Skip to contentSkip to conversation

Mention PowerPC in the L1TF documentation.

PATCH In review [email protected] 3 participants 5 messages Nov 14, 2019 – Nov 15, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Mention PowerPC in the L1TF documentation.

From: Anthony Steinhauser <[email protected]>

There is a false negative that L1TF is Intel specific while it affects also PowerPC:
https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?id=8e6b6da91ac9b9ec5a925b6cb13f287a54bd547d
Another false negative is that the kernel is unconditionally protected against L1TF attacks from userspace. That's true only on x86 but not on PowerPC where you can turn the protection off by nopti.

---
 Documentation/admin-guide/hw-vuln/l1tf.rst | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
Documentation/admin-guide/hw-vuln/l1tf.rst +9 -6
index f83212fae4d5..243e494b337a 100644
@@ -9,10 +9,11 @@ for the access, has the Present bit cleared or other reserved bits set.
 Affected processors
 -------------------
 
-This vulnerability affects a wide range of Intel processors. The
-vulnerability is not present on:
+This vulnerability affects a wide range of Intel and PowerPC processors.
+The vulnerability is not present on:
 
-   - Processors from AMD, Centaur and other non Intel vendors
+   - Processors from AMD, Centaur and other non Intel vendors except for
+     PowerPC
 
    - Older processor models, where the CPU family is < 6
 
@@ -125,7 +126,7 @@ mitigations are active. The relevant sysfs file is:
 
 /sys/devices/system/cpu/vulnerabilities/l1tf
 
-The possible values in this file are:
+The possible values in this file on x86 are:
 
   ===========================   ===============================
   'Not affected'		The processor is not vulnerable
@@ -158,8 +159,10 @@ The resulting grade of protection is discussed in the following sections.
 Host mitigation mechanism
 -------------------------
 
-The kernel is unconditionally protected against L1TF attacks from malicious
-user space running on the host.
+On x86 the kernel is unconditionally protected against L1TF attacks from
+malicious user space running on the host. On PowerPC the kernel is
+protected by flushing the L1D cache on each transition from kernel to
+userspace unless the 'nopti' boot argument turns this mitigation off.
 
 
 Guest mitigation mechanisms
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
#1 lore raw

Anthony,

Quoted 5 lines · On Thu, 14 Nov 2019, [email protected] wrote:

From: Anthony Steinhauser <[email protected]>

There is a false negative that L1TF is Intel specific while it affects also PowerPC:
https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?id=8e6b6da91ac9b9ec5a925b6cb13f287a54bd547d

Please use the regular

commit 12-char-sha ("powerpc: .......")

notation. These links are horrible.

Quoted 3 lines

Another false negative is that the kernel is unconditionally protected against L1TF attacks from userspace. That's true only on x86 but not on PowerPC where you can turn the protection off by nopti.

Missing newline between body and SOB

Quoted 11 lines
---
 Documentation/admin-guide/hw-vuln/l1tf.rst | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
index f83212fae4d5..243e494b337a 100644
@@ -9,10 +9,11 @@ for the access, has the Present bit cleared or other reserved bits set.
 Affected processors

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 16

Quoted 10 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -3
 -------------------
 
-This vulnerability affects a wide range of Intel processors. The
-vulnerability is not present on:
+This vulnerability affects a wide range of Intel and PowerPC processors.
+The vulnerability is not present on:
 
-   - Processors from AMD, Centaur and other non Intel vendors
+   - Processors from AMD, Centaur and other non Intel vendors except for
+     PowerPC

No. This needs restructuring. The non Intel vendor means vendors which produce x86 machines (not really clear TBH), but adding these two PPC parts above and here does not make it any better.

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 18

Quoted 1 line
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
    - Older processor models, where the CPU family is < 6

Also this suggest that _ALL_ PowerPC processors are affected as there is no exception list.

So I suggest to structure this like this:

Affected processors
-------------------

1) Intel processors

Move the existing list here

2) PowerPC processors

Add some meat

Whether a processor is affected or not...

Documentation/admin-guide/hw-vuln/l1tf.rst mitigations are active. The relevant sysfs file is: · line 129

Quoted 6 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +1 -1
@@ -125,7 +126,7 @@ mitigations are active. The relevant sysfs file is:
 
 /sys/devices/system/cpu/vulnerabilities/l1tf
 
-The possible values in this file are:
+The possible values in this file on x86 are:

That commit you referenced added the sysfs output for powerpc. So that should be documented properly here as well, right?

