Skip to contentSkip to conversation

KVM: s390: disable migration mode when dirty tracking is disabled

PATCH Superseded Nico Boehr 3 participants 6 messages Jan 20, 2023 – Jan 26, 2023

Subsystems (4)

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Nico Boehr

author
#0 lore raw

PATCH 1/1 KVM: s390: disable migration mode when dirty tracking is disabled

Migration mode is a VM attribute which enables tracking of changes in storage attributes (PGSTE). It assumes dirty tracking is enabled on all memslots to keep a dirty bitmap of pages with changed storage attributes.

When enabling migration mode, we currently check that dirty tracking is enabled for all memslots. However, userspace can disable dirty tracking without disabling migration mode.

Since migration mode is pointless with dirty tracking disabled, disable migration mode whenever userspace disables dirty tracking on any slot.

Also update the documentation to clarify that dirty tracking must be enabled when enabling migration mode, which is already enforced by the code in kvm_s390_vm_start_migration().

To disable migration mode, slots_lock should be held, which is taken in kvm_set_memory_region() and thus held in kvm_arch_prepare_memory_region().

Restructure the prepare code a bit so all the sanity checking is done before disabling migration mode. This ensures migration mode isn't disabled when some sanity check fails.

Fixes 190df4a212a7 ("KVM: s390: CMMA tracking, ESSA emulation, migration mode")
---
 Documentation/virt/kvm/devices/vm.rst |  4 +++
 arch/s390/kvm/kvm-s390.c              | 41 ++++++++++++++++++---------
 2 files changed, 32 insertions(+), 13 deletions(-)
Documentation/virt/kvm/devices/vm.rst +4 -0
index 60acc39e0e93..147efec626e5 100644
@@ -302,6 +302,10 @@ Allows userspace to start migration mode, needed for PGSTE migration.
 Setting this attribute when migration mode is already active will have
 no effects.
 
+Dirty tracking must be enabled on all memslots, else -EINVAL is returned. When
+dirty tracking is disabled on any memslot, migration mode is automatically
+stopped.
+
 :Parameters: none
 :Returns:   -ENOMEM if there is not enough free memory to start migration mode;
 	    -EINVAL if the state of the VM is invalid (e.g. no memory defined);
arch/s390/kvm/kvm-s390.c +28 -13

56 lines not on this page · raw message

Signature · 1 line
2.39.0
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 19 lines · On Fri, 20 Jan 2023 08:54:06 +0100 Nico Boehr <[email protected]> wrote:

Migration mode is a VM attribute which enables tracking of changes in storage attributes (PGSTE). It assumes dirty tracking is enabled on all memslots to keep a dirty bitmap of pages with changed storage attributes.

When enabling migration mode, we currently check that dirty tracking is enabled for all memslots. However, userspace can disable dirty tracking without disabling migration mode.

Since migration mode is pointless with dirty tracking disabled, disable migration mode whenever userspace disables dirty tracking on any slot.

Also update the documentation to clarify that dirty tracking must be enabled when enabling migration mode, which is already enforced by the code in kvm_s390_vm_start_migration().

To disable migration mode, slots_lock should be held, which is taken in kvm_set_memory_region() and thus held in kvm_arch_prepare_memory_region().

Quoted 7 lines

Restructure the prepare code a bit so all the sanity checking is done before disabling migration mode. This ensures migration mode isn't disabled when some sanity check fails.

Fixes 190df4a212a7 ("KVM: s390: CMMA tracking, ESSA emulation, migration mode")
Quoted 81 lines
---
 Documentation/virt/kvm/devices/vm.rst |  4 +++
 arch/s390/kvm/kvm-s390.c              | 41 ++++++++++++++++++---------
 2 files changed, 32 insertions(+), 13 deletions(-)
Documentation/virt/kvm/devices/vm.rst +4 -0
index 60acc39e0e93..147efec626e5 100644
@@ -302,6 +302,10 @@ Allows userspace to start migration mode, needed for PGSTE migration.
 Setting this attribute when migration mode is already active will have
 no effects.
 
