mbox series

[v5,0/4] landlock: truncate support

Message ID 20220817203006.21769-1-gnoack3000@gmail.com (mailing list archive)
Headers show
Series landlock: truncate support | expand

Message

Günther Noack Aug. 17, 2022, 8:30 p.m. UTC
The goal of these patches is to work towards a more complete coverage
of file system operations that are restrictable with Landlock.

The known set of currently unsupported file system operations in
Landlock is described at [1]. Out of the operations listed there,
truncate is the only one that modifies file contents, so these patches
should make it possible to prevent the direct modification of file
contents with Landlock.

The patch introduces the truncation restriction feature as an
additional bit in the access_mask_t bitmap, in line with the existing
supported operations.

The truncation flag covers both the truncate(2) and ftruncate(2)
families of syscalls, as well as open(2) with the O_TRUNC flag.
This includes usages of creat() in the case where existing regular
files are overwritten.

Apart from Landlock, file truncation can also be restricted using
seccomp-bpf, but it is more difficult to use (requires BPF, requires
keeping up-to-date syscall lists) and it is not configurable by file
hierarchy, as Landlock is. The simplicity and flexibility of the
Landlock approach makes it worthwhile adding.

While it's possible to use the "write file" and "truncate" rights
independent of each other, it simplifies the mental model for
userspace callers to always use them together.

Specifically, the following behaviours might be surprising for users
when using these independently:

 * The commonly creat() syscall requires the truncate right when
   overwriting existing files, as it is equivalent to open(2) with
   O_TRUNC|O_CREAT|O_WRONLY.
 * The "write file" right is not always required to truncate a file,
   even through the open(2) syscall (when using O_RDONLY|O_TRUNC).

Nevertheless, keeping the two flags separate is the correct approach
to guarantee backwards compatibility for existing Landlock users.

These patches are based on version 6.0-rc1.

Best regards,
Günther

[1] https://docs.kernel.org/userspace-api/landlock.html#filesystem-flags

Past discussions:
V1: https://lore.kernel.org/all/20220707200612.132705-1-gnoack3000@gmail.com/
V2: https://lore.kernel.org/all/20220712211405.14705-1-gnoack3000@gmail.com/
V3: https://lore.kernel.org/all/20220804193746.9161-1-gnoack3000@gmail.com/
V4: https://lore.kernel.org/all/20220814192603.7387-1-gnoack3000@gmail.com/

Changelog:

V5:
* Documentation
  * Fix wording in userspace-api headers and in landlock.rst.
  * Move the truncation limitation section one to the bottom.
  * Move all .rst changes into the documentation commit.
* selftests
  * Remove _metadata argument from helpers where it became unnecessary.
  * Open writable file descriptors at the top of both tests, before Landlock
    is enabled, to exercise ftruncate() independently from open().
  * Simplify test_ftruncate and decouple it from exercising open().
  * test_creat(): Return errno on close() failure (it does not conflict).
  * Fix /* comment style */
  * Reorder blocks of EXPECT_EQ checks to be consistent within a test.
  * Add missing |O_TRUNC to a check in one test.
  * Put the truncate_unhandled test before the other.

V4:
 * Documentation
   * Clarify wording and syntax as discussed in review.
   * Use a less confusing error message in the example.
 * selftests:
   * Stop using ASSERT_EQ in test helpers, return EBADFD instead.
     (This is an intentionally uncommon error code, so that the source
     of the error is clear and the test can distinguish test setup
     failures from failures in the actual system call under test.)
 * samples/Documentation:
   * Use additional clarifying comments in the kernel backwards
     compatibility logic.

V3:
 * selftests:
   * Explicitly test ftruncate with readonly file descriptors
     (returns EINVAL).
   * Extract test_ftruncate, test_truncate, test_creat helpers,
     which simplified the previously mixed usage of EXPECT/ASSERT.
   * Test creat() behaviour as part of the big truncation test.
   * Stop testing the truncate64(2) and ftruncate64(2) syscalls.
     This simplifies the tests a bit. The kernel implementations are the
     same as for truncate(2) and ftruncate(2), so there is little benefit
     from testing them exhaustively. (We aren't testing all open(2)
     variants either.)
 * samples/landlock/sandboxer.c:
   * Use switch() to implement best effort mode.
 * Documentation:
   * Give more background on surprising truncation behaviour.
   * Use switch() in the example too, to stay in-line with the sample tool.
   * Small fixes in header file to address previous comments.
* misc:
  * Fix some typos and const usages.

V2:
 * Documentation: Mention the truncation flag where needed.
 * Documentation: Point out connection between truncation and file writing.
 * samples: Add file truncation to the landlock/sandboxer.c sample tool.
 * selftests: Exercise open(2) with O_TRUNC and creat(2) exhaustively.
 * selftests: Exercise truncation syscalls when the truncate right
   is not handled by Landlock.

Günther Noack (4):
  landlock: Support file truncation
  selftests/landlock: Selftests for file truncation support
  samples/landlock: Extend sample tool to support
    LANDLOCK_ACCESS_FS_TRUNCATE
  landlock: Document Landlock's file truncation support

 Documentation/userspace-api/landlock.rst     |  52 +++-
 include/uapi/linux/landlock.h                |  17 +-
 samples/landlock/sandboxer.c                 |  23 +-
 security/landlock/fs.c                       |   9 +-
 security/landlock/limits.h                   |   2 +-
 security/landlock/syscalls.c                 |   2 +-
 tools/testing/selftests/landlock/base_test.c |   2 +-
 tools/testing/selftests/landlock/fs_test.c   | 257 ++++++++++++++++++-
 8 files changed, 336 insertions(+), 28 deletions(-)


base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
--
2.37.2

Comments

Mickaël Salaün Sept. 1, 2022, 5:10 p.m. UTC | #1
Hmm, I think there is an issue with this series. Landlock only enforces 
restrictions at open time or when dealing with user-supplied file paths 
(relative or absolute). The use of the path_truncate hook in this series 
doesn't distinguish between file descriptor from before the current 
sandbox or from after being sandboxed. For instance, if a file 
descriptor is received through a unix socket, it is assumed that this is 
legitimate and no Landlock restriction apply on it, which is not the 
case with this series anymore. It is the same for files opened before 
the process sandbox itself.

To be able to follow the current semantic, I think we should control the 
truncate access at open time (or when dealing with a user-supplied path) 
but not on any file descriptor as it is currently done.


