Skip to contentSkip to conversation

Fix userfaultfd_api to return EINVAL as expected

PATCH Superseded Audra Mitchell 3 participants 14 messages Jun 21, 2024 – Jun 26, 2024

Subsystems (4)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/3 Fix userfaultfd_api to return EINVAL as expected

Currently if we request a feature that is not set in the Kernel config we fail silently and return all the available features. However, the man page indicates we should return an EINVAL.

We need to fix this issue since we can end up with a Kernel warning should a program request the feature UFFD_FEATURE_WP_UNPOPULATED on a kernel with the config not set with this feature.

 [  200.812896] WARNING: CPU: 91 PID: 13634 at mm/memory.c:1660 zap_pte_range+0x43d/0x660
 [  200.820738] Modules linked in:
 [  200.869387] CPU: 91 PID: 13634 Comm: userfaultfd Kdump: loaded Not tainted 6.9.0-rc5+ #8
 [  200.877477] Hardware name: Dell Inc. PowerEdge R6525/0N7YGH, BIOS 2.7.3 03/30/2022
 [  200.885052] RIP: 0010:zap_pte_range+0x43d/0x660
Fixes e06f1e1dd499 ("userfaultfd: wp: enabled write protection in userfaultfd API")
---
 fs/userfaultfd.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
fs/userfaultfd.c +6 -1
index eee7320ab0b0..17e409ceaa33 100644
@@ -2057,7 +2057,7 @@ static int userfaultfd_api(struct userfaultfd_ctx *ctx,
 		goto out;
 	features = uffdio_api.features;
 	ret = -EINVAL;
-	if (uffdio_api.api != UFFD_API || (features & ~UFFD_API_FEATURES))
+	if (uffdio_api.api != UFFD_API)
 		goto err_out;
 	ret = -EPERM;
 	if ((features & UFFD_FEATURE_EVENT_FORK) && !capable(CAP_SYS_PTRACE))
@@ -2081,6 +2081,11 @@ static int userfaultfd_api(struct userfaultfd_ctx *ctx,
 	uffdio_api.features &= ~UFFD_FEATURE_WP_UNPOPULATED;
 	uffdio_api.features &= ~UFFD_FEATURE_WP_ASYNC;
 #endif
+
+	ret = -EINVAL;
+	if (features & ~uffdio_api.features)
+		goto err_out;
+
 	uffdio_api.ioctls = UFFD_API_IOCTLS;
 	ret = -EFAULT;
 	if (copy_to_user(buf, &uffdio_api, sizeof(uffdio_api)))
Signature · 1 line
2.44.0
#1 lore raw

PATCH v2 2/3 Update uffd-stress to handle EINVAL for unset config features

Now that we have updated userfaultfd_api to correctly return EIVAL when a feature is requested but not available, let's fix the uffd-stress test to only set the UFFD_FEATURE_WP_UNPOPULATED feature when the config is set. In addition, still run the test if the CONFIG_PTE_MARKER_UFFD_WP is not set, just dont use the corresponding UFFD_FEATURE_WP_UNPOPULATED feature.

---
 tools/testing/selftests/mm/uffd-stress.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)
tools/testing/selftests/mm/uffd-stress.c +9 -5
index f78bab0f3d45..b9b6d858eab8 100644
@@ -38,6 +38,8 @@
 
 #ifdef __NR_userfaultfd
 
+uint64_t features;
+
 #define BOUNCE_RANDOM		(1<<0)
 #define BOUNCE_RACINGFAULTS	(1<<1)
 #define BOUNCE_VERIFY		(1<<2)
@@ -247,10 +249,14 @@ static int userfaultfd_stress(void)
 	unsigned long nr;
 	struct uffd_args args[nr_cpus];
 	uint64_t mem_size = nr_pages * page_size;
+	int flags = 0;
 
 	memset(args, 0, sizeof(struct uffd_args) * nr_cpus);
 
