[PATCH 01/21] binder: use EPOLLERR from eventpoll.h

[PATCH 01/21] binder: use EPOLLERR from eventpoll.h

[Carlos Llamas]

Use EPOLLERR instead of POLLERR to make sure it is cast to the correct
__poll_t type. This fixes the following sparse issue:

  drivers/android/binder.c:5030:24: warning: incorrect type in return expression (different base types)
  drivers/android/binder.c:5030:24:    expected restricted __poll_t
  drivers/android/binder.c:5030:24:    got int

Fixes: f88982679f54 ("binder: check for binder_thread allocation failure in binder_poll()")
Cc: stable@vger.kernel.org
Cc: Eric Biggers <ebiggers@google.com>
Signed-off-by: Carlos Llamas <cmllamas@google.com>
---
 drivers/android/binder.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/android/binder.c b/drivers/android/binder.c
index 92128aae2d06..71a40a4c546f 100644
--- a/drivers/android/binder.c
+++ b/drivers/android/binder.c
@@ -5030,7 +5030,7 @@ static __poll_t binder_poll(struct file *filp,
 
 	thread = binder_get_thread(proc);
 	if (!thread)
-		return POLLERR;
+		return EPOLLERR;
 
 	binder_inner_proc_lock(thread->proc);
 	thread->looper |= BINDER_LOOPER_STATE_POLL;
-- 
2.42.0.869.gea05f2083d-goog

[PATCH 02/21] binder: fix use-after-free in shinker's callback

[Carlos Llamas]

The mmap read lock is used during the shrinker's callback, which means
that using alloc->vma pointer isn't safe as it can race with munmap().
As of commit dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in
munmap") the mmap lock is downgraded after the vma has been isolated.

I was able to reproduce this issue by manually adding some delays and
triggering page reclaiming through the shrinker's debug sysfs. The
following KASAN report confirms the UAF:

  ==================================================================
  BUG: KASAN: slab-use-after-free in zap_page_range_single+0x470/0x4b8
  Read of size 8 at addr ffff356ed50e50f0 by task bash/478

  CPU: 1 PID: 478 Comm: bash Not tainted 6.6.0-rc5-00055-g1c8b86a3799f-dirty #70
  Hardware name: linux,dummy-virt (DT)
  Call trace:
   zap_page_range_single+0x470/0x4b8
   binder_alloc_free_page+0x608/0xadc
   __list_lru_walk_one+0x130/0x3b0
   list_lru_walk_node+0xc4/0x22c
   binder_shrink_scan+0x108/0x1dc
   shrinker_debugfs_scan_write+0x2b4/0x500
   full_proxy_write+0xd4/0x140
   vfs_write+0x1ac/0x758
   ksys_write+0xf0/0x1dc
   __arm64_sys_write+0x6c/0x9c

  Allocated by task 492:
   kmem_cache_alloc+0x130/0x368
   vm_area_alloc+0x2c/0x190
   mmap_region+0x258/0x18bc
   do_mmap+0x694/0xa60
   vm_mmap_pgoff+0x170/0x29c
   ksys_mmap_pgoff+0x290/0x3a0
   __arm64_sys_mmap+0xcc/0x144

  Freed by task 491:
   kmem_cache_free+0x17c/0x3c8
   vm_area_free_rcu_cb+0x74/0x98
   rcu_core+0xa38/0x26d4
   rcu_core_si+0x10/0x1c
   __do_softirq+0x2fc/0xd24

  Last potentially related work creation:
   __call_rcu_common.constprop.0+0x6c/0xba0
   call_rcu+0x10/0x1c
   vm_area_free+0x18/0x24
   remove_vma+0xe4/0x118
   do_vmi_align_munmap.isra.0+0x718/0xb5c
   do_vmi_munmap+0xdc/0x1fc
   __vm_munmap+0x10c/0x278
   __arm64_sys_munmap+0x58/0x7c

Fix this issue by performing instead a vma_lookup() which will fail to
find the vma that was isolated before the mmap lock downgrade. Note that
this option has better performance than upgrading to a mmap write lock
which would increase contention. Plus, mmap_write_trylock() has been
recently removed anyway.