On 17/08/2022 22:30, Günther Noack wrote:
> The goal of these patches is to work towards a more complete coverage
> of file system operations that are restrictable with Landlock.
> 
> The known set of currently unsupported file system operations in
> Landlock is described at [1]. Out of the operations listed there,
> truncate is the only one that modifies file contents, so these patches
> should make it possible to prevent the direct modification of file
> contents with Landlock.
> 
> The patch introduces the truncation restriction feature as an
> additional bit in the access_mask_t bitmap, in line with the existing
> supported operations.
> 
> The truncation flag covers both the truncate(2) and ftruncate(2)
> families of syscalls, as well as open(2) with the O_TRUNC flag.
> This includes usages of creat() in the case where existing regular
> files are overwritten.
> 
> Apart from Landlock, file truncation can also be restricted using
> seccomp-bpf, but it is more difficult to use (requires BPF, requires
> keeping up-to-date syscall lists) and it is not configurable by file
> hierarchy, as Landlock is. The simplicity and flexibility of the
> Landlock approach makes it worthwhile adding.
> 
> While it's possible to use the "write file" and "truncate" rights
> independent of each other, it simplifies the mental model for
> userspace callers to always use them together.
> 
> Specifically, the following behaviours might be surprising for users
> when using these independently:
> 
>   * The commonly creat() syscall requires the truncate right when
>     overwriting existing files, as it is equivalent to open(2) with
>     O_TRUNC|O_CREAT|O_WRONLY.
>   * The "write file" right is not always required to truncate a file,
>     even through the open(2) syscall (when using O_RDONLY|O_TRUNC).
> 
> Nevertheless, keeping the two flags separate is the correct approach
> to guarantee backwards compatibility for existing Landlock users.
> 
> These patches are based on version 6.0-rc1.
> 
> Best regards,
> Günther
> 
> [1] https://docs.kernel.org/userspace-api/landlock.html#filesystem-flags
> 
> Past discussions:
> V1: https://lore.kernel.org/all/20220707200612.132705-1-gnoack3000@gmail.com/
> V2: https://lore.kernel.org/all/20220712211405.14705-1-gnoack3000@gmail.com/
> V3: https://lore.kernel.org/all/20220804193746.9161-1-gnoack3000@gmail.com/
> V4: https://lore.kernel.org/all/20220814192603.7387-1-gnoack3000@gmail.com/
> 
> Changelog:
> 
> V5:
> * Documentation
>    * Fix wording in userspace-api headers and in landlock.rst.
>    * Move the truncation limitation section one to the bottom.
>    * Move all .rst changes into the documentation commit.
> * selftests
>    * Remove _metadata argument from helpers where it became unnecessary.
>    * Open writable file descriptors at the top of both tests, before Landlock
>      is enabled, to exercise ftruncate() independently from open().
>    * Simplify test_ftruncate and decouple it from exercising open().
>    * test_creat(): Return errno on close() failure (it does not conflict).
>    * Fix /* comment style */
>    * Reorder blocks of EXPECT_EQ checks to be consistent within a test.
>    * Add missing |O_TRUNC to a check in one test.
>    * Put the truncate_unhandled test before the other.
> 
> V4:
>   * Documentation
>     * Clarify wording and syntax as discussed in review.
>     * Use a less confusing error message in the example.
>   * selftests:
>     * Stop using ASSERT_EQ in test helpers, return EBADFD instead.
>       (This is an intentionally uncommon error code, so that the source
>       of the error is clear and the test can distinguish test setup
>       failures from failures in the actual system call under test.)
>   * samples/Documentation:
>     * Use additional clarifying comments in the kernel backwards
>       compatibility logic.
> 
> V3:
>   * selftests:
>     * Explicitly test ftruncate with readonly file descriptors
>       (returns EINVAL).
>     * Extract test_ftruncate, test_truncate, test_creat helpers,
>       which simplified the previously mixed usage of EXPECT/ASSERT.
>     * Test creat() behaviour as part of the big truncation test.
>     * Stop testing the truncate64(2) and ftruncate64(2) syscalls.
>       This simplifies the tests a bit. The kernel implementations are the
>       same as for truncate(2) and ftruncate(2), so there is little benefit
>       from testing them exhaustively. (We aren't testing all open(2)
>       variants either.)
>   * samples/landlock/sandboxer.c:
>     * Use switch() to implement best effort mode.
>   * Documentation:
>     * Give more background on surprising truncation behaviour.
>     * Use switch() in the example too, to stay in-line with the sample tool.
>     * Small fixes in header file to address previous comments.
> * misc:
>    * Fix some typos and const usages.
> 
> V2:
>   * Documentation: Mention the truncation flag where needed.
>   * Documentation: Point out connection between truncation and file writing.
>   * samples: Add file truncation to the landlock/sandboxer.c sample tool.
>   * selftests: Exercise open(2) with O_TRUNC and creat(2) exhaustively.
>   * selftests: Exercise truncation syscalls when the truncate right
>     is not handled by Landlock.
> 
> Günther Noack (4):
>    landlock: Support file truncation
>    selftests/landlock: Selftests for file truncation support
>    samples/landlock: Extend sample tool to support
>      LANDLOCK_ACCESS_FS_TRUNCATE
>    landlock: Document Landlock's file truncation support
> 
>   Documentation/userspace-api/landlock.rst     |  52 +++-
>   include/uapi/linux/landlock.h                |  17 +-
>   samples/landlock/sandboxer.c                 |  23 +-
>   security/landlock/fs.c                       |   9 +-
>   security/landlock/limits.h                   |   2 +-
>   security/landlock/syscalls.c                 |   2 +-
>   tools/testing/selftests/landlock/base_test.c |   2 +-
>   tools/testing/selftests/landlock/fs_test.c   | 257 ++++++++++++++++++-
>   8 files changed, 336 insertions(+), 28 deletions(-)
> 
> 
> base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> --
> 2.37.2
Günther Noack Sept. 2, 2022, 5:32 a.m. UTC | #2
On Thu, Sep 01, 2022 at 07:10:38PM +0200, Mickaël Salaün wrote:
> Hmm, I think there is an issue with this series. Landlock only enforces
> restrictions at open time or when dealing with user-supplied file paths
> (relative or absolute).

Argh, ok. That sounds like a desirable property, although it would
mean reworking the patch set.

> The use of the path_truncate hook in this series
> doesn't distinguish between file descriptor from before the current sandbox
> or from after being sandboxed. For instance, if a file descriptor is
> received through a unix socket, it is assumed that this is legitimate and no
> Landlock restriction apply on it, which is not the case with this series
> anymore. It is the same for files opened before the process sandbox itself.
>
> To be able to follow the current semantic, I think we should control the
> truncate access at open time (or when dealing with a user-supplied path) but
> not on any file descriptor as it is currently done.

OK - so let me try to make a constructive proposal. We have previously
identified a few operations where a truncation happens, and I would
propose that the following Landlock rights should be needed for these:

* truncate() (operating on char *path): Require LL_ACCESS_FS_TRUNCATE
* ftruncate() (operating on fd): No Landlock rights required
* open() for reading with O_TRUNC: Require LL_ACCESS_FS_TRUNCATE
* open() for writing with O_TRUNC: Require LL_ACCESS_FS_WRITE_FILE

The rationale goes as follows:

* ftruncate() is already adequately protected by the
  LL_ACCESS_FS_WRITE_FILE right. ftruncate is only permitted on fds
  that are open for writing.
* truncate() is not Landlock-restrictable in Landlock ABI v1,
  so needs to be covered by the new truncate right.
* open() for reading with O_TRUNC is also not Landlock-restrictable in
  Landlock ABI v1, so needs to be covered by the new truncate right.
* open() for writing with O_TRUNC is also not Landlock-restrictable in
  Landlock ABI v1. BUT: A caller who can open the file for writing
  will also be able to ftruncate it - so it doesn't really make sense
  to ask for a different Landlock right here.

Does that approach make sense to you?

I think in terms of changs required for it, it sounds like it would
require a change to the path_truncate LSM hook to distinguish the
cases above.

Do you want a new patch on top of the existing one, or should I rather
create a new version of the old truncate patch set?