-	if (uffd_test_ctx_init(UFFD_FEATURE_WP_UNPOPULATED, NULL))
+	if (features & UFFD_FEATURE_WP_UNPOPULATED && test_type == TEST_ANON)
+		flags = UFFD_FEATURE_WP_UNPOPULATED;
+
+	if (uffd_test_ctx_init(flags, NULL))
 		err("context init failed");
 
 	if (posix_memalign(&area, page_size, page_size))
@@ -385,8 +391,6 @@ static void set_test_type(const char *type)
 
 static void parse_test_type_arg(const char *raw_type)
 {
-	uint64_t features = UFFD_API_FEATURES;
-
 	set_test_type(raw_type);
 
 	if (!test_type)
@@ -409,8 +413,8 @@ static void parse_test_type_arg(const char *raw_type)
 	 * feature.
 	 */
 
-	if (userfaultfd_open(&features))
-		err("Userfaultfd open failed");
+	if (uffd_get_features(&features))
+		err("failed to get available features");
 
 	test_uffdio_wp = test_uffdio_wp &&
 		(features & UFFD_FEATURE_PAGEFAULT_FLAG_WP);
Signature · 1 line
2.44.0
reviewer Acked-by on v2 2/3
#2 lore raw

Re: [PATCH v2 2/3] Update uffd-stress to handle EINVAL for unset config features

Quoted 8 lines · Audra Mitchell

Now that we have updated userfaultfd_api to correctly return EIVAL when a feature is requested but not available, let's fix the uffd-stress test to only set the UFFD_FEATURE_WP_UNPOPULATED feature when the config is set. In addition, still run the test if the CONFIG_PTE_MARKER_UFFD_WP is not set, just dont use the corresponding UFFD_FEATURE_WP_UNPOPULATED feature.

Signature · 1 line
Peter Xu
#3 lore raw

PATCH v2 3/3 Turn off test_uffdio_wp if CONFIG_PTE_MARKER_UFFD_WP is not configured.

In reply to Audra Mitchell

If CONFIG_PTE_MARKER_UFFD_WP is disabled, then testing with test_uffdio_up enables calling uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. The kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP on anonymous vmas.

---
 tools/testing/selftests/mm/uffd-stress.c | 3 +++
 1 file changed, 3 insertions(+)
tools/testing/selftests/mm/uffd-stress.c +3 -0
index b9b6d858eab8..2601c9dfadd6 100644
@@ -419,6 +419,9 @@ static void parse_test_type_arg(const char *raw_type)
 	test_uffdio_wp = test_uffdio_wp &&
 		(features & UFFD_FEATURE_PAGEFAULT_FLAG_WP);
 
+	if (test_type != TEST_ANON && !(features & UFFD_FEATURE_WP_UNPOPULATED))
+		test_uffdio_wp = false;
+
 	close(uffd);
 	uffd = -1;
 }
Signature · 1 line
2.44.0
reviewer
#4 lore raw

Re: [PATCH v2 3/3] Turn off test_uffdio_wp if CONFIG_PTE_MARKER_UFFD_WP is not configured.

Quoted 1 line · Audra Mitchell

If CONFIG_PTE_MARKER_UFFD_WP is disabled, then testing with test_uffdio_up

Here you're talking about pte markers, then..

Quoted 14 lines

enables calling uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. The kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP on anonymous vmas.

---
 tools/testing/selftests/mm/uffd-stress.c | 3 +++
 1 file changed, 3 insertions(+)
tools/testing/selftests/mm/uffd-stress.c +0 -0
index b9b6d858eab8..2601c9dfadd6 100644

tools/testing/selftests/mm/uffd-stress.c static void parse_test_type_arg(const char *raw_type) · line 423

Quoted 5 lines
tools/testing/selftests/mm/uffd-stress.c +2 -0
@@ -419,6 +419,9 @@ static void parse_test_type_arg(const char *raw_type)
 	test_uffdio_wp = test_uffdio_wp &&
 		(features & UFFD_FEATURE_PAGEFAULT_FLAG_WP);
 
+	if (test_type != TEST_ANON && !(features & UFFD_FEATURE_WP_UNPOPULATED))
+		test_uffdio_wp = false;

... here you're checking against wp_unpopulated. I'm slightly confused.