+Dirty tracking must be enabled on all memslots, else -EINVAL is returned. When
+dirty tracking is disabled on any memslot, migration mode is automatically
+stopped.
+
 :Parameters: none
 :Returns:   -ENOMEM if there is not enough free memory to start migration mode;
 	    -EINVAL if the state of the VM is invalid (e.g. no memory defined);
arch/s390/kvm/kvm-s390.c +28 -13
index e4890e04b210..4785f002cd93 100644
@@ -5628,28 +5628,43 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
 				   enum kvm_mr_change change)
 {
 	gpa_t size;
+	int rc;
 
 	/* When we are protected, we should not change the memory slots */
 	if (kvm_s390_pv_get_handle(kvm))
 		return -EINVAL;
 
-	if (change == KVM_MR_DELETE || change == KVM_MR_FLAGS_ONLY)
-		return 0;
+	if (change != KVM_MR_DELETE && change != KVM_MR_FLAGS_ONLY) {
+		/* A few sanity checks. We can have memory slots which have to be
+		 * located/ended at a segment boundary (1MB). The memory in userland is
+		 * ok to be fragmented into various different vmas. It is okay to mmap()
+		 * and munmap() stuff in this slot after doing this call at any time
+		 */
 
-	/* A few sanity checks. We can have memory slots which have to be
-	   located/ended at a segment boundary (1MB). The memory in userland is
-	   ok to be fragmented into various different vmas. It is okay to mmap()
-	   and munmap() stuff in this slot after doing this call at any time */
+		if (new->userspace_addr & 0xffffful)
+			return -EINVAL;
 
-	if (new->userspace_addr & 0xffffful)
-		return -EINVAL;
+		size = new->npages * PAGE_SIZE;
+		if (size & 0xffffful)
+			return -EINVAL;
 
-	size = new->npages * PAGE_SIZE;
-	if (size & 0xffffful)
-		return -EINVAL;
+		if ((new->base_gfn * PAGE_SIZE) + size > kvm->arch.mem_limit)
+			return -EINVAL;
+	}
 
-	if ((new->base_gfn * PAGE_SIZE) + size > kvm->arch.mem_limit)
-		return -EINVAL;
+	/* Turn off migration mode when userspace disables dirty page logging.
+	 * Migration mode expects dirty page logging being enabled to store
+	 * its dirty bitmap.
+	 */
+	if (kvm->arch.migration_mode) {
+		if ((old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&
+		    !(new->flags & KVM_MEM_LOG_DIRTY_PAGES)) {
+			rc = kvm_s390_vm_stop_migration(kvm);
+
+			if (rc)
+				pr_warn("Failed to stop migration mode\n");
+		}
+	}
 
 	return 0;
 }
maintainer
#2 lore raw
In reply to Nico Boehr
Quoted 10 lines · Nico Boehr

Migration mode is a VM attribute which enables tracking of changes in storage attributes (PGSTE). It assumes dirty tracking is enabled on all memslots to keep a dirty bitmap of pages with changed storage attributes.

When enabling migration mode, we currently check that dirty tracking is enabled for all memslots. However, userspace can disable dirty tracking without disabling migration mode.

Since migration mode is pointless with dirty tracking disabled, disable migration mode whenever userspace disables dirty tracking on any slot.

Will userspace be able to handle the sudden -EINVAL rcs on KVM_S390_GET_CMMA_BITS and KVM_S390_SET_CMMA_BITS?

I.e. what allows us to simply turn it off without the userspace knowing about it?

Quoted 26 lines

Also update the documentation to clarify that dirty tracking must be enabled when enabling migration mode, which is already enforced by the code in kvm_s390_vm_start_migration().

To disable migration mode, slots_lock should be held, which is taken in kvm_set_memory_region() and thus held in kvm_arch_prepare_memory_region().

Restructure the prepare code a bit so all the sanity checking is done before disabling migration mode. This ensures migration mode isn't disabled when some sanity check fails.

Fixes 190df4a212a7 ("KVM: s390: CMMA tracking, ESSA emulation, migration mode")
---
  Documentation/virt/kvm/devices/vm.rst |  4 +++
  arch/s390/kvm/kvm-s390.c              | 41 ++++++++++++++++++---------
  2 files changed, 32 insertions(+), 13 deletions(-)
Documentation/virt/kvm/devices/vm.rst +0 -0
index 60acc39e0e93..147efec626e5 100644

Documentation/virt/kvm/devices/vm.rst Allows userspace to start migration mode, needed for PGSTE migration. · line 307

Quoted 6 lines
Documentation/virt/kvm/devices/vm.rst +3 -0
@@ -302,6 +302,10 @@ Allows userspace to start migration mode, needed for PGSTE migration.
  Setting this attribute when migration mode is already active will have
  no effects.
 
