Skip to contentSkip to conversation

syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

PATCH Superseded Yang Xu 3 participants 7 messages Dec 18, 2019 – Jan 7, 2020

lore
Share & tools
lore

Yang Xu

author
#0 lore raw

PATCH v2 1/1 syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

Since commit cefdca0a86be ("userfaultfd/sysctl: add vm.unprivileged_userfaultfd"). , it adds a global sysctl knob "vm.unprivileged_userfaultfd" to control whether unprivileged users can use the userfaultfd system calls. Set this to 1 to allow unprivileged users to use the userfaultfd system calls, or set this to 0 to restrict userfaultfd to only privileged users (with SYS_CAP_PTRACE capability). The default value is 1. Add hint about it.

---
 .../syscalls/userfaultfd/userfaultfd01.c      | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)
testcases/kernel/syscalls/userfaultfd/userfaultfd01.c +13 -6
index a5e142209..4e178b4f8 100644
@@ -82,12 +82,19 @@ static void run(void)
 
 	set_pages();
 
-	uffd = sys_userfaultfd(O_CLOEXEC | O_NONBLOCK);
-
-	if (uffd == -1)
-		tst_brk(TBROK | TERRNO,
-			"Could not create userfault file descriptor");
-
+	TEST(sys_userfaultfd(O_CLOEXEC | O_NONBLOCK));
+
+	if (TST_RET == -1) {
+		if (TST_ERR == EPERM) {
+			tst_res(TCONF, "Hint: check /proc/sys/vm/unprivileged_userfaultfd");
+			tst_brk(TCONF | TTERRNO,
+				"userfaultfd() requires CAP_SYS_PTRACE on this system");
+		} else
+			tst_brk(TBROK | TTERRNO,
+				"Could not create userfault file descriptor");
+	}
+
+	uffd = TST_RET;
 	uffdio_api.api = UFFD_API;
 	uffdio_api.features = 0;
 	SAFE_IOCTL(uffd, UFFDIO_API, &uffdio_api);
Signature · 1 line
2.18.0
author
#1 lore raw

PATCH v2 1/1 syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

Hi
Ping.

Quoted 43 lines

Since commit cefdca0a86be ("userfaultfd/sysctl: add vm.unprivileged_userfaultfd"). , it adds a global sysctl knob "vm.unprivileged_userfaultfd" to control whether unprivileged users can use the userfaultfd system calls. Set this to 1 to allow unprivileged users to use the userfaultfd system calls, or set this to 0 to restrict userfaultfd to only privileged users (with SYS_CAP_PTRACE capability). The default value is 1. Add hint about it.

---
  .../syscalls/userfaultfd/userfaultfd01.c      | 19 +++++++++++++------
  1 file changed, 13 insertions(+), 6 deletions(-)
testcases/kernel/syscalls/userfaultfd/userfaultfd01.c +13 -6
index a5e142209..4e178b4f8 100644
@@ -82,12 +82,19 @@ static void run(void)
 
  	set_pages();
 
-	uffd = sys_userfaultfd(O_CLOEXEC | O_NONBLOCK);
-
-	if (uffd == -1)
-		tst_brk(TBROK | TERRNO,
-			"Could not create userfault file descriptor");
-
+	TEST(sys_userfaultfd(O_CLOEXEC | O_NONBLOCK));
+
+	if (TST_RET == -1) {
+		if (TST_ERR == EPERM) {
+			tst_res(TCONF, "Hint: check /proc/sys/vm/unprivileged_userfaultfd");
+			tst_brk(TCONF | TTERRNO,
+				"userfaultfd() requires CAP_SYS_PTRACE on this system");
+		} else
+			tst_brk(TBROK | TTERRNO,
+				"Could not create userfault file descriptor");
+	}
+
+	uffd = TST_RET;
  	uffdio_api.api = UFFD_API;
  	uffdio_api.features = 0;
  	SAFE_IOCTL(uffd, UFFDIO_API, &uffdio_api);
#2 lore raw

PATCH v2 1/1 syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

In reply to Yang Xu
Quoted 11 lines · Yang Xu

Since commit cefdca0a86be ("userfaultfd/sysctl: add vm.unprivileged_userfaultfd").
, it adds a global sysctl knob "vm.unprivileged_userfaultfd" to control whether
unprivileged users can use the userfaultfd system calls. Set this to 1 to allow
unprivileged users to use the userfaultfd system calls, or set this to 0 to restrict userfaultfd to only privileged users (with SYS_CAP_PTRACE capability). The
default value is 1. Add hint about it.

Can we do the "vm.unprivileged_userfaultfd" check in the setup() and do set to 1 if it exists?
And maybe we need more tests for the global sysctl knob "vm.unprivileged_userfaultfd".