Are you running this test over shmem/hugetlb when the WP feature isn't supported?

I'm wondering whether you're looking for UFFD_FEATURE_WP_HUGETLBFS_SHMEM instead.

Thanks,

Quoted 7 lines
tools/testing/selftests/mm/uffd-stress.c +1 -0
+
 	close(uffd);
 	uffd = -1;
 }
Signature · 1 line
2.44.0
Signature · 1 line
Peter Xu
#5 lore raw

Re: [PATCH v2 3/3] Turn off test_uffdio_wp if CONFIG_PTE_MARKER_UFFD_WP is not configured.

Quoted 27 lines · Peter Xu
Quoted 1 line · Audra Mitchell

If CONFIG_PTE_MARKER_UFFD_WP is disabled, then testing with test_uffdio_up

Here you're talking about pte markers, then..

Quoted 14 lines

enables calling uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. The kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP on anonymous vmas.

---
 tools/testing/selftests/mm/uffd-stress.c | 3 +++
 1 file changed, 3 insertions(+)
tools/testing/selftests/mm/uffd-stress.c +0 -0
index b9b6d858eab8..2601c9dfadd6 100644

tools/testing/selftests/mm/uffd-stress.c static void parse_test_type_arg(const char *raw_type) · line 423

Quoted 5 lines
tools/testing/selftests/mm/uffd-stress.c +2 -0
@@ -419,6 +419,9 @@ static void parse_test_type_arg(const char *raw_type)
 	test_uffdio_wp = test_uffdio_wp &&
 		(features & UFFD_FEATURE_PAGEFAULT_FLAG_WP);
 
+	if (test_type != TEST_ANON && !(features & UFFD_FEATURE_WP_UNPOPULATED))
+		test_uffdio_wp = false;

... here you're checking against wp_unpopulated. I'm slightly confused.

Quoted 5 lines

Are you running this test over shmem/hugetlb when the WP feature isn't supported?

I'm wondering whether you're looking for UFFD_FEATURE_WP_HUGETLBFS_SHMEM instead.

I can confirm, its all really confusing... So in userfaultfd_api, we disable three features if CONFIG_PTE_MARKER_UFFD_WP is not enabled- including UFFD_FEATURE_WP_UNPOPULATED:

#ifndef CONFIG_PTE_MARKER_UFFD_WP
        uffdio_api.features &= ~UFFD_FEATURE_WP_HUGETLBFS_SHMEM;
        uffdio_api.features &= ~UFFD_FEATURE_WP_UNPOPULATED;
        uffdio_api.features &= ~UFFD_FEATURE_WP_ASYNC;

#endif

If you run the userfaultfd selftests with the run_vmtests script we get several failures stemming from trying to call uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. However, the kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP - which is set when you pass the UFFDIO_REGISTER_MODE_WP flag - on anonymous vmas.

In parse_test_type_arg() I added the features check against UFFD_FEATURE_WP_UNPOPULATED as it seemed the most well know feature/flag. I'm more than happy to take any suggestions and adapt them if you have any!

Thanks in advance and happy Monday!

-- Audra

Quoted 13 lines

Thanks,

Quoted 7 lines
tools/testing/selftests/mm/uffd-stress.c +1 -0
+
 	close(uffd);
 	uffd = -1;
 }
Signature · 1 line
2.44.0
Signature · 1 line
Peter Xu
reviewer
#6 lore raw

Re: [PATCH v2 3/3] Turn off test_uffdio_wp if CONFIG_PTE_MARKER_UFFD_WP is not configured.

Quoted 51 lines · Audra Mitchell
Quoted 27 lines · Peter Xu
Quoted 1 line · Audra Mitchell

If CONFIG_PTE_MARKER_UFFD_WP is disabled, then testing with test_uffdio_up

Here you're talking about pte markers, then..

Quoted 14 lines

enables calling uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. The kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP on anonymous vmas.