--Günther

> On 17/08/2022 22:30, Günther Noack wrote:
> > The goal of these patches is to work towards a more complete coverage
> > of file system operations that are restrictable with Landlock.
> >
> > The known set of currently unsupported file system operations in
> > Landlock is described at [1]. Out of the operations listed there,
> > truncate is the only one that modifies file contents, so these patches
> > should make it possible to prevent the direct modification of file
> > contents with Landlock.
> >
> > The patch introduces the truncation restriction feature as an
> > additional bit in the access_mask_t bitmap, in line with the existing
> > supported operations.
> >
> > The truncation flag covers both the truncate(2) and ftruncate(2)
> > families of syscalls, as well as open(2) with the O_TRUNC flag.
> > This includes usages of creat() in the case where existing regular
> > files are overwritten.
> >
> > Apart from Landlock, file truncation can also be restricted using
> > seccomp-bpf, but it is more difficult to use (requires BPF, requires
> > keeping up-to-date syscall lists) and it is not configurable by file
> > hierarchy, as Landlock is. The simplicity and flexibility of the
> > Landlock approach makes it worthwhile adding.
> >
> > While it's possible to use the "write file" and "truncate" rights
> > independent of each other, it simplifies the mental model for
> > userspace callers to always use them together.
> >
> > Specifically, the following behaviours might be surprising for users
> > when using these independently:
> >
> >   * The commonly creat() syscall requires the truncate right when
> >     overwriting existing files, as it is equivalent to open(2) with
> >     O_TRUNC|O_CREAT|O_WRONLY.
> >   * The "write file" right is not always required to truncate a file,
> >     even through the open(2) syscall (when using O_RDONLY|O_TRUNC).
> >
> > Nevertheless, keeping the two flags separate is the correct approach
> > to guarantee backwards compatibility for existing Landlock users.
> >
> > These patches are based on version 6.0-rc1.
> >
> > Best regards,
> > Günther
> >
> > [1] https://docs.kernel.org/userspace-api/landlock.html#filesystem-flags
> >
> > Past discussions:
> > V1: https://lore.kernel.org/all/20220707200612.132705-1-gnoack3000@gmail.com/
> > V2: https://lore.kernel.org/all/20220712211405.14705-1-gnoack3000@gmail.com/
> > V3: https://lore.kernel.org/all/20220804193746.9161-1-gnoack3000@gmail.com/
> > V4: https://lore.kernel.org/all/20220814192603.7387-1-gnoack3000@gmail.com/
> >
> > Changelog:
> >
> > V5:
> > * Documentation
> >    * Fix wording in userspace-api headers and in landlock.rst.
> >    * Move the truncation limitation section one to the bottom.
> >    * Move all .rst changes into the documentation commit.
> > * selftests
> >    * Remove _metadata argument from helpers where it became unnecessary.
> >    * Open writable file descriptors at the top of both tests, before Landlock
> >      is enabled, to exercise ftruncate() independently from open().
> >    * Simplify test_ftruncate and decouple it from exercising open().
> >    * test_creat(): Return errno on close() failure (it does not conflict).
> >    * Fix /* comment style */
> >    * Reorder blocks of EXPECT_EQ checks to be consistent within a test.
> >    * Add missing |O_TRUNC to a check in one test.
> >    * Put the truncate_unhandled test before the other.
> >
> > V4:
> >   * Documentation
> >     * Clarify wording and syntax as discussed in review.
> >     * Use a less confusing error message in the example.
> >   * selftests:
> >     * Stop using ASSERT_EQ in test helpers, return EBADFD instead.
> >       (This is an intentionally uncommon error code, so that the source
> >       of the error is clear and the test can distinguish test setup
> >       failures from failures in the actual system call under test.)
> >   * samples/Documentation:
> >     * Use additional clarifying comments in the kernel backwards
> >       compatibility logic.
> >
> > V3:
> >   * selftests:
> >     * Explicitly test ftruncate with readonly file descriptors
> >       (returns EINVAL).
> >     * Extract test_ftruncate, test_truncate, test_creat helpers,
> >       which simplified the previously mixed usage of EXPECT/ASSERT.
> >     * Test creat() behaviour as part of the big truncation test.
> >     * Stop testing the truncate64(2) and ftruncate64(2) syscalls.
> >       This simplifies the tests a bit. The kernel implementations are the
> >       same as for truncate(2) and ftruncate(2), so there is little benefit
> >       from testing them exhaustively. (We aren't testing all open(2)
> >       variants either.)
> >   * samples/landlock/sandboxer.c:
> >     * Use switch() to implement best effort mode.
> >   * Documentation:
> >     * Give more background on surprising truncation behaviour.
> >     * Use switch() in the example too, to stay in-line with the sample tool.
> >     * Small fixes in header file to address previous comments.
> > * misc:
> >    * Fix some typos and const usages.
> >
> > V2:
> >   * Documentation: Mention the truncation flag where needed.
> >   * Documentation: Point out connection between truncation and file writing.
> >   * samples: Add file truncation to the landlock/sandboxer.c sample tool.
> >   * selftests: Exercise open(2) with O_TRUNC and creat(2) exhaustively.
> >   * selftests: Exercise truncation syscalls when the truncate right
> >     is not handled by Landlock.
> >
> > Günther Noack (4):
> >    landlock: Support file truncation
> >    selftests/landlock: Selftests for file truncation support
> >    samples/landlock: Extend sample tool to support
> >      LANDLOCK_ACCESS_FS_TRUNCATE
> >    landlock: Document Landlock's file truncation support
> >
> >   Documentation/userspace-api/landlock.rst     |  52 +++-
> >   include/uapi/linux/landlock.h                |  17 +-
> >   samples/landlock/sandboxer.c                 |  23 +-
> >   security/landlock/fs.c                       |   9 +-
> >   security/landlock/limits.h                   |   2 +-
> >   security/landlock/syscalls.c                 |   2 +-
> >   tools/testing/selftests/landlock/base_test.c |   2 +-
> >   tools/testing/selftests/landlock/fs_test.c   | 257 ++++++++++++++++++-
> >   8 files changed, 336 insertions(+), 28 deletions(-)
> >
> >
> > base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> > --
> > 2.37.2

--
Günther Noack Sept. 2, 2022, 6:16 a.m. UTC | #3
On Fri, Sep 02, 2022 at 07:32:49AM +0200, Günther Noack wrote:
> On Thu, Sep 01, 2022 at 07:10:38PM +0200, Mickaël Salaün wrote:
> > Hmm, I think there is an issue with this series. Landlock only enforces
> > restrictions at open time or when dealing with user-supplied file paths
> > (relative or absolute).
>
> Argh, ok. That sounds like a desirable property, although it would
> mean reworking the patch set.
>
> > The use of the path_truncate hook in this series
> > doesn't distinguish between file descriptor from before the current sandbox
> > or from after being sandboxed. For instance, if a file descriptor is
> > received through a unix socket, it is assumed that this is legitimate and no
> > Landlock restriction apply on it, which is not the case with this series
> > anymore. It is the same for files opened before the process sandbox itself.
> >
> > To be able to follow the current semantic, I think we should control the
> > truncate access at open time (or when dealing with a user-supplied path) but
> > not on any file descriptor as it is currently done.
>
> OK - so let me try to make a constructive proposal. We have previously
> identified a few operations where a truncation happens, and I would
> propose that the following Landlock rights should be needed for these:
>
> * truncate() (operating on char *path): Require LL_ACCESS_FS_TRUNCATE
> * ftruncate() (operating on fd): No Landlock rights required
> * open() for reading with O_TRUNC: Require LL_ACCESS_FS_TRUNCATE
> * open() for writing with O_TRUNC: Require LL_ACCESS_FS_WRITE_FILE

Thinking about it again, another alternative would be to require
TRUNCATE as well when opening a file for writing - it would be
logical, because the resulting FD can be truncated. It would also
require people to provide the truncate right in order to open files
for writing, but this may be the logical thing to do.

Let me know what you think!

—Günther

>
> The rationale goes as follows:
>
> * ftruncate() is already adequately protected by the
>   LL_ACCESS_FS_WRITE_FILE right. ftruncate is only permitted on fds
>   that are open for writing.
> * truncate() is not Landlock-restrictable in Landlock ABI v1,
>   so needs to be covered by the new truncate right.
> * open() for reading with O_TRUNC is also not Landlock-restrictable in
>   Landlock ABI v1, so needs to be covered by the new truncate right.
> * open() for writing with O_TRUNC is also not Landlock-restrictable in
>   Landlock ABI v1. BUT: A caller who can open the file for writing
>   will also be able to ftruncate it - so it doesn't really make sense
>   to ask for a different Landlock right here.
>
> Does that approach make sense to you?
>
> I think in terms of changs required for it, it sounds like it would
> require a change to the path_truncate LSM hook to distinguish the
> cases above.
>
> Do you want a new patch on top of the existing one, or should I rather
> create a new version of the old truncate patch set?
>
> --Günther
>
> > On 17/08/2022 22:30, Günther Noack wrote:
> > > The goal of these patches is to work towards a more complete coverage
> > > of file system operations that are restrictable with Landlock.
> > >
> > > The known set of currently unsupported file system operations in
> > > Landlock is described at [1]. Out of the operations listed there,
> > > truncate is the only one that modifies file contents, so these patches
> > > should make it possible to prevent the direct modification of file
> > > contents with Landlock.
> > >
> > > The patch introduces the truncation restriction feature as an
> > > additional bit in the access_mask_t bitmap, in line with the existing
> > > supported operations.
> > >
> > > The truncation flag covers both the truncate(2) and ftruncate(2)
> > > families of syscalls, as well as open(2) with the O_TRUNC flag.
> > > This includes usages of creat() in the case where existing regular
> > > files are overwritten.
> > >
> > > Apart from Landlock, file truncation can also be restricted using
> > > seccomp-bpf, but it is more difficult to use (requires BPF, requires
> > > keeping up-to-date syscall lists) and it is not configurable by file
> > > hierarchy, as Landlock is. The simplicity and flexibility of the
> > > Landlock approach makes it worthwhile adding.
> > >
> > > While it's possible to use the "write file" and "truncate" rights
> > > independent of each other, it simplifies the mental model for
> > > userspace callers to always use them together.
> > >
> > > Specifically, the following behaviours might be surprising for users
> > > when using these independently:
> > >
> > >   * The commonly creat() syscall requires the truncate right when
> > >     overwriting existing files, as it is equivalent to open(2) with
> > >     O_TRUNC|O_CREAT|O_WRONLY.
> > >   * The "write file" right is not always required to truncate a file,
> > >     even through the open(2) syscall (when using O_RDONLY|O_TRUNC).
> > >
> > > Nevertheless, keeping the two flags separate is the correct approach
> > > to guarantee backwards compatibility for existing Landlock users.
> > >
> > > These patches are based on version 6.0-rc1.
> > >
> > > Best regards,
> > > Günther
> > >
> > > [1] https://docs.kernel.org/userspace-api/landlock.html#filesystem-flags
> > >
> > > Past discussions:
> > > V1: https://lore.kernel.org/all/20220707200612.132705-1-gnoack3000@gmail.com/
> > > V2: https://lore.kernel.org/all/20220712211405.14705-1-gnoack3000@gmail.com/
> > > V3: https://lore.kernel.org/all/20220804193746.9161-1-gnoack3000@gmail.com/
> > > V4: https://lore.kernel.org/all/20220814192603.7387-1-gnoack3000@gmail.com/
> > >
> > > Changelog:
> > >
> > > V5:
> > > * Documentation
> > >    * Fix wording in userspace-api headers and in landlock.rst.
> > >    * Move the truncation limitation section one to the bottom.
> > >    * Move all .rst changes into the documentation commit.
> > > * selftests
> > >    * Remove _metadata argument from helpers where it became unnecessary.
> > >    * Open writable file descriptors at the top of both tests, before Landlock
> > >      is enabled, to exercise ftruncate() independently from open().
> > >    * Simplify test_ftruncate and decouple it from exercising open().
> > >    * test_creat(): Return errno on close() failure (it does not conflict).
> > >    * Fix /* comment style */
> > >    * Reorder blocks of EXPECT_EQ checks to be consistent within a test.
> > >    * Add missing |O_TRUNC to a check in one test.
> > >    * Put the truncate_unhandled test before the other.
> > >
> > > V4:
> > >   * Documentation
> > >     * Clarify wording and syntax as discussed in review.
> > >     * Use a less confusing error message in the example.
> > >   * selftests:
> > >     * Stop using ASSERT_EQ in test helpers, return EBADFD instead.
> > >       (This is an intentionally uncommon error code, so that the source
> > >       of the error is clear and the test can distinguish test setup
> > >       failures from failures in the actual system call under test.)
> > >   * samples/Documentation:
> > >     * Use additional clarifying comments in the kernel backwards
> > >       compatibility logic.
> > >
> > > V3:
> > >   * selftests:
> > >     * Explicitly test ftruncate with readonly file descriptors
> > >       (returns EINVAL).
> > >     * Extract test_ftruncate, test_truncate, test_creat helpers,
> > >       which simplified the previously mixed usage of EXPECT/ASSERT.
> > >     * Test creat() behaviour as part of the big truncation test.
> > >     * Stop testing the truncate64(2) and ftruncate64(2) syscalls.
> > >       This simplifies the tests a bit. The kernel implementations are the
> > >       same as for truncate(2) and ftruncate(2), so there is little benefit
> > >       from testing them exhaustively. (We aren't testing all open(2)
> > >       variants either.)
> > >   * samples/landlock/sandboxer.c:
> > >     * Use switch() to implement best effort mode.
> > >   * Documentation:
> > >     * Give more background on surprising truncation behaviour.
> > >     * Use switch() in the example too, to stay in-line with the sample tool.
> > >     * Small fixes in header file to address previous comments.
> > > * misc:
> > >    * Fix some typos and const usages.
> > >
> > > V2:
> > >   * Documentation: Mention the truncation flag where needed.
> > >   * Documentation: Point out connection between truncation and file writing.
> > >   * samples: Add file truncation to the landlock/sandboxer.c sample tool.
> > >   * selftests: Exercise open(2) with O_TRUNC and creat(2) exhaustively.
> > >   * selftests: Exercise truncation syscalls when the truncate right
> > >     is not handled by Landlock.
> > >
> > > Günther Noack (4):
> > >    landlock: Support file truncation
> > >    selftests/landlock: Selftests for file truncation support
> > >    samples/landlock: Extend sample tool to support
> > >      LANDLOCK_ACCESS_FS_TRUNCATE
> > >    landlock: Document Landlock's file truncation support
> > >
> > >   Documentation/userspace-api/landlock.rst     |  52 +++-
> > >   include/uapi/linux/landlock.h                |  17 +-
> > >   samples/landlock/sandboxer.c                 |  23 +-
> > >   security/landlock/fs.c                       |   9 +-
> > >   security/landlock/limits.h                   |   2 +-
> > >   security/landlock/syscalls.c                 |   2 +-
> > >   tools/testing/selftests/landlock/base_test.c |   2 +-
> > >   tools/testing/selftests/landlock/fs_test.c   | 257 ++++++++++++++++++-
> > >   8 files changed, 336 insertions(+), 28 deletions(-)
> > >
> > >
> > > base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
> > > --
> > > 2.37.2
>
> --