+Dirty tracking must be enabled on all memslots, else -EINVAL is returned. When
+dirty tracking is disabled on any memslot, migration mode is automatically
+stopped.

Do we also need to add a warning to the CMMA IOCTLs?

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5631

Quoted 13 lines
Documentation/virt/kvm/devices/vm.rst +1 -0
+
  :Parameters: none
  :Returns:   -ENOMEM if there is not enough free memory to start migration mode;
  	    -EINVAL if the state of the VM is invalid (e.g. no memory defined);
arch/s390/kvm/kvm-s390.c +1 -0
index e4890e04b210..4785f002cd93 100644
@@ -5628,28 +5628,43 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
  				   enum kvm_mr_change change)
  {
  	gpa_t size;
+	int rc;

Not sure why you added rc even though it doesn't need to be used.

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5642

Quoted 13 lines
arch/s390/kvm/kvm-s390.c +6 -2
  	/* When we are protected, we should not change the memory slots */
  	if (kvm_s390_pv_get_handle(kvm))
  		return -EINVAL;
 
-	if (change == KVM_MR_DELETE || change == KVM_MR_FLAGS_ONLY)
-		return 0;
+	if (change != KVM_MR_DELETE && change != KVM_MR_FLAGS_ONLY) {
+		/* A few sanity checks. We can have memory slots which have to be
+		 * located/ended at a segment boundary (1MB). The memory in userland is
+		 * ok to be fragmented into various different vmas. It is okay to mmap()
+		 * and munmap() stuff in this slot after doing this call at any time
+		 */

This isn't net code, we usually start our comments on a "*" line.

Quoted 24 lines
arch/s390/kvm/kvm-s390.c +9 -11
-	/* A few sanity checks. We can have memory slots which have to be
-	   located/ended at a segment boundary (1MB). The memory in userland is
-	   ok to be fragmented into various different vmas. It is okay to mmap()
-	   and munmap() stuff in this slot after doing this call at any time */
+		if (new->userspace_addr & 0xffffful)
+			return -EINVAL;
 
-	if (new->userspace_addr & 0xffffful)
-		return -EINVAL;
+		size = new->npages * PAGE_SIZE;
+		if (size & 0xffffful)
+			return -EINVAL;
 
-	size = new->npages * PAGE_SIZE;
-	if (size & 0xffffful)
-		return -EINVAL;
+		if ((new->base_gfn * PAGE_SIZE) + size > kvm->arch.mem_limit)
+			return -EINVAL;
+	}
 
-	if ((new->base_gfn * PAGE_SIZE) + size > kvm->arch.mem_limit)
-		return -EINVAL;
+	/* Turn off migration mode when userspace disables dirty page logging.

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5665

Quoted 10 lines
arch/s390/kvm/kvm-s390.c +10 -0
+	 * Migration mode expects dirty page logging being enabled to store
+	 * its dirty bitmap.
+	 */
+	if (kvm->arch.migration_mode) {
+		if ((old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&
+		    !(new->flags & KVM_MEM_LOG_DIRTY_PAGES)) {
+			rc = kvm_s390_vm_stop_migration(kvm);
+
+			if (rc)
+				pr_warn("Failed to stop migration mode\n");

As the results were rather catastrophic it might make more sense to use WARN_ONCE() and condense these 3 lines into one.

Quoted 5 lines
arch/s390/kvm/kvm-s390.c +2 -0
+		}
+	}
 
  	return 0;
  }
maintainer
#3 lore raw
Quoted 9 lines · On Wed, 25 Jan 2023 14:55:59 +0100 Janosch Frank <[email protected]> wrote:
Quoted 8 lines · Nico Boehr

Migration mode is a VM attribute which enables tracking of changes in storage attributes (PGSTE). It assumes dirty tracking is enabled on all memslots to keep a dirty bitmap of pages with changed storage attributes.

When enabling migration mode, we currently check that dirty tracking is enabled for all memslots. However, userspace can disable dirty tracking without disabling migration mode.

Quoted 8 lines
Quoted 2 lines

Since migration mode is pointless with dirty tracking disabled, disable migration mode whenever userspace disables dirty tracking on any slot.

Will userspace be able to handle the sudden -EINVAL rcs on KVM_S390_GET_CMMA_BITS and KVM_S390_SET_CMMA_BITS?

