[PATCH v3 0/8] Getting rid of get_unused_fd() / use O_CLOEXEC

[PATCH v3 0/8] Getting rid of get_unused_fd() / use O_CLOEXEC

[Yann Droneaud]

Hi,

With help from subsystem maintainers, I've managed to get some of
get_unused_fd() calls converted to use get_unused_fd_flags(O_CLOEXEC)
instead. [ANDROID][IB][VFIO]

KVM subsystem maintainers helped me to change calls to anon_inode_getfd()
to use O_CLOEXEC by default. [KVM]

Some maintainers applied my patches to convert get_unused_fd() to
get_unused_fd_flags(0) were using O_CLOEXEC wasn't possible without breaking ABI.
[SCTP][XFS]

So, still in the hope to get rid of get_unused_fd(), please find a another
attempt to remove get_unused_fd() macro and encourage subsystems to use
get_unused_fd_flags(O_CLOEXEC) or anon_inode_getfd(O_CLOEXEC) by default
were appropriate.

The patchset convert all calls to get_unused_fd()
to get_unused_fd_flags(0) before removing get_unused_fd() macro.

Without get_unused_fd() macro, more subsystems are likely to use
anon_inode_getfd() and be teached to provide an API that let userspace
choose the opening flags of the file descriptor.

Additionally I'm suggesting Industrial IO (IIO) subsystem to use
anon_inode_getfd(O_CLOEXEC): it's the only subsystem using anon_inode_getfd()
with a fixed set of flags not including O_CLOEXEC.

Not using O_CLOEXEC by default or not letting userspace provide the "open" flags
should be considered bad practice from security point of view:
in most case O_CLOEXEC must be used to not leak file descriptor across exec().
Using O_CLOEXEC by default when flags are not provided by userspace allows it
to choose, without race, if the file descriptor is going to be inherited
across exec().

Status:

In linux-next tag 20130906, they're currently:

- 22 calls to get_unused_fd_flags()     (+3)
     not counting get_unused_fd() and anon_inode_getfd()
-  7 calls to get_unused_fd()           (-3)
- 11 calls to anon_inode_getfd()        (0)

Changes from patchset v2 [PATCHSETv2]:

- android/sw_sync: use get_unused_fd_flags(O_CLOEXEC) instead of get_unused_fd()
  DROPPED: applied upstream

- android/sync: use get_unused_fd_flags(O_CLOEXEC) instead of get_unused_fd()
  DROPPED: applied upstream

- vfio: use get_unused_fd_flags(0) instead of get_unused_fd()
  DROPPED: applied upstream.
  Additionally subsystem maintainer applied another patch on top
  to set the flags to O_CLOEXEC.

- industrialio: use anon_inode_getfd() with O_CLOEXEC flag
  NEW: propose to use O_CLOEXEC as default flag.

Links:

[ANDROID]
  http://lkml.kernel.org/r/CACSP8SjXGMk2_kX_+RgzqqQwqKernvF1Wt3K5tw991W5dfAnCA@mail.gmail.com
  http://lkml.kernel.org/r/CACSP8SjZcpcpEtQHzcGYhf-MP7QGo0XpN7-uN7rmD=vNtopG=w@mail.gmail.com

[IB]
  http://lkml.kernel.org/r/CAL1RGDXV1_BjSLrQDFdVQ1_D75+bMtOtikHOUp8VBiy_OJUf=w@mail.gmail.com

[VFIO]
  http://lkml.kernel.org/r/20130822171744.1297.13711.stgit@bling.home

[KVM]
  http://lkml.kernel.org/r/5219A8FC.8090307@redhat.com
  http://lkml.kernel.org/r/3557EF65-4327-4DAE-999A-B0EE13C433F5@suse.de
  http://lkml.kernel.org/r/20130826102023.GA8218@redhat.com

[SCTP]
  http://lkml.kernel.org/r/51D312E8.6090702@gmail.com
  http://lkml.kernel.org/r/20130702.161428.1703028286206350504.davem@davemloft.net

[XFS]
  http://lkml.kernel.org/r/20130709205321.GV20932@sgi.com

[PATCHSETv2]
  http://lkml.kernel.org/r/cover.1376327678.git.ydroneaud@opteya.com

