2.34.1

[PATCH v3 5.15] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: f2b75f1368af ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 04ac18ff022f..3c7d64c454b3 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -592,6 +592,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (c->x86 == 0x19) {
 		switch (p.ucode_rev >> 8) {
-- 
2.34.1

[PATCH v3 6.6] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: 90293047df18 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 5fcdfbb792bd..b5a234eef471 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -549,6 +549,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (cpu_has(c, X86_FEATURE_ZEN3) ||
 	    cpu_has(c, X86_FEATURE_ZEN4)) {
-- 
2.34.1

[PATCH v3 5.10] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: 78192f511f40 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index e67d7603449b..bf07b2c5418a 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -599,6 +599,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (c->x86 == 0x19) {
 		switch (p.ucode_rev >> 8) {
-- 
2.34.1

[PATCH v3 6.12] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: 7a0395f6607a5 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index efd42ee9d1cc..289ff197b1b3 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -378,6 +378,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (cpu_has(c, X86_FEATURE_ZEN3) ||
 	    cpu_has(c, X86_FEATURE_ZEN4)) {
-- 
2.34.1

[PATCH v3 6.6] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: f2b75f1368af ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 04ac18ff022f..3c7d64c454b3 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -592,6 +592,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (c->x86 == 0x19) {
 		switch (p.ucode_rev >> 8) {
-- 
2.34.1

[PATCH v3 6.6] x86/bugs: Fix use of possibly uninit value in amd_check_tsa_microcode()

Posted by Michael Zhivich 2 months, 2 weeks ago

For kernels compiled with CONFIG_INIT_STACK_NONE=y, the value of __reserved
field in zen_patch_rev union on the stack may be garbage.  If so, it will
prevent correct microcode check when consulting p.ucode_rev, resulting in
incorrect mitigation selection.

This is a stable-only fix.

Cc: <stable@vger.kernel.org>
Signed-off-by: Michael Zhivich <mzhivich@akamai.com>
Fixes: 78192f511f40 ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
---

Changes in v3:
- separate "fixes" tag for each stable

 arch/x86/kernel/cpu/amd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index e67d7603449b..bf07b2c5418a 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -599,6 +599,8 @@ static bool amd_check_tsa_microcode(void)
 	p.model		= c->x86_model;
 	p.ext_model	= c->x86_model >> 4;
 	p.stepping	= c->x86_stepping;
+	/* reserved bits are expected to be 0 in test below */
+	p.__reserved	= 0;
 
 	if (c->x86 == 0x19) {
 		switch (p.ucode_rev >> 8) {
-- 
2.34.1