---
 tools/testing/selftests/mm/uffd-stress.c | 3 +++
 1 file changed, 3 insertions(+)
tools/testing/selftests/mm/uffd-stress.c +0 -0
index b9b6d858eab8..2601c9dfadd6 100644

tools/testing/selftests/mm/uffd-stress.c static void parse_test_type_arg(const char *raw_type) · line 423

Quoted 5 lines
tools/testing/selftests/mm/uffd-stress.c +2 -0
@@ -419,6 +419,9 @@ static void parse_test_type_arg(const char *raw_type)
 	test_uffdio_wp = test_uffdio_wp &&
 		(features & UFFD_FEATURE_PAGEFAULT_FLAG_WP);
 
+	if (test_type != TEST_ANON && !(features & UFFD_FEATURE_WP_UNPOPULATED))
+		test_uffdio_wp = false;

... here you're checking against wp_unpopulated. I'm slightly confused.

Quoted 5 lines

Are you running this test over shmem/hugetlb when the WP feature isn't supported?

I'm wondering whether you're looking for UFFD_FEATURE_WP_HUGETLBFS_SHMEM instead.

I can confirm, its all really confusing... So in userfaultfd_api, we disable three features if CONFIG_PTE_MARKER_UFFD_WP is not enabled- including UFFD_FEATURE_WP_UNPOPULATED:

#ifndef CONFIG_PTE_MARKER_UFFD_WP
        uffdio_api.features &= ~UFFD_FEATURE_WP_HUGETLBFS_SHMEM;
        uffdio_api.features &= ~UFFD_FEATURE_WP_UNPOPULATED;
        uffdio_api.features &= ~UFFD_FEATURE_WP_ASYNC;

#endif

If you run the userfaultfd selftests with the run_vmtests script we get several failures stemming from trying to call uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. However, the kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP - which is set when you pass the UFFDIO_REGISTER_MODE_WP flag - on anonymous vmas.

Quoted 3 lines

In parse_test_type_arg() I added the features check against UFFD_FEATURE_WP_UNPOPULATED as it seemed the most well know feature/flag. I'm more than happy to take any suggestions and adapt them if you have any!

There're documents for these features in the headers:

* UFFD_FEATURE_WP_HUGETLBFS_SHMEM indicates that userfaultfd
* write-protection mode is supported on both shmem and hugetlbfs. *
* UFFD_FEATURE_WP_UNPOPULATED indicates that userfaultfd
* write-protection mode will always apply to unpopulated pages
* (i.e. empty ptes). This will be the default behavior for shmem
* & hugetlbfs, so this flag only affects anonymous memory behavior
* when userfault write-protection mode is registered.

While in this context ("test_type != TEST_ANON") IIUC the accurate feature to check is UFFD_FEATURE_WP_HUGETLBFS_SHMEM.

In most kernels they should behave the same indeed, but note that since UNPOPULATED was introduced later than shmem/hugetlb support, it means on some kernel the result of checking these two features will be different.

Thanks,

Signature · 1 line
Peter Xu
maintainer
#7 lore raw

Re: [PATCH v2 3/3] Turn off test_uffdio_wp if CONFIG_PTE_MARKER_UFFD_WP is not configured.

Quoted 27 lines · On Mon, 24 Jun 2024 10:42:00 -0400 Peter Xu <[email protected]> wrote:
Quoted 12 lines
        uffdio_api.features &= ~UFFD_FEATURE_WP_HUGETLBFS_SHMEM;
        uffdio_api.features &= ~UFFD_FEATURE_WP_UNPOPULATED;
        uffdio_api.features &= ~UFFD_FEATURE_WP_ASYNC;

#endif

If you run the userfaultfd selftests with the run_vmtests script we get several failures stemming from trying to call uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. However, the kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP - which is set when you pass the UFFDIO_REGISTER_MODE_WP flag - on anonymous vmas.

Quoted 3 lines

In parse_test_type_arg() I added the features check against UFFD_FEATURE_WP_UNPOPULATED as it seemed the most well know feature/flag. I'm more than happy to take any suggestions and adapt them if you have any!