I.e. what allows us to simply turn it off without the userspace knowing about it?

if we are here, userspace is disabling dirty tracking without having disabled migration mode. it should have disabled migration mode before disabling dirty tracking. also, migration mode does not actually impact userspace (at least at this point). it's just an indication from userspace to the kernel that userspace is trying to migrate. disabling dirty tracking is a rather explicit way for userspace to tell the kernel that the migration is over (one way or the other)

Quoted 57 lines
Quoted 27 lines

Also update the documentation to clarify that dirty tracking must be enabled when enabling migration mode, which is already enforced by the code in kvm_s390_vm_start_migration().

To disable migration mode, slots_lock should be held, which is taken in kvm_set_memory_region() and thus held in kvm_arch_prepare_memory_region().

Restructure the prepare code a bit so all the sanity checking is done before disabling migration mode. This ensures migration mode isn't disabled when some sanity check fails.

Fixes 190df4a212a7 ("KVM: s390: CMMA tracking, ESSA emulation, migration mode")
---
  Documentation/virt/kvm/devices/vm.rst |  4 +++
  arch/s390/kvm/kvm-s390.c              | 41 ++++++++++++++++++---------
  2 files changed, 32 insertions(+), 13 deletions(-)
Documentation/virt/kvm/devices/vm.rst +0 -0
index 60acc39e0e93..147efec626e5 100644

Documentation/virt/kvm/devices/vm.rst Allows userspace to start migration mode, needed for PGSTE migration. · line 307

Quoted 6 lines
Documentation/virt/kvm/devices/vm.rst +3 -0
@@ -302,6 +302,10 @@ Allows userspace to start migration mode, needed for PGSTE migration.
  Setting this attribute when migration mode is already active will have
  no effects.
 
+Dirty tracking must be enabled on all memslots, else -EINVAL is returned. When
+dirty tracking is disabled on any memslot, migration mode is automatically
+stopped.  

Do we also need to add a warning to the CMMA IOCTLs?

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5631

Quoted 13 lines
Documentation/virt/kvm/devices/vm.rst +1 -0
+
  :Parameters: none
  :Returns:   -ENOMEM if there is not enough free memory to start migration mode;
  	    -EINVAL if the state of the VM is invalid (e.g. no memory defined);
arch/s390/kvm/kvm-s390.c +1 -0
index e4890e04b210..4785f002cd93 100644
@@ -5628,28 +5628,43 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
  				   enum kvm_mr_change change)
  {
  	gpa_t size;
+	int rc;  

Not sure why you added rc even though it doesn't need to be used.

Quoted 5 lines
arch/s390/kvm/kvm-s390.c +0 -0
  	/* When we are protected, we should not change the memory slots */
  	if (kvm_s390_pv_get_handle(kvm))
  		return -EINVAL;
 
Quoted 10 lines

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5642

Quoted 8 lines
arch/s390/kvm/kvm-s390.c +6 -2
-	if (change == KVM_MR_DELETE || change == KVM_MR_FLAGS_ONLY)
-		return 0;
+	if (change != KVM_MR_DELETE && change != KVM_MR_FLAGS_ONLY) {
+		/* A few sanity checks. We can have memory slots which have to be
+		 * located/ended at a segment boundary (1MB). The memory in userland is
+		 * ok to be fragmented into various different vmas. It is okay to mmap()
+		 * and munmap() stuff in this slot after doing this call at any time
+		 */  

This isn't net code, we usually start our comments on a "*" line.

like this:

/*
* blah
*/

I missed this when I reviewed the patch

Quoted 22 lines
Quoted 22 lines
arch/s390/kvm/kvm-s390.c +8 -9
-	/* A few sanity checks. We can have memory slots which have to be
-	   located/ended at a segment boundary (1MB). The memory in userland is
-	   ok to be fragmented into various different vmas. It is okay to mmap()
-	   and munmap() stuff in this slot after doing this call at any time */
+		if (new->userspace_addr & 0xffffful)
+			return -EINVAL;
 
-	if (new->userspace_addr & 0xffffful)
-		return -EINVAL;
+		size = new->npages * PAGE_SIZE;
+		if (size & 0xffffful)
+			return -EINVAL;
 
-	size = new->npages * PAGE_SIZE;
-	if (size & 0xffffful)
-		return -EINVAL;
+		if ((new->base_gfn * PAGE_SIZE) + size > kvm->arch.mem_limit)
+			return -EINVAL;
+	}
 