Yann Droneaud (8):
  ia64: use get_unused_fd_flags(0) instead of get_unused_fd()
  ppc/cell: use get_unused_fd_flags(0) instead of get_unused_fd()
  binfmt_misc: use get_unused_fd_flags(0) instead of get_unused_fd()
  file: use get_unused_fd_flags(0) instead of get_unused_fd()
  fanotify: use get_unused_fd_flags(0) instead of get_unused_fd()
  events: use get_unused_fd_flags(0) instead of get_unused_fd()
  file: remove get_unused_fd()
  industrialio: use anon_inode_getfd() with O_CLOEXEC flag

 arch/ia64/kernel/perfmon.c                | 2 +-
 arch/powerpc/platforms/cell/spufs/inode.c | 4 ++--
 drivers/iio/industrialio-event.c          | 2 +-
 fs/binfmt_misc.c                          | 2 +-
 fs/file.c                                 | 2 +-
 fs/notify/fanotify/fanotify_user.c        | 2 +-
 include/linux/file.h                      | 1 -
 kernel/events/core.c                      | 2 +-
 8 files changed, 8 insertions(+), 9 deletions(-)

-- 
1.8.3.1

[PATCH v3 1/8] ia64: use get_unused_fd_flags(0) instead of get_unused_fd()

[Yann Droneaud]

Macro get_unused_fd() is used to allocate a file descriptor with
default flags. Those default flags (0) can be "unsafe":
O_CLOEXEC must be used by default to not leak file descriptor
across exec().

Instead of macro get_unused_fd(), functions anon_inode_getfd()
or get_unused_fd_flags() should be used with flags given by userspace.
If not possible, flags should be set to O_CLOEXEC to provide userspace
with a default safe behavor.

In a further patch, get_unused_fd() will be removed so that
new code start using anon_inode_getfd() or get_unused_fd_flags()
with correct flags.

This patch replaces calls to get_unused_fd() with equivalent call to
get_unused_fd_flags(0) to preserve current behavor for existing code.

The hard coded flag value (0) should be reviewed on a per-subsystem basis,
and, if possible, set to O_CLOEXEC.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 arch/ia64/kernel/perfmon.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/ia64/kernel/perfmon.c b/arch/ia64/kernel/perfmon.c
index 5a9ff1c..64757c1 100644
--- a/arch/ia64/kernel/perfmon.c
+++ b/arch/ia64/kernel/perfmon.c
@@ -2668,7 +2668,7 @@ pfm_context_create(pfm_context_t *ctx, void *arg, int count, struct pt_regs *reg
 
 	ret = -ENOMEM;
 
-	fd = get_unused_fd();
+	fd = get_unused_fd_flags(0);
 	if (fd < 0)
 		return fd;
 
-- 
1.8.3.1

[PATCH v3 2/8] ppc/cell: use get_unused_fd_flags(0) instead of get_unused_fd()

[Yann Droneaud]

Macro get_unused_fd() is used to allocate a file descriptor with
default flags. Those default flags (0) can be "unsafe":
O_CLOEXEC must be used by default to not leak file descriptor
across exec().

Instead of macro get_unused_fd(), functions anon_inode_getfd()
or get_unused_fd_flags() should be used with flags given by userspace.
If not possible, flags should be set to O_CLOEXEC to provide userspace
with a default safe behavor.

In a further patch, get_unused_fd() will be removed so that
new code start using anon_inode_getfd() or get_unused_fd_flags()
with correct flags.

This patch replaces calls to get_unused_fd() with equivalent call to
get_unused_fd_flags(0) to preserve current behavor for existing code.

The hard coded flag value (0) should be reviewed on a per-subsystem basis,
and, if possible, set to O_CLOEXEC.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 arch/powerpc/platforms/cell/spufs/inode.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/platforms/cell/spufs/inode.c b/arch/powerpc/platforms/cell/spufs/inode.c
index 87ba7cf..51effce 100644
--- a/arch/powerpc/platforms/cell/spufs/inode.c
+++ b/arch/powerpc/platforms/cell/spufs/inode.c
@@ -301,7 +301,7 @@ static int spufs_context_open(struct path *path)
 	int ret;
 	struct file *filp;
 
-	ret = get_unused_fd();
+	ret = get_unused_fd_flags(0);
 	if (ret < 0)
 		return ret;
 
@@ -518,7 +518,7 @@ static int spufs_gang_open(struct path *path)
 	int ret;
 	struct file *filp;
 
-	ret = get_unused_fd();
+	ret = get_unused_fd_flags(0);
 	if (ret < 0)
 		return ret;
 
-- 
1.8.3.1

[PATCH v3 3/8] binfmt_misc: use get_unused_fd_flags(0) instead of get_unused_fd()

[Yann Droneaud]

Macro get_unused_fd() is used to allocate a file descriptor with
default flags. Those default flags (0) can be "unsafe":
O_CLOEXEC must be used by default to not leak file descriptor
across exec().

