All of lore.kernel.org
 help / color / mirror / Atom feed
From: Hui Peng <benquike@gmail.com>
To: gregkh@linuxfoundation.org, arve@android.com, tkjos@android.com,
	maco@android.com, brauner@kernel.org, cmllamas@google.com
Cc: maco@google.com, linux-kernel@vger.kernel.org,
	stable@vger.kernel.org, Hui Peng <benquike@gmail.com>
Subject: [PATCH v2 1/2] binder: defer fd_install() until after copy_to_user() in binder_thread_read()
Date: Thu, 24 Sep 2026 10:00:18 +0000	[thread overview]
Message-ID: <20260924100019.3389555-2-benquike@gmail.com> (raw)
In-Reply-To: <20260919213650.3316812-1-benquike@gmail.com>

In binder_thread_read(), binder_apply_fd_fixups() installs translated file
descriptors into current->files via fd_install() BEFORE put_user() and
copy_to_user() copy the BR_TRANSACTION / BR_REPLY command to userspace.

If put_user() or copy_to_user() returns -EFAULT, any file descriptors in
t->fd_fixups have already been installed into the recipient process's file
descriptor table and removed from t->fd_fixups, even though the transaction
failed and the sender receives BR_FAILED_REPLY.

Split fd_install() out of binder_apply_fd_fixups() into
binder_fd_fixups_install(t) and invoke it only after put_user() and
copy_to_user() succeed.

Tested in QEMU against Linux 7.3.0-rc3 using a multi-process C reproducer:
a sender thread transmits a pipe write-end FD via BINDER_TYPE_FD in a
one-way transaction to a recipient thread whose read buffer spans an
unmapped virtual memory page, triggering -EFAULT in copy_to_user(): on the
unfixed kernel, the pipe write-end FD was prematurely installed into
current->files (polling the read-end did NOT produce POLLHUP); whereas
with this fix applied, fd_install() is deferred past copy_to_user(), so
the FD is not installed on -EFAULT (polling the read-end produces POLLHUP).

Fixes: 44d8047f1d87 ("binder: use standard functions to allocate fds")
Cc: stable@vger.kernel.org
Reviewed-by: Carlos Llamas <cmllamas@google.com>
Assisted-by: LLM
Signed-off-by: Hui Peng <benquike@gmail.com>
---
Changes in v2:
- Split out as patch 1/2 as requested by Greg Kroah-Hartman and Carlos
  Llamas.
- Added QEMU test procedure details in commit message body.
- Added Reviewed-by tag from Carlos Llamas.

 drivers/android/binder.c | 22 ++++++++++++++++------
 1 file changed, 16 insertions(+), 6 deletions(-)

diff --git a/drivers/android/binder.c b/drivers/android/binder.c
index fb9ff072bd6c..ee01e6a06631 100644
--- a/drivers/android/binder.c
+++ b/drivers/android/binder.c
@@ -4705,7 +4705,7 @@ static int binder_wait_for_work(struct binder_thread *thread,
 static int binder_apply_fd_fixups(struct binder_proc *proc,
 				  struct binder_transaction *t)
 {
-	struct binder_txn_fd_fixup *fixup, *tmp;
+	struct binder_txn_fd_fixup *fixup;
 	int ret = 0;
 
 	list_for_each_entry(fixup, &t->fd_fixups, fixup_entry) {
@@ -4730,19 +4730,25 @@ static int binder_apply_fd_fixups(struct binder_proc *proc,
 			goto err;
 		}
 	}
-	list_for_each_entry_safe(fixup, tmp, &t->fd_fixups, fixup_entry) {
-		fd_install(fixup->target_fd, fixup->file);
-		list_del(&fixup->fixup_entry);
-		kfree(fixup);
-	}
 
-	return ret;
+	return 0;
 
 err:
 	binder_free_txn_fixups(t);
 	return ret;
 }
 
+static void binder_fd_fixups_install(struct binder_transaction *t)
+{
+	struct binder_txn_fd_fixup *fixup, *tmp;
+
+	list_for_each_entry_safe(fixup, tmp, &t->fd_fixups, fixup_entry) {
+		fd_install(fixup->target_fd, fixup->file);
+		list_del(&fixup->fixup_entry);
+		kfree(fixup);
+	}
+}
+
 static int binder_thread_read(struct binder_proc *proc,
 			      struct binder_thread *thread,
 			      binder_uintptr_t binder_buffer, size_t size,
@@ -5141,6 +5147,8 @@ static int binder_thread_read(struct binder_proc *proc,
 		}
 		ptr += trsize;
 
+		binder_fd_fixups_install(t);
+
 		trace_binder_transaction_received(t);
 		binder_stat_br(proc, thread, cmd);
 		binder_debug(BINDER_DEBUG_TRANSACTION,

  parent reply	other threads:[~2026-09-24 10:00 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-19 21:36 [PATCH] binder: fix buffer/node leak and premature fd_install on read -EFAULT Hui Peng
2026-09-23 22:54 ` Carlos Llamas
2026-09-24 10:00 ` [PATCH v2 0/2] binder: fix premature fd_install() and buffer leak " Hui Peng
2026-10-01 12:37   ` Greg KH
2026-09-24 10:00 ` Hui Peng [this message]
2026-09-24 10:00 ` [PATCH v2 2/2] binder: free transaction buffer via binder_free_buf() " Hui Peng

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260924100019.3389555-2-benquike@gmail.com \
    --to=benquike@gmail.com \
    --cc=arve@android.com \
    --cc=brauner@kernel.org \
    --cc=cmllamas@google.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=maco@android.com \
    --cc=maco@google.com \
    --cc=stable@vger.kernel.org \
    --cc=tkjos@android.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.