From: Mauro Carvalho Chehab <mchehab@s-opensource.com>
To: Hans Verkuil <hverkuil@xs4all.nl>
Cc: linux-media@vger.kernel.org, Hans Verkuil <hans.verkuil@cisco.com>
Subject: Re: [RFCv11 PATCH 05/29] media-request: add request ioctls
Date: Tue, 10 Apr 2018 07:57:56 -0300 [thread overview]
Message-ID: <20180410075756.3975ed22@vento.lan> (raw)
In-Reply-To: <20180409142026.19369-6-hverkuil@xs4all.nl>
Em Mon, 9 Apr 2018 16:20:02 +0200
Hans Verkuil <hverkuil@xs4all.nl> escreveu:
> From: Hans Verkuil <hans.verkuil@cisco.com>
>
> Implement the MEDIA_REQUEST_IOC_QUEUE and MEDIA_REQUEST_IOC_REINIT
> ioctls.
>
> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
> ---
> drivers/media/media-request.c | 80 +++++++++++++++++++++++++++++++++++++++++--
> 1 file changed, 78 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/media/media-request.c b/drivers/media/media-request.c
> index dffc290e4ada..27739ff7cb09 100644
> --- a/drivers/media/media-request.c
> +++ b/drivers/media/media-request.c
> @@ -118,10 +118,86 @@ static unsigned int media_request_poll(struct file *filp,
> return 0;
> }
>
> +static long media_request_ioctl_queue(struct media_request *req)
> +{
> + struct media_device *mdev = req->mdev;
> + unsigned long flags;
> + int ret = 0;
> +
> + dev_dbg(mdev->dev, "request: queue %s\n", req->debug_str);
> +
> + spin_lock_irqsave(&req->lock, flags);
> + if (req->state != MEDIA_REQUEST_STATE_IDLE) {
> + dev_dbg(mdev->dev,
> + "request: unable to queue %s, request in state %s\n",
> + req->debug_str, media_request_state_str(req->state));
> + spin_unlock_irqrestore(&req->lock, flags);
> + return -EINVAL;
> + }
> + req->state = MEDIA_REQUEST_STATE_QUEUEING;
> +
> + spin_unlock_irqrestore(&req->lock, flags);
> +
> + /*
> + * Ensure the request that is validated will be the one that gets queued
> + * next by serialising the queueing process.
> + */
> + mutex_lock(&mdev->req_queue_mutex);
The locking here seems really weird. IMHO, it should lock before
touching state, as otherwise race conditions may happen.
As I suggested before, I would use an atomic type for state, and get rid
of the spin lock (as it seems that it is meant to be used just
for state).
> +
> + ret = mdev->ops->req_queue(req);
> + spin_lock_irqsave(&req->lock, flags);
> + req->state = ret ? MEDIA_REQUEST_STATE_IDLE : MEDIA_REQUEST_STATE_QUEUED;
> + spin_unlock_irqrestore(&req->lock, flags);
> + mutex_unlock(&mdev->req_queue_mutex);
> +
Here, you have both mutex and spin locked. This is a strong indication
that locks are not well designed, are you're using two different locks
to protect the same data.
> + if (ret) {
> + dev_dbg(mdev->dev, "request: can't queue %s (%d)\n",
> + req->debug_str, ret);
> + } else {
> + media_request_get(req);
> + }
> +
> + return ret;
> +}
IMHO, the above construction hides the main code inside an if due
to an error condition. Doing this makes it clearer that, under normal
circumstances, you're doing a kref_get() call:
if (ret) {
dev_dbg(mdev->dev, "request: can't queue %s (%d)\n",
req->debug_str, ret);
return ret;
}
kref_get(&req->kref); // This is a way more easier to read than media_request_get(req);
return 0;
}
Another related issue: IMHO, kref_get() should be called here with the
mutex hold.
> +static long media_request_ioctl_reinit(struct media_request *req)
> +{
> + struct media_device *mdev = req->mdev;
> + unsigned long flags;
> +
> + spin_lock_irqsave(&req->lock, flags);
> + if (req->state != MEDIA_REQUEST_STATE_IDLE &&
> + req->state != MEDIA_REQUEST_STATE_COMPLETE) {
> + dev_dbg(mdev->dev,
> + "request: %s not in idle or complete state, cannot reinit\n",
> + req->debug_str);
> + spin_unlock_irqrestore(&req->lock, flags);
> + return -EINVAL;
> + }
> + req->state = MEDIA_REQUEST_STATE_CLEANING;
> + spin_unlock_irqrestore(&req->lock, flags);
> +
> + media_request_clean(req);
> +
> + spin_lock_irqsave(&req->lock, flags);
> + req->state = MEDIA_REQUEST_STATE_IDLE;
> + spin_unlock_irqrestore(&req->lock, flags);
This code should be called with the mutex hold.