Fixes: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap")
Cc: stable@vger.kernel.org
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Carlos Llamas <cmllamas@google.com>
---
 drivers/android/binder_alloc.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
index e3db8297095a..c4d60d81221b 100644
--- a/drivers/android/binder_alloc.c
+++ b/drivers/android/binder_alloc.c
@@ -1005,7 +1005,9 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
 		goto err_mmget;
 	if (!mmap_read_trylock(mm))
 		goto err_mmap_read_lock_failed;
-	vma = binder_alloc_get_vma(alloc);
+	vma = vma_lookup(mm, page_addr);
+	if (vma && vma != binder_alloc_get_vma(alloc))
+		goto err_invalid_vma;
 
 	list_lru_isolate(lru, item);
 	spin_unlock(lock);
@@ -1031,6 +1033,8 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
 	mutex_unlock(&alloc->mutex);
 	return LRU_REMOVED_RETRY;
 
+err_invalid_vma:
+	mmap_read_unlock(mm);
 err_mmap_read_lock_failed:
 	mmput_async(mm);
 err_mmget:
-- 
2.42.0.869.gea05f2083d-goog

[PATCH 05/21] binder: fix trivial typo of binder_free_buf_locked()

[Carlos Llamas]

Fix minor misspelling of the function in the comment section.

No functional changes in this patch.

Cc: stable@vger.kernel.org
Fixes: 0f966cba95c7 ("binder: add flag to clear buffer on txn complete")
Signed-off-by: Carlos Llamas <cmllamas@google.com>
---
 drivers/android/binder_alloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
index 9b28d0f9666d..cd720bb5c9ce 100644
--- a/drivers/android/binder_alloc.c
+++ b/drivers/android/binder_alloc.c
@@ -707,7 +707,7 @@ void binder_alloc_free_buf(struct binder_alloc *alloc,
 	/*
 	 * We could eliminate the call to binder_alloc_clear_buf()
 	 * from binder_alloc_deferred_release() by moving this to
-	 * binder_alloc_free_buf_locked(). However, that could
+	 * binder_free_buf_locked(). However, that could
 	 * increase contention for the alloc mutex if clear_on_free
 	 * is used frequently for large buffers. The mutex is not
 	 * needed for correctness here.
-- 
2.42.0.869.gea05f2083d-goog

[PATCH 06/21] binder: fix comment on binder_alloc_new_buf() return value

[Carlos Llamas]

Update the comments of binder_alloc_new_buf() to reflect that the return
value of the function is now ERR_PTR(-errno) on failure.

No functional changes in this patch.

Cc: stable@vger.kernel.org
Fixes: 57ada2fb2250 ("binder: add log information for binder transaction failures")
Signed-off-by: Carlos Llamas <cmllamas@google.com>
---
 drivers/android/binder_alloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
index cd720bb5c9ce..0e8312f4b771 100644
--- a/drivers/android/binder_alloc.c
+++ b/drivers/android/binder_alloc.c
@@ -558,7 +558,7 @@ static struct binder_buffer *binder_alloc_new_buf_locked(
  * is the sum of the three given sizes (each rounded up to
  * pointer-sized boundary)
  *
- * Return:	The allocated buffer or %NULL if error
+ * Return:	The allocated buffer or %ERR_PTR(-errno) if error
  */
 struct binder_buffer *binder_alloc_new_buf(struct binder_alloc *alloc,
 					   size_t data_size,
-- 
2.42.0.869.gea05f2083d-goog

Re: [PATCH 02/21] binder: fix use-after-free in shinker's callback

[Liam R. Howlett]

* Carlos Llamas <cmllamas@google.com> [231102 15:00]:
> The mmap read lock is used during the shrinker's callback, which means
> that using alloc->vma pointer isn't safe as it can race with munmap().

I think you know my feelings about the safety of that pointer from
previous discussions.

> As of commit dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in
> munmap") the mmap lock is downgraded after the vma has been isolated.
> 
> I was able to reproduce this issue by manually adding some delays and
> triggering page reclaiming through the shrinker's debug sysfs. The
> following KASAN report confirms the UAF:
> 
>   ==================================================================
>   BUG: KASAN: slab-use-after-free in zap_page_range_single+0x470/0x4b8
>   Read of size 8 at addr ffff356ed50e50f0 by task bash/478
> 
>   CPU: 1 PID: 478 Comm: bash Not tainted 6.6.0-rc5-00055-g1c8b86a3799f-dirty #70
>   Hardware name: linux,dummy-virt (DT)
>   Call trace:
>    zap_page_range_single+0x470/0x4b8
>    binder_alloc_free_page+0x608/0xadc
>    __list_lru_walk_one+0x130/0x3b0
>    list_lru_walk_node+0xc4/0x22c
>    binder_shrink_scan+0x108/0x1dc
>    shrinker_debugfs_scan_write+0x2b4/0x500
>    full_proxy_write+0xd4/0x140
>    vfs_write+0x1ac/0x758
>    ksys_write+0xf0/0x1dc
>    __arm64_sys_write+0x6c/0x9c
> 
>   Allocated by task 492:
>    kmem_cache_alloc+0x130/0x368
>    vm_area_alloc+0x2c/0x190
>    mmap_region+0x258/0x18bc
>    do_mmap+0x694/0xa60
>    vm_mmap_pgoff+0x170/0x29c
>    ksys_mmap_pgoff+0x290/0x3a0
>    __arm64_sys_mmap+0xcc/0x144
> 
>   Freed by task 491:
>    kmem_cache_free+0x17c/0x3c8
>    vm_area_free_rcu_cb+0x74/0x98
>    rcu_core+0xa38/0x26d4
>    rcu_core_si+0x10/0x1c
>    __do_softirq+0x2fc/0xd24
> 
>   Last potentially related work creation:
>    __call_rcu_common.constprop.0+0x6c/0xba0
>    call_rcu+0x10/0x1c
>    vm_area_free+0x18/0x24
>    remove_vma+0xe4/0x118
>    do_vmi_align_munmap.isra.0+0x718/0xb5c
>    do_vmi_munmap+0xdc/0x1fc
>    __vm_munmap+0x10c/0x278
>    __arm64_sys_munmap+0x58/0x7c
> 
> Fix this issue by performing instead a vma_lookup() which will fail to
> find the vma that was isolated before the mmap lock downgrade. Note that
> this option has better performance than upgrading to a mmap write lock
> which would increase contention. Plus, mmap_write_trylock() has been
> recently removed anyway.
> 
> Fixes: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap")
> Cc: stable@vger.kernel.org
> Cc: Liam Howlett <liam.howlett@oracle.com>
> Cc: Minchan Kim <minchan@kernel.org>
> Signed-off-by: Carlos Llamas <cmllamas@google.com>
> ---
>  drivers/android/binder_alloc.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
> index e3db8297095a..c4d60d81221b 100644
> --- a/drivers/android/binder_alloc.c
> +++ b/drivers/android/binder_alloc.c
> @@ -1005,7 +1005,9 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
>  		goto err_mmget;
>  	if (!mmap_read_trylock(mm))
>  		goto err_mmap_read_lock_failed;
> -	vma = binder_alloc_get_vma(alloc);
> +	vma = vma_lookup(mm, page_addr);
> +	if (vma && vma != binder_alloc_get_vma(alloc))
> +		goto err_invalid_vma;

Doesn't this need to be:
if (!vma || vma != binder_alloc_get_vma(alloc))

This way, we catch a different vma and a NULL vma.

Or even, just:
if (vma != binder_alloc_get_vma(alloc))

if the alloc vma cannot be NULL?