Instead of macro get_unused_fd(), functions anon_inode_getfd()
or get_unused_fd_flags() should be used with flags given by userspace.
If not possible, flags should be set to O_CLOEXEC to provide userspace
with a default safe behavor.

In a further patch, get_unused_fd() will be removed so that
new code start using anon_inode_getfd() or get_unused_fd_flags()
with correct flags.

This patch replaces calls to get_unused_fd() with equivalent call to
get_unused_fd_flags(0) to preserve current behavor for existing code.

The hard coded flag value (0) should be reviewed on a per-subsystem basis,
and, if possible, set to O_CLOEXEC.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 fs/binfmt_misc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/binfmt_misc.c b/fs/binfmt_misc.c
index 1c740e1..052f6dc 100644
--- a/fs/binfmt_misc.c
+++ b/fs/binfmt_misc.c
@@ -138,7 +138,7 @@ static int load_misc_binary(struct linux_binprm *bprm)
 		/* if the binary should be opened on behalf of the
 		 * interpreter than keep it open and assign descriptor
 		 * to it */
- 		fd_binary = get_unused_fd();
+		fd_binary = get_unused_fd_flags(0);
  		if (fd_binary < 0) {
  			retval = fd_binary;
  			goto _ret;
-- 
1.8.3.1

[PATCH v3 4/8] file: use get_unused_fd_flags(0) instead of get_unused_fd()

[Yann Droneaud]

Macro get_unused_fd() is used to allocate a file descriptor with
default flags. Those default flags (0) can be "unsafe":
O_CLOEXEC must be used by default to not leak file descriptor
across exec().

Instead of macro get_unused_fd(), functions anon_inode_getfd()
or get_unused_fd_flags() should be used with flags given by userspace.
If not possible, flags should be set to O_CLOEXEC to provide userspace
with a default safe behavor.

In a further patch, get_unused_fd() will be removed so that
new code start using anon_inode_getfd() or get_unused_fd_flags()
with correct flags.

This patch replaces calls to get_unused_fd() with equivalent call to
get_unused_fd_flags(0) to preserve current behavor for existing code.

The hard coded flag value (0) should be reviewed on a per-subsystem basis,
and, if possible, set to O_CLOEXEC.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 fs/file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/file.c b/fs/file.c
index 4a78f98..1420d28 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -897,7 +897,7 @@ SYSCALL_DEFINE1(dup, unsigned int, fildes)
 	struct file *file = fget_raw(fildes);
 
 	if (file) {
-		ret = get_unused_fd();
+		ret = get_unused_fd_flags(0);
 		if (ret >= 0)
 			fd_install(ret, file);
 		else
-- 
1.8.3.1

[PATCH v3 5/8] fanotify: use get_unused_fd_flags(0) instead of get_unused_fd()

[Yann Droneaud]

Macro get_unused_fd() is used to allocate a file descriptor with
default flags. Those default flags (0) can be "unsafe":
O_CLOEXEC must be used by default to not leak file descriptor
across exec().

Instead of macro get_unused_fd(), functions anon_inode_getfd()
or get_unused_fd_flags() should be used with flags given by userspace.
If not possible, flags should be set to O_CLOEXEC to provide userspace
with a default safe behavor.

In a further patch, get_unused_fd() will be removed so that
new code start using anon_inode_getfd() or get_unused_fd_flags()
with correct flags.

This patch replaces calls to get_unused_fd() with equivalent call to
get_unused_fd_flags(0) to preserve current behavor for existing code.