Quoted 49 lines
---
 .../syscalls/userfaultfd/userfaultfd01.c      | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)
testcases/kernel/syscalls/userfaultfd/userfaultfd01.c +13 -6
index a5e142209..4e178b4f8 100644
@@ -82,12 +82,19 @@ static void run(void)
 
        set_pages();
 
-       uffd = sys_userfaultfd(O_CLOEXEC | O_NONBLOCK);
-
-       if (uffd == -1)
-               tst_brk(TBROK | TERRNO,
-                       "Could not create userfault file descriptor");
-
+       TEST(sys_userfaultfd(O_CLOEXEC | O_NONBLOCK));
+
+       if (TST_RET == -1) {
+               if (TST_ERR == EPERM) {
+                       tst_res(TCONF, "Hint: check
 /proc/sys/vm/unprivileged_userfaultfd");
+                       tst_brk(TCONF | TTERRNO,
+                               "userfaultfd() requires CAP_SYS_PTRACE on
 this system");
+               } else
+                       tst_brk(TBROK | TTERRNO,
+                               "Could not create userfault file
 descriptor");
+       }
+
+       uffd = TST_RET;
        uffdio_api.api = UFFD_API;
        uffdio_api.features = 0;
        SAFE_IOCTL(uffd, UFFDIO_API, &uffdio_api);

--
2.18.0

--
Mailing list info: https://lists.linux.it/listinfo/ltp

Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20200106/90744333/attachment.htm>
author
#3 lore raw

PATCH v2 1/1 syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

Hi Li

Quoted 18 lines

On Wed, Dec 18, 2019 at 12:02 PM Yang Xu <[email protected] <mailto:[email protected]>> wrote:

Since commit cefdca0a86be ("userfaultfd/sysctl: add vm.unprivileged_userfaultfd").
, it adds a global sysctl knob "vm.unprivileged_userfaultfd" to control whether
unprivileged users can use the userfaultfd system calls. Set this to 1 to allow
unprivileged users to use the userfaultfd system calls, or set this to 0 to
restrict userfaultfd to only privileged users (with SYS_CAP_PTRACE capability). The
default value is 1. Add hint about it.

Quoted 2 lines

Can we do the "vm.unprivileged_userfaultfd" check in the setup() and do set to 1 if it exists?

I remembered Jan Stancek has a patch about bpf hint about unprivileged_bpf_disabled, I do as same as that patch did.

Also, month agos about acct02 discussion, Cyril points about adjusting the threshold value of resume and suppend to make case passes.

So, I have a question that we have unified standards about these cases (bpf, acct02, affected by sysctl)? Report TCONF or modify argument to make case passes?

@Cyril and @Jan Stancek What do you think about it?

Quoted 2 lines

And maybe?we need more tests for the global sysctl knob "vm.unprivileged_userfaultfd".Eeven though we don't have case to test unprivileged_bpf_disabled. I

still think testing unprivileged_userfaultfd is meaningful and we can begin with it.

Best Regards
Yang Xu

Quoted 56 lines

    Signed-off-by: Yang Xu <[email protected]
    <mailto:[email protected]>>
    ---
     ?.../syscalls/userfaultfd/userfaultfd01.c? ? ? | 19 +++++++++++++------
     ?1 file changed, 13 insertions(+), 6 deletions(-)

    diff --git a/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    b/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    index a5e142209..4e178b4f8 100644
    --- a/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    +++ b/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    @@ -82,12 +82,19 @@ static void run(void)

     ? ? ? ? set_pages();

    -? ? ? ?uffd = sys_userfaultfd(O_CLOEXEC | O_NONBLOCK);
    -
    -? ? ? ?if (uffd == -1)
    -? ? ? ? ? ? ? ?tst_brk(TBROK | TERRNO,
    -? ? ? ? ? ? ? ? ? ? ? ?"Could not create userfault file descriptor");
    -
    +? ? ? ?TEST(sys_userfaultfd(O_CLOEXEC | O_NONBLOCK));
    +
    +? ? ? ?if (TST_RET == -1) {
    +? ? ? ? ? ? ? ?if (TST_ERR == EPERM) {
    +? ? ? ? ? ? ? ? ? ? ? ?tst_res(TCONF, "Hint: check
    /proc/sys/vm/unprivileged_userfaultfd");
    +? ? ? ? ? ? ? ? ? ? ? ?tst_brk(TCONF | TTERRNO,
    +? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?"userfaultfd() requires
    CAP_SYS_PTRACE on this system");
    +? ? ? ? ? ? ? ?} else
    +? ? ? ? ? ? ? ? ? ? ? ?tst_brk(TBROK | TTERRNO,
    +? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?"Could not create userfault file
    descriptor");
    +? ? ? ?}
    +
    +? ? ? ?uffd = TST_RET;
     ? ? ? ? uffdio_api.api = UFFD_API;
     ? ? ? ? uffdio_api.features = 0;
     ? ? ? ? SAFE_IOCTL(uffd, UFFDIO_API, &uffdio_api);
    -- 
    2.18.0




    -- 
    Mailing list info: https://lists.linux.it/listinfo/ltp
Signature · 2 lines
Regards,
Li Wang
#4 lore raw

PATCH v2 1/1 syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

Quoted 19 lines · Yang Xu

Hi Li

Quoted 18 lines

On Wed, Dec 18, 2019 at 12:02 PM Yang Xu <[email protected] <mailto:[email protected]>> wrote:

Since commit cefdca0a86be ("userfaultfd/sysctl: add vm.unprivileged_userfaultfd").
, it adds a global sysctl knob "vm.unprivileged_userfaultfd" to control whether
unprivileged users can use the userfaultfd system calls. Set this to 1 to allow
unprivileged users to use the userfaultfd system calls, or set this to 0 to
restrict userfaultfd to only privileged users (with SYS_CAP_PTRACE capability). The
default value is 1. Add hint about it.