>  
>  	list_lru_isolate(lru, item);
>  	spin_unlock(lock);
> @@ -1031,6 +1033,8 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
>  	mutex_unlock(&alloc->mutex);
>  	return LRU_REMOVED_RETRY;
>  
> +err_invalid_vma:
> +	mmap_read_unlock(mm);
>  err_mmap_read_lock_failed:
>  	mmput_async(mm);
>  err_mmget:
> -- 
> 2.42.0.869.gea05f2083d-goog
> 

Re: [PATCH 02/21] binder: fix use-after-free in shinker's callback

[Carlos Llamas]

On Thu, Nov 02, 2023 at 03:20:51PM -0400, Liam R. Howlett wrote:
> * Carlos Llamas <cmllamas@google.com> [231102 15:00]:
> > The mmap read lock is used during the shrinker's callback, which means
> > that using alloc->vma pointer isn't safe as it can race with munmap().
> 
> I think you know my feelings about the safety of that pointer from
> previous discussions.
> 

Yeah. The work here is not done. We actually already store the vm_start
address in alloc->buffer, so in theory we don't even need to swap the
alloc->vma pointer we could just drop it. So, I agree with you.

I want to include this saftey "fix" along with some other work that uses
the page fault handler and get_user_pages_remote(). I've tried a quick
prototype of this and it works fine.

> > diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
> > index e3db8297095a..c4d60d81221b 100644
> > --- a/drivers/android/binder_alloc.c
> > +++ b/drivers/android/binder_alloc.c
> > @@ -1005,7 +1005,9 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
> >  		goto err_mmget;
> >  	if (!mmap_read_trylock(mm))
> >  		goto err_mmap_read_lock_failed;
> > -	vma = binder_alloc_get_vma(alloc);
> > +	vma = vma_lookup(mm, page_addr);
> > +	if (vma && vma != binder_alloc_get_vma(alloc))
> > +		goto err_invalid_vma;
> 
> Doesn't this need to be:
> if (!vma || vma != binder_alloc_get_vma(alloc))
> 
> This way, we catch a different vma and a NULL vma.
> 
> Or even, just:
> if (vma != binder_alloc_get_vma(alloc))
> 
> if the alloc vma cannot be NULL?
> 

If the vma_lookup() is NULL then we still need to isolate and free the
given binder page and we obviously skip the zap() in this case.

However, if we receive a random unexpected vma because of a corrupted
address or similar, then the whole process is skipped.

Thus, why we use the check above.

--
Carlos Llamas

Re: [PATCH 02/21] binder: fix use-after-free in shinker's callback

[Liam R. Howlett]

* Carlos Llamas <cmllamas@google.com> [231102 16:09]:

...
> 
> > > diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
> > > index e3db8297095a..c4d60d81221b 100644
> > > --- a/drivers/android/binder_alloc.c
> > > +++ b/drivers/android/binder_alloc.c
> > > @@ -1005,7 +1005,9 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
> > >  		goto err_mmget;
> > >  	if (!mmap_read_trylock(mm))
> > >  		goto err_mmap_read_lock_failed;
> > > -	vma = binder_alloc_get_vma(alloc);
> > > +	vma = vma_lookup(mm, page_addr);
> > > +	if (vma && vma != binder_alloc_get_vma(alloc))
> > > +		goto err_invalid_vma;
> > 
> > Doesn't this need to be:
> > if (!vma || vma != binder_alloc_get_vma(alloc))
> > 
> > This way, we catch a different vma and a NULL vma.
> > 
> > Or even, just:
> > if (vma != binder_alloc_get_vma(alloc))
> > 
> > if the alloc vma cannot be NULL?
> > 
> 
> If the vma_lookup() is NULL then we still need to isolate and free the
> given binder page and we obviously skip the zap() in this case.

I would have thought if there was no VMA, then the entire process could
be avoided.  Thanks for clarifying.

