Linux USB
 help / color / mirror / Atom feed
From: Rituparna Warwatkar <rwarwatkar@gmail.com>
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: linux-usb@vger.kernel.org, linux-kernel@vger.kernel.org,
	Rituparna Warwatkar <rwarwatkar@gmail.com>,
	syzbot+54927260acba030187a6@syzkaller.appspotmail.com
Subject: [PATCH v2] usb: gadget: uvc: align XU descriptor pointers to fix kmemleak reports
Date: Tue,  4 Aug 2026 07:29:47 +0200	[thread overview]
Message-ID: <20260804052947.15277-1-rwarwatkar@gmail.com> (raw)

kmemleak reports the baSourceID and bmControls arrays allocated by the
UVC extension-unit configfs attributes as leaked, e.g.:

  BUG: memory leak
  unreferenced object 0xffff888114fee2c0 (size 8):
    __kmalloc_noprof
    uvcg_extension_ba_source_id_store
    configfs_write_iter
    vfs_write
    ksys_write

The arrays are not actually leaked: they are reachable through
xu->desc.baSourceID / xu->desc.bmControls and are freed when the
extension unit is removed. The problem is that struct
uvcg_extension_unit_descriptor is marked __packed, so these two heap
pointers are stored at unaligned offsets. kmemleak only scans memory on
pointer-aligned boundaries, so it never sees the pointers and reports
the arrays as unreferenced.

Unlike the UAPI struct uvc_extension_unit_descriptor, this is an
in-memory staging structure: baSourceID and bmControls are pointers,
not inline arrays, and the wire descriptor is assembled field by field
in UVC_COPY_XU_DESCRIPTOR(). So __packed is not needed for layout
correctness and only serves to misalign the pointers.

Drop __packed and move the two remaining scalar members (bControlSize
and iExtension) ahead of the pointers so that baSourceID lands on a
natural 8-byte boundary. This keeps the pointers aligned and visible to
kmemleak while leaving the structure hole-free and the same size as
before (40 bytes on 64-bit). The bLength..bNrInPins prefix is unchanged,
so the "memcpy(dst, desc, 22)" in UVC_COPY_XU_DESCRIPTOR() and the wire
descriptor layout are unaffected.

Reported-by: syzbot+54927260acba030187a6@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=54927260acba030187a6
Fixes: 0525210c9840 ("usb: gadget: uvc: Allow definition of XUs in configfs")
Suggested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Rituparna Warwatkar <rwarwatkar@gmail.com>
---
Changes in v2:
 - Rather than only dropping __packed (which left padding holes and
   grew the struct), reorder the members so bControlSize and
   iExtension precede the two pointers, grouping baSourceID and
   bmControls at a natural 8-byte boundary. The struct stays 40 bytes
   with no padding, and the bLength..bNrInPins prefix (and thus
   UVC_COPY_XU_DESCRIPTOR() and the wire layout) is unchanged.

 drivers/usb/gadget/function/uvc_configfs.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/gadget/function/uvc_configfs.h b/drivers/usb/gadget/function/uvc_configfs.h
index 9391614135e..049fb9e14e4 100644
--- a/drivers/usb/gadget/function/uvc_configfs.h
+++ b/drivers/usb/gadget/function/uvc_configfs.h
@@ -172,11 +172,11 @@ struct uvcg_extension_unit_descriptor {
        u8 guidExtensionCode[16];
        u8 bNumControls;
        u8 bNrInPins;
-       u8 *baSourceID;
        u8 bControlSize;
-       u8 *bmControls;
        u8 iExtension;
-} __packed;
+       u8 *baSourceID;
+       u8 *bmControls;
+};

 struct uvcg_extension {
        struct config_item item;
--
2.47.3


             reply	other threads:[~2026-08-04  5:30 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-08-04  5:29 Rituparna Warwatkar [this message]
2026-08-04  6:28 ` [PATCH v2] usb: gadget: uvc: align XU descriptor pointers to fix kmemleak reports Greg Kroah-Hartman

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=20260804052947.15277-1-rwarwatkar@gmail.com \
    --to=rwarwatkar@gmail.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-usb@vger.kernel.org \
    --cc=syzbot+54927260acba030187a6@syzkaller.appspotmail.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox