From: John Keeping <john@metanate.com>
To: Linyu Yuan <quic_linyyuan@quicinc.com>
Cc: Felipe Balbi <balbi@kernel.org>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
linux-usb@vger.kernel.org, Michael Wu <michael@allwinnertech.com>
Subject: Re: [PATCH v1 1/2] usb: gadget: ffs: change ep->status safe in ffs_epfile_io()
Date: Tue, 31 May 2022 12:44:48 +0100 [thread overview]
Message-ID: <YpX/sILPw90Uvq+S@donbot> (raw)
In-Reply-To: <1653989775-14267-2-git-send-email-quic_linyyuan@quicinc.com>
On Tue, May 31, 2022 at 05:36:14PM +0800, Linyu Yuan wrote:
> If a task read/write data in blocking mode, it will wait the completion
> in ffs_epfile_io(), if function unbind occurs, ffs_func_unbind() will
> kfree ffs ep, once the task wake up, it still dereference the ffs ep to
> obtain the request status.
>
> Fix it by moving the request status to io_data which is stack-safe.
>
> Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com>
One minor issue below, but otherwise I like this approach and it will
help solving a similar issue on the AIO path which I spotted while
looking into this (ffs_func_unbind() calls ffs_func_eps_disable()
setting epfile->ep = NULL before draining the completion workqueue where
ffs_user_copy_worker() relies on epfile->ep being non-null).
> ---
> drivers/usb/gadget/function/f_fs.c | 32 +++++++++++++++++---------------
> 1 file changed, 17 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/usb/gadget/function/f_fs.c b/drivers/usb/gadget/function/f_fs.c
> index 4585ee3..dcba835 100644
> --- a/drivers/usb/gadget/function/f_fs.c
> +++ b/drivers/usb/gadget/function/f_fs.c
> @@ -122,8 +122,6 @@ struct ffs_ep {
> struct usb_endpoint_descriptor *descs[3];
>
> u8 num;
> -
> - int status; /* P: epfile->mutex */
> };
>
> struct ffs_epfile {
> @@ -227,6 +225,9 @@ struct ffs_io_data {
> bool use_sg;
>
> struct ffs_data *ffs;
> +
> + int status;
> + struct completion done;
> };
>
> struct ffs_desc_helper {
> @@ -707,12 +708,12 @@ static const struct file_operations ffs_ep0_operations = {
>
> static void ffs_epfile_io_complete(struct usb_ep *_ep, struct usb_request *req)
> {
> + struct ffs_io_data *io_data = req->context;
> +
> ENTER();
> - if (req->context) {
> - struct ffs_ep *ep = _ep->driver_data;
> - ep->status = req->status ? req->status : req->actual;
> - complete(req->context);
> - }
> +
> + io_data->status = req->status ? req->status : req->actual;
> + complete(&io_data->done);
> }
>
> static ssize_t ffs_copy_to_iter(void *data, int data_len, struct iov_iter *iter)
> @@ -1050,7 +1051,6 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
> WARN(1, "%s: data_len == -EINVAL\n", __func__);
> ret = -EINVAL;
> } else if (!io_data->aio) {
> - DECLARE_COMPLETION_ONSTACK(done);
> bool interrupted = false;
>
> req = ep->req;
> @@ -1066,7 +1066,8 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
>
> io_data->buf = data;
>
> - req->context = &done;
> + init_completion(&io_data->done);
> + req->context = io_data;
> req->complete = ffs_epfile_io_complete;
>
> ret = usb_ep_queue(ep->ep, req, GFP_ATOMIC);
> @@ -1075,7 +1076,7 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
>
> spin_unlock_irq(&epfile->ffs->eps_lock);
>
> - if (wait_for_completion_interruptible(&done)) {
> + if (wait_for_completion_interruptible(&io_data->done)) {
> /*
> * To avoid race condition with ffs_epfile_io_complete,
> * dequeue the request first then check
> @@ -1083,17 +1084,18 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
> * condition with req->complete callback.
> */
> usb_ep_dequeue(ep->ep, req);
> - wait_for_completion(&done);
> - interrupted = ep->status < 0;
> + wait_for_completion(&io_data->done);
> + interrupted = true;
This is a change in behaviour - shouldn't this be:
interrupted = io_data->status;
?
Otherwise data is lost unnecessarily if the request completes
successfully in the small window between cancellation and being
dequeued.
> }
>
> if (interrupted)
> ret = -EINTR;
> - else if (io_data->read && ep->status > 0)
> - ret = __ffs_epfile_read_data(epfile, data, ep->status,
> + else if (io_data->read && io_data->status > 0)
> + ret = __ffs_epfile_read_data(epfile, data, io_data->status,
> &io_data->data);
> else
> - ret = ep->status;
> + ret = io_data->status;
> +
> goto error_mutex;
> } else if (!(req = usb_ep_alloc_request(ep->ep, GFP_ATOMIC))) {
> ret = -ENOMEM;
> --
> 2.7.4
>
next prev parent reply other threads:[~2022-05-31 11:45 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-31 9:36 [PATCH v1 0/2] usb: f_fs: safe operation in ffs_epfile_io() Linyu Yuan
2022-05-31 9:36 ` [PATCH v1 1/2] usb: gadget: ffs: change ep->status safe " Linyu Yuan
2022-05-31 11:44 ` John Keeping [this message]
2022-05-31 13:06 ` Linyu Yuan
2022-05-31 9:36 ` [PATCH v1 2/2] usb: gadget: ffs: change ep->ep " Linyu Yuan
2022-05-31 11:47 ` John Keeping
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=YpX/sILPw90Uvq+S@donbot \
--to=john@metanate.com \
--cc=balbi@kernel.org \
--cc=gregkh@linuxfoundation.org \
--cc=linux-usb@vger.kernel.org \
--cc=michael@allwinnertech.com \
--cc=quic_linyyuan@quicinc.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