Documentation/admin-guide/hw-vuln/l1tf.rst The resulting grade of protection is discussed in the following sections. · line 165

Quoted 12 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -2
   ===========================   ===============================
   'Not affected'		The processor is not vulnerable
@@ -158,8 +159,10 @@ The resulting grade of protection is discussed in the following sections.
 Host mitigation mechanism
 -------------------------
 
-The kernel is unconditionally protected against L1TF attacks from malicious
-user space running on the host.
+On x86 the kernel is unconditionally protected against L1TF attacks from
+malicious user space running on the host. On PowerPC the kernel is
+protected by flushing the L1D cache on each transition from kernel to
+userspace unless the 'nopti' boot argument turns this mitigation off.

Please make this clearly visually separated. Just glueing this together is hard to read.

What about the l1tf boot param? Is it x86 only? If so, then this wants to be added to the the documentation as well.

What about the guest to host issue on PPC? Not affected or same mitigation or what?

We really spent a lot of time to write understandable and useful documentation. Just sprinkling a few powerpc'isms into it is really not cutting it.

This needs proper structuring so that it's obvious for the intended audience (administrators, users) which part is applicable to which architecture or to both.

Thanks,

tglx

#2 lore raw

OK, I don't intend to work on it to that extent, so you can consider this abandoned.

Quoted 112 lines · Thomas Gleixner

Anthony,

Quoted 5 lines · On Thu, 14 Nov 2019, [email protected] wrote:

From: Anthony Steinhauser <[email protected]>

There is a false negative that L1TF is Intel specific while it affects also PowerPC:
https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?id=8e6b6da91ac9b9ec5a925b6cb13f287a54bd547d

Please use the regular

   commit 12-char-sha ("powerpc: .......")

notation. These links are horrible.

Quoted 3 lines

Another false negative is that the kernel is unconditionally protected against L1TF attacks from userspace. That's true only on x86 but not on PowerPC where you can turn the protection off by nopti.

Missing newline between body and SOB

Quoted 11 lines
---
 Documentation/admin-guide/hw-vuln/l1tf.rst | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
index f83212fae4d5..243e494b337a 100644
@@ -9,10 +9,11 @@ for the access, has the Present bit cleared or other reserved bits set.
 Affected processors

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 16

Quoted 10 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -3
 -------------------
 
-This vulnerability affects a wide range of Intel processors. The
-vulnerability is not present on:
+This vulnerability affects a wide range of Intel and PowerPC processors.
+The vulnerability is not present on:
 
-   - Processors from AMD, Centaur and other non Intel vendors
+   - Processors from AMD, Centaur and other non Intel vendors except for
+     PowerPC

No. This needs restructuring. The non Intel vendor means vendors which produce x86 machines (not really clear TBH), but adding these two PPC parts above and here does not make it any better.

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 18

Quoted 1 line
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
    - Older processor models, where the CPU family is < 6

Also this suggest that _ALL_ PowerPC processors are affected as there is no exception list.

So I suggest to structure this like this:

Affected processors
-------------------

1) Intel processors

Move the existing list here

2) PowerPC processors

Add some meat

Whether a processor is affected or not...

Documentation/admin-guide/hw-vuln/l1tf.rst mitigations are active. The relevant sysfs file is: · line 129

Quoted 6 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +1 -1
@@ -125,7 +126,7 @@ mitigations are active. The relevant sysfs file is:
 
 /sys/devices/system/cpu/vulnerabilities/l1tf
 
-The possible values in this file are:
+The possible values in this file on x86 are:

That commit you referenced added the sysfs output for powerpc. So that should be documented properly here as well, right?

Documentation/admin-guide/hw-vuln/l1tf.rst The resulting grade of protection is discussed in the following sections. · line 165

Quoted 12 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -2
   ===========================   ===============================
   'Not affected'             The processor is not vulnerable
@@ -158,8 +159,10 @@ The resulting grade of protection is discussed in the following sections.
 Host mitigation mechanism
 -------------------------
 
-The kernel is unconditionally protected against L1TF attacks from malicious
-user space running on the host.
+On x86 the kernel is unconditionally protected against L1TF attacks from
+malicious user space running on the host. On PowerPC the kernel is
+protected by flushing the L1D cache on each transition from kernel to
+userspace unless the 'nopti' boot argument turns this mitigation off.