--
Mickaël Salaün Sept. 2, 2022, 8:40 a.m. UTC | #4
On 02/09/2022 08:16, Günther Noack wrote:
> On Fri, Sep 02, 2022 at 07:32:49AM +0200, Günther Noack wrote:
>> On Thu, Sep 01, 2022 at 07:10:38PM +0200, Mickaël Salaün wrote:
>>> Hmm, I think there is an issue with this series. Landlock only enforces
>>> restrictions at open time or when dealing with user-supplied file paths
>>> (relative or absolute).
>>
>> Argh, ok. That sounds like a desirable property, although it would
>> mean reworking the patch set.
>>
>>> The use of the path_truncate hook in this series
>>> doesn't distinguish between file descriptor from before the current sandbox
>>> or from after being sandboxed. For instance, if a file descriptor is
>>> received through a unix socket, it is assumed that this is legitimate and no
>>> Landlock restriction apply on it, which is not the case with this series
>>> anymore. It is the same for files opened before the process sandbox itself.
>>>
>>> To be able to follow the current semantic, I think we should control the
>>> truncate access at open time (or when dealing with a user-supplied path) but
>>> not on any file descriptor as it is currently done.
>>
>> OK - so let me try to make a constructive proposal. We have previously
>> identified a few operations where a truncation happens, and I would
>> propose that the following Landlock rights should be needed for these:
>>
>> * truncate() (operating on char *path): Require LL_ACCESS_FS_TRUNCATE
>> * ftruncate() (operating on fd): No Landlock rights required
>> * open() for reading with O_TRUNC: Require LL_ACCESS_FS_TRUNCATE
>> * open() for writing with O_TRUNC: Require LL_ACCESS_FS_WRITE_FILE
> 
> Thinking about it again, another alternative would be to require
> TRUNCATE as well when opening a file for writing - it would be
> logical, because the resulting FD can be truncated. It would also
> require people to provide the truncate right in order to open files
> for writing, but this may be the logical thing to do.

Another alternative would be to keep the current semantic but ignore 
file descriptors from not-sandboxed processes. This could be possible by 
following the current file->f_mode logic but using the Landlock's 
file->f_security instead to store if the file descriptor was opened in a 
context allowing it to be truncated: file opened outside of a landlocked 
process, or in a sandbox allowing LANDLOCK_ACCESS_FS_TRUNCATE on the 
related path.


> 
> Let me know what you think!
> 
> —Günther
> 
>>
>> The rationale goes as follows:
>>
>> * ftruncate() is already adequately protected by the
>>    LL_ACCESS_FS_WRITE_FILE right. ftruncate is only permitted on fds
>>    that are open for writing.
>> * truncate() is not Landlock-restrictable in Landlock ABI v1,
>>    so needs to be covered by the new truncate right.
>> * open() for reading with O_TRUNC is also not Landlock-restrictable in
>>    Landlock ABI v1, so needs to be covered by the new truncate right.
>> * open() for writing with O_TRUNC is also not Landlock-restrictable in
>>    Landlock ABI v1. BUT: A caller who can open the file for writing
>>    will also be able to ftruncate it - so it doesn't really make sense
>>    to ask for a different Landlock right here.
>>
>> Does that approach make sense to you?
>>
>> I think in terms of changs required for it, it sounds like it would
>> require a change to the path_truncate LSM hook to distinguish the
>> cases above.

Yes, it requires some changes to the path_truncate hook. I think 
providing a struct file, when available, as a second argument looks good.

Serge, Paul, what do you think about that?


>>
>> Do you want a new patch on top of the existing one, or should I rather
>> create a new version of the old truncate patch set?

Please create a sixth patch series also including my (slight) changes.


>>
>> --Günther
>>
>>> On 17/08/2022 22:30, Günther Noack wrote:
>>>> The goal of these patches is to work towards a more complete coverage
>>>> of file system operations that are restrictable with Landlock.
>>>>
>>>> The known set of currently unsupported file system operations in
>>>> Landlock is described at [1]. Out of the operations listed there,
>>>> truncate is the only one that modifies file contents, so these patches
>>>> should make it possible to prevent the direct modification of file
>>>> contents with Landlock.
>>>>
>>>> The patch introduces the truncation restriction feature as an
>>>> additional bit in the access_mask_t bitmap, in line with the existing
>>>> supported operations.
>>>>
>>>> The truncation flag covers both the truncate(2) and ftruncate(2)
>>>> families of syscalls, as well as open(2) with the O_TRUNC flag.
>>>> This includes usages of creat() in the case where existing regular
>>>> files are overwritten.
>>>>
>>>> Apart from Landlock, file truncation can also be restricted using
>>>> seccomp-bpf, but it is more difficult to use (requires BPF, requires
>>>> keeping up-to-date syscall lists) and it is not configurable by file
>>>> hierarchy, as Landlock is. The simplicity and flexibility of the
>>>> Landlock approach makes it worthwhile adding.
>>>>
>>>> While it's possible to use the "write file" and "truncate" rights
>>>> independent of each other, it simplifies the mental model for
>>>> userspace callers to always use them together.
>>>>
>>>> Specifically, the following behaviours might be surprising for users
>>>> when using these independently:
>>>>
>>>>    * The commonly creat() syscall requires the truncate right when
>>>>      overwriting existing files, as it is equivalent to open(2) with
>>>>      O_TRUNC|O_CREAT|O_WRONLY.
>>>>    * The "write file" right is not always required to truncate a file,
>>>>      even through the open(2) syscall (when using O_RDONLY|O_TRUNC).
>>>>
>>>> Nevertheless, keeping the two flags separate is the correct approach
>>>> to guarantee backwards compatibility for existing Landlock users.
>>>>
>>>> These patches are based on version 6.0-rc1.
>>>>
>>>> Best regards,
>>>> Günther
>>>>
>>>> [1] https://docs.kernel.org/userspace-api/landlock.html#filesystem-flags
>>>>
>>>> Past discussions:
>>>> V1: https://lore.kernel.org/all/20220707200612.132705-1-gnoack3000@gmail.com/
>>>> V2: https://lore.kernel.org/all/20220712211405.14705-1-gnoack3000@gmail.com/
>>>> V3: https://lore.kernel.org/all/20220804193746.9161-1-gnoack3000@gmail.com/
>>>> V4: https://lore.kernel.org/all/20220814192603.7387-1-gnoack3000@gmail.com/
>>>>
>>>> Changelog:
>>>>
>>>> V5:
>>>> * Documentation
>>>>     * Fix wording in userspace-api headers and in landlock.rst.
>>>>     * Move the truncation limitation section one to the bottom.
>>>>     * Move all .rst changes into the documentation commit.
>>>> * selftests
>>>>     * Remove _metadata argument from helpers where it became unnecessary.
>>>>     * Open writable file descriptors at the top of both tests, before Landlock
>>>>       is enabled, to exercise ftruncate() independently from open().
>>>>     * Simplify test_ftruncate and decouple it from exercising open().
>>>>     * test_creat(): Return errno on close() failure (it does not conflict).
>>>>     * Fix /* comment style */
>>>>     * Reorder blocks of EXPECT_EQ checks to be consistent within a test.
>>>>     * Add missing |O_TRUNC to a check in one test.
>>>>     * Put the truncate_unhandled test before the other.
>>>>
>>>> V4:
>>>>    * Documentation
>>>>      * Clarify wording and syntax as discussed in review.
>>>>      * Use a less confusing error message in the example.
>>>>    * selftests:
>>>>      * Stop using ASSERT_EQ in test helpers, return EBADFD instead.
>>>>        (This is an intentionally uncommon error code, so that the source
>>>>        of the error is clear and the test can distinguish test setup
>>>>        failures from failures in the actual system call under test.)
>>>>    * samples/Documentation:
>>>>      * Use additional clarifying comments in the kernel backwards
>>>>        compatibility logic.
>>>>
>>>> V3:
>>>>    * selftests:
>>>>      * Explicitly test ftruncate with readonly file descriptors
>>>>        (returns EINVAL).
>>>>      * Extract test_ftruncate, test_truncate, test_creat helpers,
>>>>        which simplified the previously mixed usage of EXPECT/ASSERT.
>>>>      * Test creat() behaviour as part of the big truncation test.
>>>>      * Stop testing the truncate64(2) and ftruncate64(2) syscalls.
>>>>        This simplifies the tests a bit. The kernel implementations are the
>>>>        same as for truncate(2) and ftruncate(2), so there is little benefit
>>>>        from testing them exhaustively. (We aren't testing all open(2)
>>>>        variants either.)
>>>>    * samples/landlock/sandboxer.c:
>>>>      * Use switch() to implement best effort mode.
>>>>    * Documentation:
>>>>      * Give more background on surprising truncation behaviour.
>>>>      * Use switch() in the example too, to stay in-line with the sample tool.
>>>>      * Small fixes in header file to address previous comments.
>>>> * misc:
>>>>     * Fix some typos and const usages.
>>>>
>>>> V2:
>>>>    * Documentation: Mention the truncation flag where needed.
>>>>    * Documentation: Point out connection between truncation and file writing.
>>>>    * samples: Add file truncation to the landlock/sandboxer.c sample tool.
>>>>    * selftests: Exercise open(2) with O_TRUNC and creat(2) exhaustively.
>>>>    * selftests: Exercise truncation syscalls when the truncate right
>>>>      is not handled by Landlock.
>>>>
>>>> Günther Noack (4):
>>>>     landlock: Support file truncation
>>>>     selftests/landlock: Selftests for file truncation support
>>>>     samples/landlock: Extend sample tool to support
>>>>       LANDLOCK_ACCESS_FS_TRUNCATE
>>>>     landlock: Document Landlock's file truncation support
>>>>
>>>>    Documentation/userspace-api/landlock.rst     |  52 +++-
>>>>    include/uapi/linux/landlock.h                |  17 +-
>>>>    samples/landlock/sandboxer.c                 |  23 +-
>>>>    security/landlock/fs.c                       |   9 +-
>>>>    security/landlock/limits.h                   |   2 +-
>>>>    security/landlock/syscalls.c                 |   2 +-
>>>>    tools/testing/selftests/landlock/base_test.c |   2 +-
>>>>    tools/testing/selftests/landlock/fs_test.c   | 257 ++++++++++++++++++-
>>>>    8 files changed, 336 insertions(+), 28 deletions(-)
>>>>
>>>>
>>>> base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
>>>> --
>>>> 2.37.2
>>
>> --
> 
> --
xiujianfeng Sept. 2, 2022, 12:26 p.m. UTC | #5
Hi,

在 2022/9/2 1:10, Mickaël Salaün 写道:
> Hmm, I think there is an issue with this series. Landlock only enforces 
> restrictions at open time or when dealing with user-supplied file paths 
> (relative or absolute). The use of the path_truncate hook in this series 
> doesn't distinguish between file descriptor from before the current 
> sandbox or from after being sandboxed. For instance, if a file 
> descriptor is received through a unix socket, it is assumed that this is 
> legitimate and no Landlock restriction apply on it, which is not the 
> case with this series anymore. It is the same for files opened before 
> the process sandbox itself.

so I think this issue also exists in the chown/chmod series, right? 
there is a testcase in that patchset verify the corresponding rights 
inside the sanbox with a fd opened before sanboxing.
> 
> To be able to follow the current semantic, I think we should control the 
> truncate access at open time (or when dealing with a user-supplied path) 
> but not on any file descriptor as it is currently done. >
> 
> On 17/08/2022 22:30, Günther Noack wrote:
>> The goal of these patches is to work towards a more complete coverage
>> of file system operations that are restrictable with Landlock.
>>
>> The known set of currently unsupported file system operations in
>> Landlock is described at [1]. Out of the operations listed there,
>> truncate is the only one that modifies file contents, so these patches
>> should make it possible to prevent the direct modification of file
>> contents with Landlock.
>>
>> The patch introduces the truncation restriction feature as an
>> additional bit in the access_mask_t bitmap, in line with the existing
>> supported operations.
>>
>> The truncation flag covers both the truncate(2) and ftruncate(2)
>> families of syscalls, as well as open(2) with the O_TRUNC flag.
>> This includes usages of creat() in the case where existing regular
>> files are overwritten.
>>
>> Apart from Landlock, file truncation can also be restricted using
>> seccomp-bpf, but it is more difficult to use (requires BPF, requires
>> keeping up-to-date syscall lists) and it is not configurable by file
>> hierarchy, as Landlock is. The simplicity and flexibility of the
>> Landlock approach makes it worthwhile adding.
>>
>> While it's possible to use the "write file" and "truncate" rights
>> independent of each other, it simplifies the mental model for
>> userspace callers to always use them together.
>>
>> Specifically, the following behaviours might be surprising for users
>> when using these independently:
>>
>>   * The commonly creat() syscall requires the truncate right when
>>     overwriting existing files, as it is equivalent to open(2) with
>>     O_TRUNC|O_CREAT|O_WRONLY.
>>   * The "write file" right is not always required to truncate a file,
>>     even through the open(2) syscall (when using O_RDONLY|O_TRUNC).
>>
>> Nevertheless, keeping the two flags separate is the correct approach
>> to guarantee backwards compatibility for existing Landlock users.
>>
>> These patches are based on version 6.0-rc1.
>>
>> Best regards,
>> Günther
>>
>> [1] https://docs.kernel.org/userspace-api/landlock.html#filesystem-flags
>>
>> Past discussions:
>> V1: 
>> https://lore.kernel.org/all/20220707200612.132705-1-gnoack3000@gmail.com/
>> V2: 
>> https://lore.kernel.org/all/20220712211405.14705-1-gnoack3000@gmail.com/
>> V3: 
>> https://lore.kernel.org/all/20220804193746.9161-1-gnoack3000@gmail.com/
>> V4: 
>> https://lore.kernel.org/all/20220814192603.7387-1-gnoack3000@gmail.com/
>>
>> Changelog:
>>
>> V5:
>> * Documentation
>>    * Fix wording in userspace-api headers and in landlock.rst.
>>    * Move the truncation limitation section one to the bottom.
>>    * Move all .rst changes into the documentation commit.
>> * selftests
>>    * Remove _metadata argument from helpers where it became unnecessary.
>>    * Open writable file descriptors at the top of both tests, before 
>> Landlock
>>      is enabled, to exercise ftruncate() independently from open().
>>    * Simplify test_ftruncate and decouple it from exercising open().
>>    * test_creat(): Return errno on close() failure (it does not 
>> conflict).
>>    * Fix /* comment style */
>>    * Reorder blocks of EXPECT_EQ checks to be consistent within a test.
>>    * Add missing |O_TRUNC to a check in one test.
>>    * Put the truncate_unhandled test before the other.
>>
>> V4:
>>   * Documentation
>>     * Clarify wording and syntax as discussed in review.
>>     * Use a less confusing error message in the example.
>>   * selftests:
>>     * Stop using ASSERT_EQ in test helpers, return EBADFD instead.
>>       (This is an intentionally uncommon error code, so that the source
>>       of the error is clear and the test can distinguish test setup
>>       failures from failures in the actual system call under test.)
>>   * samples/Documentation:
>>     * Use additional clarifying comments in the kernel backwards
>>       compatibility logic.
>>
>> V3:
>>   * selftests:
>>     * Explicitly test ftruncate with readonly file descriptors
>>       (returns EINVAL).
>>     * Extract test_ftruncate, test_truncate, test_creat helpers,
>>       which simplified the previously mixed usage of EXPECT/ASSERT.
>>     * Test creat() behaviour as part of the big truncation test.
>>     * Stop testing the truncate64(2) and ftruncate64(2) syscalls.
>>       This simplifies the tests a bit. The kernel implementations are the
>>       same as for truncate(2) and ftruncate(2), so there is little 
>> benefit
>>       from testing them exhaustively. (We aren't testing all open(2)
>>       variants either.)
>>   * samples/landlock/sandboxer.c:
>>     * Use switch() to implement best effort mode.
>>   * Documentation:
>>     * Give more background on surprising truncation behaviour.
>>     * Use switch() in the example too, to stay in-line with the sample 
>> tool.
>>     * Small fixes in header file to address previous comments.
>> * misc:
>>    * Fix some typos and const usages.
>>
>> V2:
>>   * Documentation: Mention the truncation flag where needed.
>>   * Documentation: Point out connection between truncation and file 
>> writing.
>>   * samples: Add file truncation to the landlock/sandboxer.c sample tool.
>>   * selftests: Exercise open(2) with O_TRUNC and creat(2) exhaustively.
>>   * selftests: Exercise truncation syscalls when the truncate right
>>     is not handled by Landlock.
>>
>> Günther Noack (4):
>>    landlock: Support file truncation
>>    selftests/landlock: Selftests for file truncation support
>>    samples/landlock: Extend sample tool to support
>>      LANDLOCK_ACCESS_FS_TRUNCATE
>>    landlock: Document Landlock's file truncation support
>>
>>   Documentation/userspace-api/landlock.rst     |  52 +++-
>>   include/uapi/linux/landlock.h                |  17 +-
>>   samples/landlock/sandboxer.c                 |  23 +-
>>   security/landlock/fs.c                       |   9 +-
>>   security/landlock/limits.h                   |   2 +-
>>   security/landlock/syscalls.c                 |   2 +-
>>   tools/testing/selftests/landlock/base_test.c |   2 +-
>>   tools/testing/selftests/landlock/fs_test.c   | 257 ++++++++++++++++++-
>>   8 files changed, 336 insertions(+), 28 deletions(-)
>>
>>
>> base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868
>> -- 
>> 2.37.2
> .
Mickaël Salaün Sept. 2, 2022, 1:12 p.m. UTC | #6
On 02/09/2022 14:26, xiujianfeng wrote:
> Hi,
> 
> 在 2022/9/2 1:10, Mickaël Salaün 写道:
>> Hmm, I think there is an issue with this series. Landlock only enforces
>> restrictions at open time or when dealing with user-supplied file paths
>> (relative or absolute). The use of the path_truncate hook in this series
>> doesn't distinguish between file descriptor from before the current
>> sandbox or from after being sandboxed. For instance, if a file
>> descriptor is received through a unix socket, it is assumed that this is
>> legitimate and no Landlock restriction apply on it, which is not the
>> case with this series anymore. It is the same for files opened before
>> the process sandbox itself.
> 
> so I think this issue also exists in the chown/chmod series, right?
> there is a testcase in that patchset verify the corresponding rights
> inside the sanbox with a fd opened before sanboxing.

Correct. For LANDLOCK_ACCESS_FS_TRUNCATE, we need to add tests to make 
sure that:
* a sandboxed process with the truncate restriction can open a file in 
write mode, forward it to an un-sandboxed process, and make sure this 
receiver cannot truncate the file descriptor, nor its dup.
* an inherited file descriptor can be truncated even if done by a 
sandboxed process, except if it was created by a sandboxed process and 
the truncate restriction applied on it.

However, for the file metadata accesses, I suggest you first focus on 
the inode_setattr and inode_setxattr hook modifications. We'll get back 
to this FD-based restrictions later.
Günther Noack Sept. 2, 2022, 4:34 p.m. UTC | #7
On Fri, Sep 02, 2022 at 10:40:57AM +0200, Mickaël Salaün wrote:
> On 02/09/2022 08:16, Günther Noack wrote:
> > On Fri, Sep 02, 2022 at 07:32:49AM +0200, Günther Noack wrote:
> > > On Thu, Sep 01, 2022 at 07:10:38PM +0200, Mickaël Salaün wrote:
> > > > Hmm, I think there is an issue with this series. Landlock only enforces
> > > > restrictions at open time or when dealing with user-supplied file paths
> > > > (relative or absolute).
> > >
> > > Argh, ok. That sounds like a desirable property, although it would
> > > mean reworking the patch set.
> > >
> > > > The use of the path_truncate hook in this series
> > > > doesn't distinguish between file descriptor from before the current sandbox
> > > > or from after being sandboxed. For instance, if a file descriptor is
> > > > received through a unix socket, it is assumed that this is legitimate and no
> > > > Landlock restriction apply on it, which is not the case with this series
> > > > anymore. It is the same for files opened before the process sandbox itself.
> > > >
> > > > To be able to follow the current semantic, I think we should control the
> > > > truncate access at open time (or when dealing with a user-supplied path) but
> > > > not on any file descriptor as it is currently done.
> > >
> > > OK - so let me try to make a constructive proposal. We have previously
> > > identified a few operations where a truncation happens, and I would
> > > propose that the following Landlock rights should be needed for these:
> > >
> > > * truncate() (operating on char *path): Require LL_ACCESS_FS_TRUNCATE
> > > * ftruncate() (operating on fd): No Landlock rights required
> > > * open() for reading with O_TRUNC: Require LL_ACCESS_FS_TRUNCATE
> > > * open() for writing with O_TRUNC: Require LL_ACCESS_FS_WRITE_FILE
> >
> > Thinking about it again, another alternative would be to require
> > TRUNCATE as well when opening a file for writing - it would be
> > logical, because the resulting FD can be truncated. It would also
> > require people to provide the truncate right in order to open files
> > for writing, but this may be the logical thing to do.
>
> Another alternative would be to keep the current semantic but ignore file
> descriptors from not-sandboxed processes. This could be possible by
> following the current file->f_mode logic but using the Landlock's
> file->f_security instead to store if the file descriptor was opened in a
> context allowing it to be truncated: file opened outside of a landlocked
> process, or in a sandbox allowing LANDLOCK_ACCESS_FS_TRUNCATE on the related
> path.

