[PATCH bpf v2 1/2] bpf: link: refuse non-O_RDWR flags in BPF_OBJ_GET

Invoking BPF_OBJ_GET on a pinned bpf_link checks the path access
permissions based on file_flags, but the returned fd ignores flags.
This means that any user can acquire a "read-write" fd for a pinned
link with mode 0664 by invoking BPF_OBJ_GET with BPF_F_RDONLY in
file_flags. The fd can be used to invoke BPF_LINK_DETACH, etc.

Fix this by refusing non-O_RDWR flags in BPF_OBJ_GET. This works
because OBJ_GET by default returns a read write mapping and libbpf
doesn't expose a way to override this behaviour for programs
and links.

Fixes: 70ed506c3bbc ("bpf: Introduce pinnable bpf_link abstraction")
Signed-off-by: Lorenz Bauer <redacted>
---
 kernel/bpf/inode.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c
index 1576ff331ee4..dc56237d6960 100644
--- a/kernel/bpf/inode.c
+++ b/kernel/bpf/inode.c

@@ -547,7 +547,7 @@ int bpf_obj_get_user(const char __user *pathname, int flags)
 	else if (type == BPF_TYPE_MAP)
 		ret = bpf_map_new_fd(raw, f_flags);
 	else if (type == BPF_TYPE_LINK)
-		ret = bpf_link_new_fd(raw);
+		ret = (f_flags != O_RDWR) ? -EINVAL : bpf_link_new_fd(raw);
 	else
 		return -ENOENT;
 

-- 
2.27.0

As for bpf_link, refuse creating a non-O_RDWR fd. Since program fds
currently don't allow modifications this is a precaution, not a
straight up bug fix.

Signed-off-by: Lorenz Bauer <redacted>
---
 kernel/bpf/inode.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c
index dc56237d6960..d2de2abec35b 100644
--- a/kernel/bpf/inode.c
+++ b/kernel/bpf/inode.c

@@ -543,7 +543,7 @@ int bpf_obj_get_user(const char __user *pathname, int flags)
 		return PTR_ERR(raw);
 
 	if (type == BPF_TYPE_PROG)
-		ret = bpf_prog_new_fd(raw);
+		ret = (f_flags != O_RDWR) ? -EINVAL : bpf_prog_new_fd(raw);
 	else if (type == BPF_TYPE_MAP)
 		ret = bpf_map_new_fd(raw, f_flags);
 	else if (type == BPF_TYPE_LINK)

-- 
2.27.0

On Fri, Mar 26, 2021 at 9:07 AM Lorenz Bauer [off-list ref] wrote:

For both patches, shall we do the check before bpf_obj_do_get(), which is a few
lines above?

Thanks,
Song

        if (type == BPF_TYPE_PROG)
-               ret = bpf_prog_new_fd(raw);
+               ret = (f_flags != O_RDWR) ? -EINVAL : bpf_prog_new_fd(raw);
        else if (type == BPF_TYPE_MAP)
                ret = bpf_map_new_fd(raw, f_flags);
        else if (type == BPF_TYPE_LINK)
--
2.27.0

On Fri, Mar 26, 2021 at 9:05 AM Lorenz Bauer [off-list ref] wrote:

Invoking BPF_OBJ_GET on a pinned bpf_link checks the path access
permissions based on file_flags, but the returned fd ignores flags.
This means that any user can acquire a "read-write" fd for a pinned
link with mode 0664 by invoking BPF_OBJ_GET with BPF_F_RDONLY in
file_flags. The fd can be used to invoke BPF_LINK_DETACH, etc.

Fix this by refusing non-O_RDWR flags in BPF_OBJ_GET. This works
because OBJ_GET by default returns a read write mapping and libbpf
doesn't expose a way to override this behaviour for programs
and links.

Fixes: 70ed506c3bbc ("bpf: Introduce pinnable bpf_link abstraction")
Signed-off-by: Lorenz Bauer <redacted>
---

LGTM.

Acked-by: Andrii Nakryiko <andrii@kernel.org>

On Fri, Mar 26, 2021 at 1:14 PM Song Liu [off-list ref] wrote:

On Fri, Mar 26, 2021 at 9:07 AM Lorenz Bauer [off-list ref] wrote:

quoted

As for bpf_link, refuse creating a non-O_RDWR fd. Since program fds
currently don't allow modifications this is a precaution, not a
straight up bug fix.

Signed-off-by: Lorenz Bauer <redacted>
---
 kernel/bpf/inode.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c
index dc56237d6960..d2de2abec35b 100644
--- a/kernel/bpf/inode.c
+++ b/kernel/bpf/inode.c

@@ -543,7 +543,7 @@ int bpf_obj_get_user(const char __user *pathname, int flags)
                return PTR_ERR(raw);

For both patches, shall we do the check before bpf_obj_do_get(), which is a few
lines above?

Map does use f_flags, so we need to let them through. Or did you mean
to do a (type != BPF_TYPE_MAP && f_flags != O_RDWR) check?