Please make this clearly visually separated. Just glueing this together is hard to read.

What about the l1tf boot param? Is it x86 only? If so, then this wants to be added to the the documentation as well.

What about the guest to host issue on PPC? Not affected or same mitigation or what?

We really spent a lot of time to write understandable and useful documentation. Just sprinkling a few powerpc'isms into it is really not cutting it.

This needs proper structuring so that it's obvious for the intended audience (administrators, users) which part is applicable to which architecture or to both.

Thanks,

tglx

#3 lore raw
Quoted 2 lines · Anthony Steinhauser

OK, I don't intend to work on it to that extent, so you can consider this abandoned.

I or someone else at IBM will pick this up and get it massaged into a form that Thomas is happy with.

cheers

Quoted 113 lines · Thomas Gleixner
Quoted 112 lines

Anthony,

Quoted 5 lines · On Thu, 14 Nov 2019, [email protected] wrote:

From: Anthony Steinhauser <[email protected]>

There is a false negative that L1TF is Intel specific while it affects also PowerPC:
https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?id=8e6b6da91ac9b9ec5a925b6cb13f287a54bd547d

Please use the regular

   commit 12-char-sha ("powerpc: .......")

notation. These links are horrible.

Quoted 3 lines

Another false negative is that the kernel is unconditionally protected against L1TF attacks from userspace. That's true only on x86 but not on PowerPC where you can turn the protection off by nopti.

Missing newline between body and SOB

Quoted 11 lines
---
 Documentation/admin-guide/hw-vuln/l1tf.rst | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
index f83212fae4d5..243e494b337a 100644
@@ -9,10 +9,11 @@ for the access, has the Present bit cleared or other reserved bits set.
 Affected processors

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 16

Quoted 10 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -3
 -------------------
 
-This vulnerability affects a wide range of Intel processors. The
-vulnerability is not present on:
+This vulnerability affects a wide range of Intel and PowerPC processors.
+The vulnerability is not present on:
 
-   - Processors from AMD, Centaur and other non Intel vendors
+   - Processors from AMD, Centaur and other non Intel vendors except for
+     PowerPC

No. This needs restructuring. The non Intel vendor means vendors which produce x86 machines (not really clear TBH), but adding these two PPC parts above and here does not make it any better.

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 18

Quoted 1 line
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
    - Older processor models, where the CPU family is < 6

Also this suggest that _ALL_ PowerPC processors are affected as there is no exception list.

So I suggest to structure this like this:

Affected processors
-------------------

1) Intel processors

Move the existing list here

2) PowerPC processors

Add some meat

Whether a processor is affected or not...

Documentation/admin-guide/hw-vuln/l1tf.rst mitigations are active. The relevant sysfs file is: · line 129

Quoted 6 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +1 -1
@@ -125,7 +126,7 @@ mitigations are active. The relevant sysfs file is:
 
 /sys/devices/system/cpu/vulnerabilities/l1tf
 
-The possible values in this file are:
+The possible values in this file on x86 are:

That commit you referenced added the sysfs output for powerpc. So that should be documented properly here as well, right?

Documentation/admin-guide/hw-vuln/l1tf.rst The resulting grade of protection is discussed in the following sections. · line 165

Quoted 12 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -2
   ===========================   ===============================
   'Not affected'             The processor is not vulnerable
@@ -158,8 +159,10 @@ The resulting grade of protection is discussed in the following sections.
 Host mitigation mechanism
 -------------------------
 
-The kernel is unconditionally protected against L1TF attacks from malicious
-user space running on the host.
+On x86 the kernel is unconditionally protected against L1TF attacks from
+malicious user space running on the host. On PowerPC the kernel is
+protected by flushing the L1D cache on each transition from kernel to
+userspace unless the 'nopti' boot argument turns this mitigation off.

Please make this clearly visually separated. Just glueing this together is hard to read.

What about the l1tf boot param? Is it x86 only? If so, then this wants to be added to the the documentation as well.

What about the guest to host issue on PPC? Not affected or same mitigation or what?

We really spent a lot of time to write understandable and useful documentation. Just sprinkling a few powerpc'isms into it is really not cutting it.