The hard coded flag value (0) should be reviewed on a per-subsystem basis,
and, if possible, set to O_CLOEXEC.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 fs/notify/fanotify/fanotify_user.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
index e44cb64..644b9a7 100644
--- a/fs/notify/fanotify/fanotify_user.c
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -69,7 +69,7 @@ static int create_fd(struct fsnotify_group *group,
 
 	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
 
-	client_fd = get_unused_fd();
+	client_fd = get_unused_fd_flags(0);
 	if (client_fd < 0)
 		return client_fd;
 
-- 
1.8.3.1

[PATCH v3 6/8] events: use get_unused_fd_flags(0) instead of get_unused_fd()

[Yann Droneaud]

Macro get_unused_fd() is used to allocate a file descriptor with
default flags. Those default flags (0) can be "unsafe":
O_CLOEXEC must be used by default to not leak file descriptor
across exec().

Instead of macro get_unused_fd(), functions anon_inode_getfd()
or get_unused_fd_flags() should be used with flags given by userspace.
If not possible, flags should be set to O_CLOEXEC to provide userspace
with a default safe behavor.

In a further patch, get_unused_fd() will be removed so that
new code start using anon_inode_getfd() or get_unused_fd_flags()
with correct flags.

This patch replaces calls to get_unused_fd() with equivalent call to
get_unused_fd_flags(0) to preserve current behavor for existing code.

The hard coded flag value (0) should be reviewed on a per-subsystem basis,
and, if possible, set to O_CLOEXEC.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 kernel/events/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 2207efc..b12fdd3 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -6935,7 +6935,7 @@ SYSCALL_DEFINE5(perf_event_open,
 	if ((flags & PERF_FLAG_PID_CGROUP) && (pid == -1 || cpu == -1))
 		return -EINVAL;
 
-	event_fd = get_unused_fd();
+	event_fd = get_unused_fd_flags(0);
 	if (event_fd < 0)
 		return event_fd;
 
-- 
1.8.3.1

[PATCH v3 7/8] file: remove get_unused_fd()

[Yann Droneaud]

Macro get_unused_fd() allocates a file descriptor without O_CLOEXEC flag.

This can be seen as an unsafe default: in most case O_CLOEXEC
must be used to not leak file descriptor across exec().

Using O_CLOEXEC by default allows userspace to choose, without race,
if the file descriptor is going to be inherited across exec().

This patch removes get_unused_fd() so that newer kernel code use
anon_inode_getfd() or get_unused_fd_flags() with flags provided
by userspace. If flags cannot be given by userspace,
O_CLOEXEC must be the default flag.

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 include/linux/file.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/include/linux/file.h b/include/linux/file.h
index cbacf4f..8666002 100644
--- a/include/linux/file.h
+++ b/include/linux/file.h
@@ -63,7 +63,6 @@ extern void set_close_on_exec(unsigned int fd, int flag);
 extern bool get_close_on_exec(unsigned int fd);
 extern void put_filp(struct file *);
 extern int get_unused_fd_flags(unsigned flags);
-#define get_unused_fd() get_unused_fd_flags(0)
 extern void put_unused_fd(unsigned int fd);
 
 extern void fd_install(unsigned int fd, struct file *file);
-- 
1.8.3.1

[PATCH v3 8/8] industrialio: use anon_inode_getfd() with O_CLOEXEC flag

[Yann Droneaud]

IIO uses anon_inode_get() to allocate file descriptors as part
of its ioctls. But those ioctls are lacking a flag argument
allowing userspace to choose options for the newly opened file
descriptor.

In such case it's advised to use O_CLOEXEC by default so that
userspace is allowed to choose, without race, if the file descriptor
is going to be inherited across exec().

KVM usage of anon_inode_getfd() was fixed in a previous patchset [1],
so IIO is the only subsystem using anon_inode_getfd() with a fixed set
of flags not including O_CLOEXEC.

This patch set O_CLOEXEC flag on the event file descriptor created
with anon_inode_getfd() to not leak file descriptors across exec().