I'm not convinced that it'll be worth distinguishing between a FD
opened for writing and a FD opened for writing+truncation. And whether
the FD is open for writing is already tracked by default and
ftruncate() checks that.

I'm having a hard time constructing a scenario where write() should be
permitted on an FD but ftruncate() should be forbidden. It seems that
write() is the more dangerous operation of the two, with more
potential to modify a file to one's liking, whereas the modifications
possible through TRUNCATE are relatively benign?

The opposite scenario (where ftruncate() is permitted and write() is
forbidden) simply can't exist because an FD must already be writable
in order to use ftruncate(). (see man page)

Additionally, if we recall previous discussions on the truncate patch
sets, there is the very commonly used creat() syscall (a.k.a. open()
with O_CREAT|O_WRONLY|O_TRUNC), which anyway requires the Landlock
truncate right in many cases. So I still think you can't actually use
LANDLOCK_ACCESS_FS_FILE_WRITE without also providing the
LANDLOCK_ACCESS_FS_TRUNCATE right?

In conclusion, I'd be in favor of not tracking the truncate right
separately as a property of an open file descriptor. Does that
rationale sound reasonable?

--
Mickaël Salaün Sept. 3, 2022, 12:02 p.m. UTC | #8
On 02/09/2022 18:34, Günther Noack wrote:
> On Fri, Sep 02, 2022 at 10:40:57AM +0200, Mickaël Salaün wrote:
>> On 02/09/2022 08:16, Günther Noack wrote:
>>> On Fri, Sep 02, 2022 at 07:32:49AM +0200, Günther Noack wrote:
>>>> On Thu, Sep 01, 2022 at 07:10:38PM +0200, Mickaël Salaün wrote:
>>>>> Hmm, I think there is an issue with this series. Landlock only enforces
>>>>> restrictions at open time or when dealing with user-supplied file paths
>>>>> (relative or absolute).
>>>>
>>>> Argh, ok. That sounds like a desirable property, although it would
>>>> mean reworking the patch set.
>>>>
>>>>> The use of the path_truncate hook in this series
>>>>> doesn't distinguish between file descriptor from before the current sandbox
>>>>> or from after being sandboxed. For instance, if a file descriptor is
>>>>> received through a unix socket, it is assumed that this is legitimate and no
>>>>> Landlock restriction apply on it, which is not the case with this series
>>>>> anymore. It is the same for files opened before the process sandbox itself.
>>>>>
>>>>> To be able to follow the current semantic, I think we should control the
>>>>> truncate access at open time (or when dealing with a user-supplied path) but
>>>>> not on any file descriptor as it is currently done.
>>>>
>>>> OK - so let me try to make a constructive proposal. We have previously
>>>> identified a few operations where a truncation happens, and I would
>>>> propose that the following Landlock rights should be needed for these:
>>>>
>>>> * truncate() (operating on char *path): Require LL_ACCESS_FS_TRUNCATE
>>>> * ftruncate() (operating on fd): No Landlock rights required
>>>> * open() for reading with O_TRUNC: Require LL_ACCESS_FS_TRUNCATE
>>>> * open() for writing with O_TRUNC: Require LL_ACCESS_FS_WRITE_FILE
>>>
>>> Thinking about it again, another alternative would be to require
>>> TRUNCATE as well when opening a file for writing - it would be
>>> logical, because the resulting FD can be truncated. It would also
>>> require people to provide the truncate right in order to open files
>>> for writing, but this may be the logical thing to do.
>>
>> Another alternative would be to keep the current semantic but ignore file
>> descriptors from not-sandboxed processes. This could be possible by
>> following the current file->f_mode logic but using the Landlock's
>> file->f_security instead to store if the file descriptor was opened in a
>> context allowing it to be truncated: file opened outside of a landlocked
>> process, or in a sandbox allowing LANDLOCK_ACCESS_FS_TRUNCATE on the related
>> path.
> 
> I'm not convinced that it'll be worth distinguishing between a FD
> opened for writing and a FD opened for writing+truncation. And whether
> the FD is open for writing is already tracked by default and
> ftruncate() checks that.

That might be a misunderstanding. What I'm proposing is to keep the same 
semantic as this fifth patch series, only to keep scoped Landlock 
restrictions and propagate them (which is already how Landlock works).

The layout1.truncate tests should work the same except that 
test_ftruncate(file_*_fd) will always be allowed because such FD they 
are opened before the thread being sandboxed.

> 
> I'm having a hard time constructing a scenario where write() should be
> permitted on an FD but ftruncate() should be forbidden. It seems that
> write() is the more dangerous operation of the two, with more
> potential to modify a file to one's liking, whereas the modifications
> possible through TRUNCATE are relatively benign?

I don't understand, this is how this fifth series already restrict 
truncate. I'm not proposing to change the FD minimal requirement to be 
"truncatable", and it would not be possible with the LSM framework anyway.


> 
> The opposite scenario (where ftruncate() is permitted and write() is
> forbidden) simply can't exist because an FD must already be writable
> in order to use ftruncate(). (see man page)

Right, and I'm not proposing to change that. Currently, the kernel 
tracks how a FD was opened (e.g. read or write mode). I'm proposing to 
add another *complementary* Landlock-specific mode for truncate because 
it is a more fine-grained access right with Landlock, hence this patch 
series.


> 
> Additionally, if we recall previous discussions on the truncate patch
> sets, there is the very commonly used creat() syscall (a.k.a. open()
> with O_CREAT|O_WRONLY|O_TRUNC), which anyway requires the Landlock
> truncate right in many cases. So I still think you can't actually use
> LANDLOCK_ACCESS_FS_FILE_WRITE without also providing the
> LANDLOCK_ACCESS_FS_TRUNCATE right?

Hmm, that is definitely not a common case, but this series permit that, 
see test_truncate(file_t).

> 
> In conclusion, I'd be in favor of not tracking the truncate right
> separately as a property of an open file descriptor. Does that
> rationale sound reasonable?

No, but I think there is a misunderstanding. :)

The idea is first to change hook_file_open() to set 
landlock_file(file)->access = LANDLOCK_ACCESS_FS_TRUNCATE if it is 
allowed by the policy: current thread being either not in a sandbox, or 
in a sandbox that allows truncate.
Then, in hook_path_truncate(), we allow the operation if `file && 
(landlock_file(file)->access & LANDLOCK_ACCESS_FS_TRUNCATE)`. Otherwise, 
it there is only a path available (because it comes from truncate(2)), 
we (almost) call current_check_access_path(path, 
LANDLOCK_ACCESS_FS_TRUNCATE).