Quoted 16 lines

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5665

Quoted 13 lines
arch/s390/kvm/kvm-s390.c +11 -2
-	if ((new->base_gfn * PAGE_SIZE) + size > kvm->arch.mem_limit)
-		return -EINVAL;
+	/* Turn off migration mode when userspace disables dirty page logging.
+	 * Migration mode expects dirty page logging being enabled to store
+	 * its dirty bitmap.
+	 */
+	if (kvm->arch.migration_mode) {
+		if ((old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&
+		    !(new->flags & KVM_MEM_LOG_DIRTY_PAGES)) {
+			rc = kvm_s390_vm_stop_migration(kvm);
+
+			if (rc)
+				pr_warn("Failed to stop migration mode\n");  

As the results were rather catastrophic it might make more sense to use WARN_ONCE() and condense these 3 lines into one.

(in which case rc is not needed)

is WARN_ONCE even enough? the results are indeed potentially catastrophic, would it not make sense to WARN (without _ONCE) ?

Quoted 7 lines
Quoted 6 lines
arch/s390/kvm/kvm-s390.c +2 -0
+		}
+	}
 
  	return 0;
  }  
author
#4 lore raw
In reply to Janosch Frank

Quoting Janosch Frank (2023-01-25 14:55:59)

Quoted 9 lines · Nico Boehr
Quoted 8 lines

Migration mode is a VM attribute which enables tracking of changes in storage attributes (PGSTE). It assumes dirty tracking is enabled on all memslots to keep a dirty bitmap of pages with changed storage attributes.

When enabling migration mode, we currently check that dirty tracking is enabled for all memslots. However, userspace can disable dirty tracking without disabling migration mode.

Quoted 5 lines
Quoted 2 lines

Since migration mode is pointless with dirty tracking disabled, disable migration mode whenever userspace disables dirty tracking on any slot.

Will userspace be able to handle the sudden -EINVAL rcs on KVM_S390_GET_CMMA_BITS and KVM_S390_SET_CMMA_BITS?

QEMU has proper error handling on the GET_CMMA_BITS code path and will not attempt GET_CMMA_BITS after it disabled dirty tracking. So yes, userspace can handle this fine. In addition, as mentioned in the commit, it was never allowed to have migration mode without dirty tracking. It was checked when migration mode is enabled, just wasn't enforced when dirty tracking went off. The alternative would be to refuse disabling dirty tracking when migration mode is on; and that would _really_ break userspace.

Or we just leave migration mode on and check on every emulation/ioctl that a dirty bitmap is still there, which would change absolutely nothing about the return value of GET_CMMA_BITS.

Or we allocate the dirty bitmap for storage attributes independent of the dirty bitmap for pages, which increases memory usage and makes this patch quite a bit more complex, risking that we break more than what is already broken.

This approach really seems like the sane option to me.

For SET_CMMA_BIT, nothing changes.

Quoted 32 lines

I.e. what allows us to simply turn it off without the userspace knowing about it?

Quoted 29 lines

Also update the documentation to clarify that dirty tracking must be enabled when enabling migration mode, which is already enforced by the code in kvm_s390_vm_start_migration().

To disable migration mode, slots_lock should be held, which is taken in kvm_set_memory_region() and thus held in kvm_arch_prepare_memory_region().

Restructure the prepare code a bit so all the sanity checking is done before disabling migration mode. This ensures migration mode isn't disabled when some sanity check fails.

Fixes 190df4a212a7 ("KVM: s390: CMMA tracking, ESSA emulation, migration mode")
---
  Documentation/virt/kvm/devices/vm.rst |  4 +++
  arch/s390/kvm/kvm-s390.c              | 41 ++++++++++++++++++---------
  2 files changed, 32 insertions(+), 13 deletions(-)
Documentation/virt/kvm/devices/vm.rst +0 -0
index 60acc39e0e93..147efec626e5 100644
@@ -302,6 +302,10 @@ Allows userspace to start migration mode, needed for PGSTE migration.
  Setting this attribute when migration mode is already active will have
  no effects.
 
Quoted 5 lines

Documentation/virt/kvm/devices/vm.rst Allows userspace to start migration mode, needed for PGSTE migration. · line 307

Quoted 3 lines
Documentation/virt/kvm/devices/vm.rst +3 -0
+Dirty tracking must be enabled on all memslots, else -EINVAL is returned. When
+dirty tracking is disabled on any memslot, migration mode is automatically
+stopped.