Links:

- Secure File Descriptor Handling (Ulrich Drepper, 2008)
  http://udrepper.livejournal.com/20407.html

- Excuse me son, but your code is leaking !!! (Dan Walsh, March 2012)
  http://danwalsh.livejournal.com/53603.html

- [1] kvm: use anon_inode_getfd() with O_CLOEXEC flag
  http://lkml.kernel.org/r/cover.1377372576.git.ydroneaud@opteya.com

Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
---
 drivers/iio/industrialio-event.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/iio/industrialio-event.c b/drivers/iio/industrialio-event.c
index 10aa9ef..2390e3d 100644
--- a/drivers/iio/industrialio-event.c
+++ b/drivers/iio/industrialio-event.c
@@ -159,7 +159,7 @@ int iio_event_getfd(struct iio_dev *indio_dev)
 	}
 	spin_unlock_irq(&ev_int->wait.lock);
 	fd = anon_inode_getfd("iio:event",
-				&iio_event_chrdev_fileops, ev_int, O_RDONLY);
+				&iio_event_chrdev_fileops, ev_int, O_RDONLY | O_CLOEXEC);
 	if (fd < 0) {
 		spin_lock_irq(&ev_int->wait.lock);
 		__clear_bit(IIO_BUSY_BIT_POS, &ev_int->flags);
-- 
1.8.3.1

Re: [PATCH v3 8/8] industrialio: use anon_inode_getfd() with O_CLOEXEC flag

[Jonathan Cameron]

On 09/06/13 11:39, Yann Droneaud wrote:
> IIO uses anon_inode_get() to allocate file descriptors as part
> of its ioctls. But those ioctls are lacking a flag argument
> allowing userspace to choose options for the newly opened file
> descriptor.
> 
> In such case it's advised to use O_CLOEXEC by default so that
> userspace is allowed to choose, without race, if the file descriptor
> is going to be inherited across exec().
> 
> KVM usage of anon_inode_getfd() was fixed in a previous patchset [1],
> so IIO is the only subsystem using anon_inode_getfd() with a fixed set
> of flags not including O_CLOEXEC.
> 
> This patch set O_CLOEXEC flag on the event file descriptor created
> with anon_inode_getfd() to not leak file descriptors across exec().
> 
> Links:
> 
> - Secure File Descriptor Handling (Ulrich Drepper, 2008)
>   http://udrepper.livejournal.com/20407.html
> 
> - Excuse me son, but your code is leaking !!! (Dan Walsh, March 2012)
>   http://danwalsh.livejournal.com/53603.html
> 
> - [1] kvm: use anon_inode_getfd() with O_CLOEXEC flag
>   http://lkml.kernel.org/r/cover.1377372576.git.ydroneaud@opteya.com
> 
> Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
> Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com

Thanks Yann.  I had no idea about this issue but your well supported description
above made this easy to review.

Applied to the togreg branch of git://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio.git
as it's not a regression but if there is support for pushing this back to stable
I personally am not against it.

Jonathan
>
 ---
>  drivers/iio/industrialio-event.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/iio/industrialio-event.c b/drivers/iio/industrialio-event.c
> index 10aa9ef..2390e3d 100644
> --- a/drivers/iio/industrialio-event.c
> +++ b/drivers/iio/industrialio-event.c
> @@ -159,7 +159,7 @@ int iio_event_getfd(struct iio_dev *indio_dev)
>  	}
>  	spin_unlock_irq(&ev_int->wait.lock);
>  	fd = anon_inode_getfd("iio:event",
> -				&iio_event_chrdev_fileops, ev_int, O_RDONLY);
> +				&iio_event_chrdev_fileops, ev_int, O_RDONLY | O_CLOEXEC);
>  	if (fd < 0) {
>  		spin_lock_irq(&ev_int->wait.lock);
>  		__clear_bit(IIO_BUSY_BIT_POS, &ev_int->flags);
> 

Re: [PATCH v3 8/8] industrialio: use anon_inode_getfd() with O_CLOEXEC flag

[Yann Droneaud]

Hi,

Le dimanche 15 septembre 2013 à 17:26 +0100, Jonathan Cameron a écrit :
> On 09/06/13 11:39, Yann Droneaud wrote:

[...]

> >   http://lkml.kernel.org/r/cover.1377372576.git.ydroneaud@opteya.com
> > 
> > Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
> > Link: http://lkml.kernel.org/r/cover.1378460926.git.ydroneaud@opteya.com
> 
> Thanks Yann.  I had no idea about this issue but your well supported description
> above made this easy to review.
> 
> Applied to the togreg branch of git://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio.git
> as it's not a regression but if there is support for pushing this back to stable
> I personally am not against it.
> 

It's not a -stable materiel. So let it flow to v3.12 or v3.13.

Thanks for the review.

Regards.

-- 
Yann Droneaud
OPTEYA