Either way is fine with me, so:

Acked-by: Andrii Nakryiko <andrii@kernel.org>

Thanks,
Song

quoted

        if (type == BPF_TYPE_PROG)
-               ret = bpf_prog_new_fd(raw);
+               ret = (f_flags != O_RDWR) ? -EINVAL : bpf_prog_new_fd(raw);
        else if (type == BPF_TYPE_MAP)
                ret = bpf_map_new_fd(raw, f_flags);
        else if (type == BPF_TYPE_LINK)
--
2.27.0

On Fri, 26 Mar 2021 at 20:14, Song Liu [off-list ref] wrote:

On Fri, Mar 26, 2021 at 9:07 AM Lorenz Bauer [off-list ref] wrote:

quoted

As for bpf_link, refuse creating a non-O_RDWR fd. Since program fds
currently don't allow modifications this is a precaution, not a
straight up bug fix.

Signed-off-by: Lorenz Bauer <redacted>
---
 kernel/bpf/inode.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c
index dc56237d6960..d2de2abec35b 100644
--- a/kernel/bpf/inode.c
+++ b/kernel/bpf/inode.c

@@ -543,7 +543,7 @@ int bpf_obj_get_user(const char __user *pathname, int flags)
                return PTR_ERR(raw);

For both patches, shall we do the check before bpf_obj_do_get(), which is a few
lines above?

type is filled in by bpf_obj_do_get, so we can't avoid calling it. As
Andrii mentions we need to allow flags for map.

Thanks,
Song

quoted

        if (type == BPF_TYPE_PROG)
-               ret = bpf_prog_new_fd(raw);
+               ret = (f_flags != O_RDWR) ? -EINVAL : bpf_prog_new_fd(raw);
        else if (type == BPF_TYPE_MAP)
                ret = bpf_map_new_fd(raw, f_flags);
        else if (type == BPF_TYPE_LINK)
--
2.27.0

-- 
Lorenz Bauer  |  Systems Engineer
6th Floor, County Hall/The Riverside Building, SE1 7PB, UK

www.cloudflare.com

On Fri, 26 Mar 2021 at 16:05, Lorenz Bauer [off-list ref] wrote:

Invoking BPF_OBJ_GET on a pinned bpf_link checks the path access
permissions based on file_flags, but the returned fd ignores flags.
This means that any user can acquire a "read-write" fd for a pinned
link with mode 0664 by invoking BPF_OBJ_GET with BPF_F_RDONLY in
file_flags. The fd can be used to invoke BPF_LINK_DETACH, etc.

Fix this by refusing non-O_RDWR flags in BPF_OBJ_GET. This works
because OBJ_GET by default returns a read write mapping and libbpf
doesn't expose a way to override this behaviour for programs
and links.

Hi Alexei and Daniel,

I think these two patches might have fallen through the cracks, could
you take a look? I'm not sure what the etiquette is around bumping a
set, so please let me know if you'd prefer me to send the patches with
acks included or something like that.

Best

-- 
Lorenz Bauer  |  Systems Engineer
6th Floor, County Hall/The Riverside Building, SE1 7PB, UK

www.cloudflare.com

On Wed, Mar 31, 2021 at 7:04 AM Lorenz Bauer [off-list ref] wrote:

On Fri, 26 Mar 2021 at 16:05, Lorenz Bauer [off-list ref] wrote:

quoted

Invoking BPF_OBJ_GET on a pinned bpf_link checks the path access
permissions based on file_flags, but the returned fd ignores flags.
This means that any user can acquire a "read-write" fd for a pinned
link with mode 0664 by invoking BPF_OBJ_GET with BPF_F_RDONLY in
file_flags. The fd can be used to invoke BPF_LINK_DETACH, etc.

Fix this by refusing non-O_RDWR flags in BPF_OBJ_GET. This works
because OBJ_GET by default returns a read write mapping and libbpf
doesn't expose a way to override this behaviour for programs
and links.

Hi Alexei and Daniel,

I think these two patches might have fallen through the cracks, could
you take a look? I'm not sure what the etiquette is around bumping a
set, so please let me know if you'd prefer me to send the patches with
acks included or something like that.

It is still in patchworks. I didn't have time to think it through.
Sorry for delay.

On Wed, Mar 31, 2021 at 7:04 AM Lorenz Bauer [off-list ref] wrote:

On Fri, 26 Mar 2021 at 16:05, Lorenz Bauer [off-list ref] wrote:

quoted

Invoking BPF_OBJ_GET on a pinned bpf_link checks the path access
permissions based on file_flags, but the returned fd ignores flags.
This means that any user can acquire a "read-write" fd for a pinned
link with mode 0664 by invoking BPF_OBJ_GET with BPF_F_RDONLY in
file_flags. The fd can be used to invoke BPF_LINK_DETACH, etc.

Fix this by refusing non-O_RDWR flags in BPF_OBJ_GET. This works
because OBJ_GET by default returns a read write mapping and libbpf
doesn't expose a way to override this behaviour for programs
and links.

Hi Alexei and Daniel,

I think these two patches might have fallen through the cracks, could
you take a look? I'm not sure what the etiquette is around bumping a
set, so please let me know if you'd prefer me to send the patches with
acks included or something like that.

Applied both. Thanks