There're documents for these features in the headers:

* UFFD_FEATURE_WP_HUGETLBFS_SHMEM indicates that userfaultfd
* write-protection mode is supported on both shmem and hugetlbfs. *
* UFFD_FEATURE_WP_UNPOPULATED indicates that userfaultfd
* write-protection mode will always apply to unpopulated pages
* (i.e. empty ptes). This will be the default behavior for shmem
* & hugetlbfs, so this flag only affects anonymous memory behavior
* when userfault write-protection mode is registered.

Quoted 6 lines

While in this context ("test_type != TEST_ANON") IIUC the accurate feature to check is UFFD_FEATURE_WP_HUGETLBFS_SHMEM.

In most kernels they should behave the same indeed, but note that since UNPOPULATED was introduced later than shmem/hugetlb support, it means on some kernel the result of checking these two features will be different.

I'm unsure what to do with this series. Peter, your review comments are unclear - do you request updates?

Also, the patches weren't cc:linux-mm which limits the audience. I'll drop this version. Audra, please continue to move this forward.

reviewer
#8 lore raw

Re: [PATCH v2 3/3] Turn off test_uffdio_wp if CONFIG_PTE_MARKER_UFFD_WP is not configured.

Quoted 32 lines · Andrew Morton
Quoted 30 lines · On Mon, 24 Jun 2024 10:42:00 -0400 Peter Xu <[email protected]> wrote:
Quoted 12 lines
        uffdio_api.features &= ~UFFD_FEATURE_WP_HUGETLBFS_SHMEM;
        uffdio_api.features &= ~UFFD_FEATURE_WP_UNPOPULATED;
        uffdio_api.features &= ~UFFD_FEATURE_WP_ASYNC;

#endif

If you run the userfaultfd selftests with the run_vmtests script we get several failures stemming from trying to call uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. However, the kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP - which is set when you pass the UFFDIO_REGISTER_MODE_WP flag - on anonymous vmas.

Quoted 3 lines

In parse_test_type_arg() I added the features check against UFFD_FEATURE_WP_UNPOPULATED as it seemed the most well know feature/flag. I'm more than happy to take any suggestions and adapt them if you have any!

There're documents for these features in the headers:

* UFFD_FEATURE_WP_HUGETLBFS_SHMEM indicates that userfaultfd
* write-protection mode is supported on both shmem and hugetlbfs. *
* UFFD_FEATURE_WP_UNPOPULATED indicates that userfaultfd
* write-protection mode will always apply to unpopulated pages
* (i.e. empty ptes). This will be the default behavior for shmem
* & hugetlbfs, so this flag only affects anonymous memory behavior
* when userfault write-protection mode is registered.

While in this context ("test_type != TEST_ANON") IIUC the accurate feature to check is UFFD_FEATURE_WP_HUGETLBFS_SHMEM.

Quoted 6 lines
Quoted 3 lines

In most kernels they should behave the same indeed, but note that since UNPOPULATED was introduced later than shmem/hugetlb support, it means on some kernel the result of checking these two features will be different.

I'm unsure what to do with this series. Peter, your review comments are unclear - do you request updates?

Yes, or some clarification from Audra would also work.

What I was trying to say is here I think the code should check against UFFD_FEATURE_WP_HUGETLBFS_SHMEM instead.

Thanks,

Signature · 1 line
Peter Xu
#9 lore raw

Re: [PATCH v2 3/3] Turn off test_uffdio_wp if CONFIG_PTE_MARKER_UFFD_WP is not configured.

Quoted 37 lines · Peter Xu
Quoted 36 lines · Andrew Morton
Quoted 33 lines · On Mon, 24 Jun 2024 10:42:00 -0400 Peter Xu <[email protected]> wrote:
Quoted 12 lines
        uffdio_api.features &= ~UFFD_FEATURE_WP_HUGETLBFS_SHMEM;
        uffdio_api.features &= ~UFFD_FEATURE_WP_UNPOPULATED;
        uffdio_api.features &= ~UFFD_FEATURE_WP_ASYNC;

#endif

If you run the userfaultfd selftests with the run_vmtests script we get several failures stemming from trying to call uffdio_regsiter with the flag UFFDIO_REGISTER_MODE_WP. However, the kernel ensures in vma_can_userfault() that if CONFIG_PTE_MARKER_UFFD_WP is disabled, only allow the VM_UFFD_WP - which is set when you pass the UFFDIO_REGISTER_MODE_WP flag - on anonymous vmas.