> 
> However, if we receive a random unexpected vma because of a corrupted
> address or similar, then the whole process is skipped.
> 
> Thus, why we use the check above.
> 
> --
> Carlos Llamas

Re: [PATCH 01/21] binder: use EPOLLERR from eventpoll.h

[Alice Ryhl]

Carlos Llamas <cmllamas@google.com> writes:
> Use EPOLLERR instead of POLLERR to make sure it is cast to the correct
> __poll_t type. This fixes the following sparse issue:
> 
>   drivers/android/binder.c:5030:24: warning: incorrect type in return expression (different base types)
>   drivers/android/binder.c:5030:24:    expected restricted __poll_t
>   drivers/android/binder.c:5030:24:    got int
> 
> Fixes: f88982679f54 ("binder: check for binder_thread allocation failure in binder_poll()")
> Cc: stable@vger.kernel.org
> Cc: Eric Biggers <ebiggers@google.com>
> Signed-off-by: Carlos Llamas <cmllamas@google.com>

Reviewed-by: Alice Ryhl <aliceryhl@google.com>

Re: [PATCH 02/21] binder: fix use-after-free in shinker's callback

[Alice Ryhl]

Carlos Llamas <cmllamas@google.com> writes:
> The mmap read lock is used during the shrinker's callback, which means
> that using alloc->vma pointer isn't safe as it can race with munmap().
> As of commit dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in
> munmap") the mmap lock is downgraded after the vma has been isolated.
> 
> I was able to reproduce this issue by manually adding some delays and
> triggering page reclaiming through the shrinker's debug sysfs. The
> following KASAN report confirms the UAF:
> 
>   [...snip...]
> 
> Fix this issue by performing instead a vma_lookup() which will fail to
> find the vma that was isolated before the mmap lock downgrade. Note that
> this option has better performance than upgrading to a mmap write lock
> which would increase contention. Plus, mmap_write_trylock() has been
> recently removed anyway.
> 
> Fixes: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap")
> Cc: stable@vger.kernel.org
> Cc: Liam Howlett <liam.howlett@oracle.com>
> Cc: Minchan Kim <minchan@kernel.org>
> Signed-off-by: Carlos Llamas <cmllamas@google.com>

This change makes sense to me, and I agree that the code still needs to
run when the vma is null.

Reviewed-by: Alice Ryhl <aliceryhl@google.com>

Re: [PATCH 05/21] binder: fix trivial typo of binder_free_buf_locked()

[Alice Ryhl]

Carlos Llamas <cmllamas@google.com> writes:
> Fix minor misspelling of the function in the comment section.
> 
> No functional changes in this patch.
> 
> Cc: stable@vger.kernel.org
> Fixes: 0f966cba95c7 ("binder: add flag to clear buffer on txn complete")
> Signed-off-by: Carlos Llamas <cmllamas@google.com>

It's a bit confusing that the pair of methods binder_alloc_free_buf and
binder_free_buf_locked are inconsistent in whether they user the alloc_
prefix. It might be worth it to change them to be consistent?

Either way, this change LGTM.

Reviewed-by: Alice Ryhl <aliceryhl@google.com>

Re: [PATCH 06/21] binder: fix comment on binder_alloc_new_buf() return value

[Alice Ryhl]

Carlos Llamas <cmllamas@google.com> writes:
> Update the comments of binder_alloc_new_buf() to reflect that the return
> value of the function is now ERR_PTR(-errno) on failure.
> 
> No functional changes in this patch.
> 
> Cc: stable@vger.kernel.org
> Fixes: 57ada2fb2250 ("binder: add log information for binder transaction failures")
> Signed-off-by: Carlos Llamas <cmllamas@google.com>

Reviewed-by: Alice Ryhl <aliceryhl@google.com>

Re: [PATCH 05/21] binder: fix trivial typo of binder_free_buf_locked()

[Carlos Llamas]

On Tue, Nov 07, 2023 at 09:08:05AM +0000, Alice Ryhl wrote:
> It's a bit confusing that the pair of methods binder_alloc_free_buf and
> binder_free_buf_locked are inconsistent in whether they user the alloc_
> prefix. It might be worth it to change them to be consistent?

Right, the prefix is concistent across the APIs but not really for the
local function names. I wouldn't mind adding the "alloc" part, it just
seemed easier to fix the comment and backport that.

--
Carlos Llamas