Quoted 5 lines
Quoted 2 lines

Can we do the "vm.unprivileged_userfaultfd" check in the setup() and do set to 1 if it exists?

I remembered Jan Stancek has a patch about bpf hint about unprivileged_bpf_disabled, I do as same as that patch did.

I just echo the nob file "unprivileged_bpf_disabled" and find it can't be changed at runtime. So if the "vm.unprivileged_userfaultfd" is like this behavior too, probably we can only do TCONF as your original patch. AnywayI will take a close look at the kernel commit later.

# cat /proc/sys/kernel/unprivileged_bpf_disabled
1
# echo 0 > /proc/sys/kernel/unprivileged_bpf_disabled
-bash: echo: write error: Invalid argument
Quoted 79 lines

Also, month agos about acct02 discussion, Cyril points about adjusting the threshold value of resume and suppend to make case passes.

So, I have a question that we have unified standards about these cases (bpf, acct02, affected by sysctl)? Report TCONF or modify argument to make case passes?

@Cyril and @Jan Stancek What do you think about it?

Quoted 2 lines

And maybe we need more tests for the global sysctl knob "vm.unprivileged_userfaultfd".Eeven though we don't have case to test

unprivileged_bpf_disabled. I
still think testing unprivileged_userfaultfd is meaningful and we can begin with it.

Best Regards
Yang Xu

Quoted 6 lines

<mailto:[email protected]>>
---
.../syscalls/userfaultfd/userfaultfd01.c | 19

+++++++++++++------

Quoted 16 lines
      1 file changed, 13 insertions(+), 6 deletions(-)
    diff --git a/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    b/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    index a5e142209..4e178b4f8 100644
    --- a/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    +++ b/testcases/kernel/syscalls/userfaultfd/userfaultfd01.c
    @@ -82,12 +82,19 @@ static void run(void)

             set_pages();

    -       uffd = sys_userfaultfd(O_CLOEXEC | O_NONBLOCK);
    -
    -       if (uffd == -1)
    -               tst_brk(TBROK | TERRNO,
    -                       "Could not create userfault file

descriptor");

Quoted 34 lines

-
+ TEST(sys_userfaultfd(O_CLOEXEC | O_NONBLOCK));
+
+ if (TST_RET == -1) {
+ if (TST_ERR == EPERM) {
+ tst_res(TCONF, "Hint: check
/proc/sys/vm/unprivileged_userfaultfd");
+ tst_brk(TCONF | TTERRNO,
+ "userfaultfd() requires
CAP_SYS_PTRACE on this system");
+ } else
+ tst_brk(TBROK | TTERRNO,
+ "Could not create userfault file
descriptor");
+ }
+
+ uffd = TST_RET;
uffdio_api.api = UFFD_API; uffdio_api.features = 0; SAFE_IOCTL(uffd, UFFDIO_API, &uffdio_api);
--
2.18.0

--
Mailing list info: https://lists.linux.it/listinfo/ltp

--
Regards,
Li Wang

Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20200106/1f7f925e/attachment-0001.htm>
#5 lore raw

PATCH v2 1/1 syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

In reply to Yang Xu

Hi Yang,

Quoted 6 lines · Yang Xu

...

Quoted 2 lines

Can we do the "vm.unprivileged_userfaultfd" check in the setup() and do set to 1 if it exists?

I remembered Jan Stancek has a patch about bpf hint about unprivileged_bpf_disabled, I do as same as that patch did.

Quoted 7 lines

Also, month agos about acct02 discussion, Cyril points about adjusting the threshold value of resume and suppend to make case passes.

So, I have a question that we have unified standards about these cases (bpf, acct02, affected by sysctl)? Report TCONF or modify argument to make case passes?

After thinking over, to report TCONF is the wise method here because this test might run with an unprivileged user and it can not modify the knob "vm.unprivileged_userfaultfd".

Sorry and now I pull back my last comment.

Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20200107/cb5dcf9f/attachment-0001.htm>
#6 lore raw

PATCH v2 1/1 syscalls/userfaultfd01: add hint about unprivileged_userfaultfd

----- Original Message -----

Quoted 8 lines

After thinking over, to report TCONF is the wise method here because this test might run with an unprivileged user and it can not modify the knob "vm.unprivileged_userfaultfd".

Sorry and now I pull back my last comment.

Pushed, thanks.