Quoted 3 lines

In parse_test_type_arg() I added the features check against UFFD_FEATURE_WP_UNPOPULATED as it seemed the most well know feature/flag. I'm more than happy to take any suggestions and adapt them if you have any!

There're documents for these features in the headers:

* UFFD_FEATURE_WP_HUGETLBFS_SHMEM indicates that userfaultfd
* write-protection mode is supported on both shmem and hugetlbfs. *
* UFFD_FEATURE_WP_UNPOPULATED indicates that userfaultfd
* write-protection mode will always apply to unpopulated pages
* (i.e. empty ptes). This will be the default behavior for shmem
* & hugetlbfs, so this flag only affects anonymous memory behavior
* when userfault write-protection mode is registered.

While in this context ("test_type != TEST_ANON") IIUC the accurate feature to check is UFFD_FEATURE_WP_HUGETLBFS_SHMEM.

In most kernels they should behave the same indeed, but note that since UNPOPULATED was introduced later than shmem/hugetlb support, it means on some kernel the result of checking these two features will be different.

Quoted 7 lines
Quoted 2 lines

I'm unsure what to do with this series. Peter, your review comments are unclear - do you request updates?

Yes, or some clarification from Audra would also work.

What I was trying to say is here I think the code should check against UFFD_FEATURE_WP_HUGETLBFS_SHMEM instead.

I was meaning to reply back and ask if Andrew wanted me to push a v3 and change the check from UFFD_FEATURE_WP_UNPOPULATED to UFFD_FEATURE_WP_HUGETLBFS_SHMEM or if he just wanted to do it, but I'll go ahead and submit v3 with the change shortly.

Also as an aside I ran scripts/get_maintainer.pl to get the email list. I probably should have thought a little bit about why the linux-mm list was missing....

Sorry about the delay and confusion!

Quoted 5 lines

Thanks,

Signature · 1 line
Peter Xu
reviewer Acked-by on v2 1/3
#10 lore raw
In reply to Audra Mitchell
Quoted 8 lines · Audra Mitchell

Currently if we request a feature that is not set in the Kernel config we fail silently and return all the available features. However, the man page indicates we should return an EINVAL.

We need to fix this issue since we can end up with a Kernel warning should a program request the feature UFFD_FEATURE_WP_UNPOPULATED on a kernel with the config not set with this feature.

Quoted 8 lines
 [  200.812896] WARNING: CPU: 91 PID: 13634 at mm/memory.c:1660 zap_pte_range+0x43d/0x660
 [  200.820738] Modules linked in:
 [  200.869387] CPU: 91 PID: 13634 Comm: userfaultfd Kdump: loaded Not tainted 6.9.0-rc5+ #8
 [  200.877477] Hardware name: Dell Inc. PowerEdge R6525/0N7YGH, BIOS 2.7.3 03/30/2022
 [  200.885052] RIP: 0010:zap_pte_range+0x43d/0x660
Fixes e06f1e1dd499 ("userfaultfd: wp: enabled write protection in userfaultfd API")

Please prefix the subject with "mm/uffd:" or "mm/userfault:" if there's a repost.

Thanks,

Signature · 1 line
Peter Xu
maintainer
#11 lore raw
In reply to Audra Mitchell
Quoted 8 lines · On Fri, 21 Jun 2024 14:12:22 -0400 Audra Mitchell <[email protected]> wrote:

Currently if we request a feature that is not set in the Kernel config we fail silently and return all the available features. However, the man page indicates we should return an EINVAL.

We need to fix this issue since we can end up with a Kernel warning should a program request the feature UFFD_FEATURE_WP_UNPOPULATED on a kernel with the config not set with this feature.