Do we also need to add a warning to the CMMA IOCTLs?

No, it is already documented there:

Quoted 2 lines

This ioctl can fail with [...] > -EINVAL if KVM_S390_CMMA_PEEK is not set but migration mode was not enabled

[...]

Quoted 11 lines

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5631

Quoted 9 lines
arch/s390/kvm/kvm-s390.c +1 -0
index e4890e04b210..4785f002cd93 100644
@@ -5628,28 +5628,43 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
                                 enum kvm_mr_change change)
  {
      gpa_t size;
+     int rc;

Not sure why you added rc even though it doesn't need to be used.

You prefer a line which is 100 chars wide over a new variable? OK fine for me.

maintainer
#5 lore raw
Quoted 28 lines · Nico Boehr

Quoting Janosch Frank (2023-01-25 14:55:59)

Quoted 9 lines · Nico Boehr
Quoted 8 lines

Migration mode is a VM attribute which enables tracking of changes in storage attributes (PGSTE). It assumes dirty tracking is enabled on all memslots to keep a dirty bitmap of pages with changed storage attributes.

When enabling migration mode, we currently check that dirty tracking is enabled for all memslots. However, userspace can disable dirty tracking without disabling migration mode.

Quoted 5 lines
Quoted 2 lines

Since migration mode is pointless with dirty tracking disabled, disable migration mode whenever userspace disables dirty tracking on any slot.

Will userspace be able to handle the sudden -EINVAL rcs on KVM_S390_GET_CMMA_BITS and KVM_S390_SET_CMMA_BITS?

QEMU has proper error handling on the GET_CMMA_BITS code path and will not attempt GET_CMMA_BITS after it disabled dirty tracking. So yes, userspace can handle this fine. In addition, as mentioned in the commit, it was never allowed to have migration mode without dirty tracking. It was checked when migration mode is enabled, just wasn't enforced when dirty tracking went off. The alternative would be to refuse disabling dirty tracking when migration mode is on; and that would _really_ break userspace.

Or we just leave migration mode on and check on every emulation/ioctl that a dirty bitmap is still there, which would change absolutely nothing about the return value of GET_CMMA_BITS.

Quoted 5 lines

Or we allocate the dirty bitmap for storage attributes independent of the dirty bitmap for pages, which increases memory usage and makes this patch quite a bit more complex, risking that we break more than what is already broken.

This approach really seems like the sane option to me.

Jup, I'm just trying to consider all possibilities to find the best one and that includes asking all the questions.

Quoted 11 lines
Quoted 6 lines

Documentation/virt/kvm/devices/vm.rst Allows userspace to start migration mode, needed for PGSTE migration. · line 307

Quoted 4 lines
Documentation/virt/kvm/devices/vm.rst +3 -0
+Dirty tracking must be enabled on all memslots, else -EINVAL is returned. When
+dirty tracking is disabled on any memslot, migration mode is automatically
+stopped.

Do we also need to add a warning to the CMMA IOCTLs?

No, it is already documented there:

Quoted 2 lines

This ioctl can fail with [...] > -EINVAL if KVM_S390_CMMA_PEEK is not set but migration mode was not enabled

That's fine but doesn't include the part where migration mode can suddenly be disabled via a memory region change.

If we implicitly change migration mode disablement then we need to document that as much as possible to cover all bases.

Quoted 14 lines

[...]

Quoted 11 lines

arch/s390/kvm/kvm-s390.c int kvm_arch_prepare_memory_region(struct kvm *kvm, · line 5631

Quoted 9 lines
arch/s390/kvm/kvm-s390.c +1 -0
index e4890e04b210..4785f002cd93 100644
@@ -5628,28 +5628,43 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
                                  enum kvm_mr_change change)
   {
       gpa_t size;
+     int rc;

Not sure why you added rc even though it doesn't need to be used.

Quoted 1 line

You prefer a line which is 100 chars wide over a new variable? OK fine for me.

I'm not sure how you manage to get over 100 chars, did I miss something?

if (!kvm->arch.migration_mode)

	return 0;

if ((old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&

    !(new->flags & KVM_MEM_LOG_DIRTY_PAGES)) {
	WARN(kvm_s390_vm_stop_migration(kvm),
	     "Migration mode could not be disabled");

return 0;