This needs proper structuring so that it's obvious for the intended audience (administrators, users) which part is applicable to which architecture or to both.

Thanks,

tglx

#4 lore raw

Thanks a lot, Michael!

Quoted 123 lines · Michael Ellerman
Quoted 2 lines · Anthony Steinhauser

OK, I don't intend to work on it to that extent, so you can consider this abandoned.

I or someone else at IBM will pick this up and get it massaged into a form that Thomas is happy with.

cheers

Quoted 113 lines · Thomas Gleixner
Quoted 112 lines

Anthony,

Quoted 5 lines · On Thu, 14 Nov 2019, [email protected] wrote:

From: Anthony Steinhauser <[email protected]>

There is a false negative that L1TF is Intel specific while it affects also PowerPC:
https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?id=8e6b6da91ac9b9ec5a925b6cb13f287a54bd547d

Please use the regular

   commit 12-char-sha ("powerpc: .......")

notation. These links are horrible.

Quoted 3 lines

Another false negative is that the kernel is unconditionally protected against L1TF attacks from userspace. That's true only on x86 but not on PowerPC where you can turn the protection off by nopti.

Missing newline between body and SOB

Quoted 11 lines
---
 Documentation/admin-guide/hw-vuln/l1tf.rst | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
index f83212fae4d5..243e494b337a 100644
@@ -9,10 +9,11 @@ for the access, has the Present bit cleared or other reserved bits set.
 Affected processors

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 16

Quoted 10 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -3
 -------------------
 
-This vulnerability affects a wide range of Intel processors. The
-vulnerability is not present on:
+This vulnerability affects a wide range of Intel and PowerPC processors.
+The vulnerability is not present on:
 
-   - Processors from AMD, Centaur and other non Intel vendors
+   - Processors from AMD, Centaur and other non Intel vendors except for
+     PowerPC

No. This needs restructuring. The non Intel vendor means vendors which produce x86 machines (not really clear TBH), but adding these two PPC parts above and here does not make it any better.

Documentation/admin-guide/hw-vuln/l1tf.rst for the access, has the Present bit cleared or other reserved bits set. · line 18

Quoted 1 line
Documentation/admin-guide/hw-vuln/l1tf.rst +0 -0
    - Older processor models, where the CPU family is < 6

Also this suggest that _ALL_ PowerPC processors are affected as there is no exception list.

So I suggest to structure this like this:

Affected processors
-------------------

1) Intel processors

Move the existing list here

2) PowerPC processors

Add some meat

Whether a processor is affected or not...

Documentation/admin-guide/hw-vuln/l1tf.rst mitigations are active. The relevant sysfs file is: · line 129

Quoted 6 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +1 -1
@@ -125,7 +126,7 @@ mitigations are active. The relevant sysfs file is:
 
 /sys/devices/system/cpu/vulnerabilities/l1tf
 
-The possible values in this file are:
+The possible values in this file on x86 are:

That commit you referenced added the sysfs output for powerpc. So that should be documented properly here as well, right?

Documentation/admin-guide/hw-vuln/l1tf.rst The resulting grade of protection is discussed in the following sections. · line 165

Quoted 12 lines
Documentation/admin-guide/hw-vuln/l1tf.rst +4 -2
   ===========================   ===============================
   'Not affected'             The processor is not vulnerable
@@ -158,8 +159,10 @@ The resulting grade of protection is discussed in the following sections.
 Host mitigation mechanism
 -------------------------
 
-The kernel is unconditionally protected against L1TF attacks from malicious
-user space running on the host.
+On x86 the kernel is unconditionally protected against L1TF attacks from
+malicious user space running on the host. On PowerPC the kernel is
+protected by flushing the L1D cache on each transition from kernel to
+userspace unless the 'nopti' boot argument turns this mitigation off.

Please make this clearly visually separated. Just glueing this together is hard to read.

What about the l1tf boot param? Is it x86 only? If so, then this wants to be added to the the documentation as well.

What about the guest to host issue on PPC? Not affected or same mitigation or what?

We really spent a lot of time to write understandable and useful documentation. Just sprinkling a few powerpc'isms into it is really not cutting it.

This needs proper structuring so that it's obvious for the intended audience (administrators, users) which part is applicable to which architecture or to both.

Thanks,

tglx