Quoted 7 lines
 [  200.812896] WARNING: CPU: 91 PID: 13634 at mm/memory.c:1660 zap_pte_range+0x43d/0x660
 [  200.820738] Modules linked in:
 [  200.869387] CPU: 91 PID: 13634 Comm: userfaultfd Kdump: loaded Not tainted 6.9.0-rc5+ #8
 [  200.877477] Hardware name: Dell Inc. PowerEdge R6525/0N7YGH, BIOS 2.7.3 03/30/2022
 [  200.885052] RIP: 0010:zap_pte_range+0x43d/0x660
Fixes e06f1e1dd499 ("userfaultfd: wp: enabled write protection in userfaultfd API")

A userspace-triggerable WARN is bad. I added cc:stable to this.

reviewer
#12 lore raw
Quoted 16 lines · Andrew Morton
Quoted 14 lines · On Fri, 21 Jun 2024 14:12:22 -0400 Audra Mitchell <[email protected]> wrote:

Currently if we request a feature that is not set in the Kernel config we fail silently and return all the available features. However, the man page indicates we should return an EINVAL.

We need to fix this issue since we can end up with a Kernel warning should a program request the feature UFFD_FEATURE_WP_UNPOPULATED on a kernel with the config not set with this feature.

 [  200.812896] WARNING: CPU: 91 PID: 13634 at mm/memory.c:1660 zap_pte_range+0x43d/0x660
 [  200.820738] Modules linked in:
 [  200.869387] CPU: 91 PID: 13634 Comm: userfaultfd Kdump: loaded Not tainted 6.9.0-rc5+ #8
 [  200.877477] Hardware name: Dell Inc. PowerEdge R6525/0N7YGH, BIOS 2.7.3 03/30/2022
 [  200.885052] RIP: 0010:zap_pte_range+0x43d/0x660
Quoted 3 lines
Quoted 1 line
Fixes e06f1e1dd499 ("userfaultfd: wp: enabled write protection in userfaultfd API")

A userspace-triggerable WARN is bad. I added cc:stable to this.

Andrew,

Note that this change may fix a WARN, but it may also start to break userspace which might be worse if it happens, IMHO. I forgot to mention that here, but only mentioned that in v1, and from that POV not copying stable seems the right thing.

https://lore.kernel.org/all/ZjuIEH8TW2tWcqXQ@x1n/

In summary: I think we can stick with Fixes on e06f1e1dd499, but we don't copy stable. The major reason we don't copy stable here is not only about complexity of such backport, but also that there can be apps trying to pass in unsupported bits (even if the kernel didn't support it) but keep using MISSING mode only, then we shouldn't fail them easily after a stable upgrade. Just smells dangerous to backport.

I'm not sure what's the best solution for this. A sweet spot might be that we start to fix it in new kernels only but not stable ones.

Thanks,

Signature · 1 line
Peter Xu
maintainer
#13 lore raw
Quoted 13 lines · On Sun, 23 Jun 2024 11:26:37 -0400 Peter Xu <[email protected]> wrote:
Quoted 3 lines
Quoted 1 line
Fixes e06f1e1dd499 ("userfaultfd: wp: enabled write protection in userfaultfd API")

A userspace-triggerable WARN is bad. I added cc:stable to this.

Andrew,

Note that this change may fix a WARN, but it may also start to break userspace which might be worse if it happens, IMHO. I forgot to mention that here, but only mentioned that in v1, and from that POV not copying stable seems the right thing.

https://lore.kernel.org/all/ZjuIEH8TW2tWcqXQ@x1n/

Quoted 7 lines

In summary: I think we can stick with Fixes on e06f1e1dd499, but we don't copy stable. The major reason we don't copy stable here is not only about complexity of such backport, but also that there can be apps trying to pass in unsupported bits (even if the kernel didn't support it) but keep using MISSING mode only, then we shouldn't fail them easily after a stable upgrade. Just smells dangerous to backport.

OK. And I'll move it into the 6.11-rc1 queue, for the next merge window.