> + return 0;
> +}
> +
> static long media_request_ioctl(struct file *filp, unsigned int cmd,
> - unsigned long __arg)
> + unsigned long arg)
> {
> - return -ENOIOCTLCMD;
> + struct media_request *req = filp->private_data;
> +
> + switch (cmd) {
> + case MEDIA_REQUEST_IOC_QUEUE:
> + return media_request_ioctl_queue(req);
> + case MEDIA_REQUEST_IOC_REINIT:
> + return media_request_ioctl_reinit(req);
> + default:
> + return -ENOIOCTLCMD;
> + }
> }
>
> static const struct file_operations request_fops = {
Thanks,
Mauro
next prev parent reply other threads:[~2018-04-10 10:58 UTC|newest]
Thread overview: 135+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-04-09 14:19 [RFCv11 PATCH 00/29] Request API Hans Verkuil
2018-04-09 14:19 ` [RFCv11 PATCH 01/29] v4l2-device.h: always expose mdev Hans Verkuil
2018-04-17 4:34 ` Alexandre Courbot
2018-04-09 14:19 ` [RFCv11 PATCH 02/29] uapi/linux/media.h: add request API Hans Verkuil
2018-04-10 5:26 ` Tomasz Figa
2018-04-23 9:55 ` Hans Verkuil
2018-04-10 9:38 ` Mauro Carvalho Chehab
2018-04-10 11:00 ` Sakari Ailus
2018-04-23 11:41 ` Hans Verkuil
2018-04-23 13:32 ` Mauro Carvalho Chehab
2018-04-17 4:34 ` Alexandre Courbot
2018-04-09 14:20 ` [RFCv11 PATCH 03/29] media-request: allocate media requests Hans Verkuil
2018-04-10 5:35 ` Tomasz Figa
2018-04-10 9:54 ` Mauro Carvalho Chehab
2018-04-23 9:59 ` Hans Verkuil
2018-04-10 9:52 ` Mauro Carvalho Chehab
2018-04-10 11:14 ` Sakari Ailus
2018-04-11 10:13 ` Mauro Carvalho Chehab
2018-04-23 9:46 ` Hans Verkuil
2018-04-23 11:49 ` Hans Verkuil
2018-04-23 13:35 ` Mauro Carvalho Chehab
2018-04-17 4:34 ` Alexandre Courbot
2018-04-09 14:20 ` [RFCv11 PATCH 04/29] media-request: core request support Hans Verkuil
2018-04-10 8:21 ` Tomasz Figa
2018-04-10 13:04 ` Sakari Ailus
2018-04-23 11:24 ` Hans Verkuil
2018-04-23 12:14 ` Hans Verkuil
2018-04-10 10:32 ` Mauro Carvalho Chehab
2018-04-10 12:32 ` Sakari Ailus
2018-04-10 14:51 ` Mauro Carvalho Chehab
2018-04-11 13:21 ` Sakari Ailus
2018-04-11 13:49 ` Mauro Carvalho Chehab
2018-04-11 15:02 ` Sakari Ailus
2018-04-11 15:17 ` Mauro Carvalho Chehab
2018-04-11 15:35 ` Sakari Ailus
2018-04-11 16:13 ` Mauro Carvalho Chehab
2018-04-12 7:11 ` Sakari Ailus
2018-04-23 12:43 ` Hans Verkuil
2018-05-07 10:05 ` Sakari Ailus
2018-04-23 12:23 ` Hans Verkuil
2018-04-23 14:07 ` Mauro Carvalho Chehab
2018-04-10 12:17 ` Sakari Ailus
2018-04-17 4:35 ` Alexandre Courbot
2018-04-23 14:28 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 05/29] media-request: add request ioctls Hans Verkuil
2018-04-10 8:59 ` Tomasz Figa
2018-04-12 7:35 ` Sakari Ailus
2018-04-23 12:40 ` Hans Verkuil
2018-04-23 11:34 ` Hans Verkuil
2018-04-10 10:57 ` Mauro Carvalho Chehab [this message]
2018-04-12 10:40 ` Sakari Ailus
2018-04-23 13:45 ` Hans Verkuil
2018-04-12 7:31 ` Sakari Ailus
2018-04-09 14:20 ` [RFCv11 PATCH 06/29] media-request: add media_request_find Hans Verkuil
2018-04-10 11:04 ` Mauro Carvalho Chehab
2018-04-12 10:52 ` Sakari Ailus
2018-04-12 12:08 ` Sakari Ailus
2018-04-23 13:50 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 07/29] media-request: add media_request_object_find Hans Verkuil
2018-04-10 11:07 ` Mauro Carvalho Chehab
2018-04-10 11:10 ` Hans Verkuil
2018-04-12 12:23 ` Sakari Ailus
2018-04-23 13:55 ` Hans Verkuil
2018-04-17 4:36 ` Alexandre Courbot
2018-04-23 14:32 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 08/29] videodev2.h: add request_fd field to v4l2_ext_controls Hans Verkuil
2018-04-10 11:11 ` Mauro Carvalho Chehab
2018-04-12 12:24 ` Sakari Ailus
2018-04-09 14:20 ` [RFCv11 PATCH 09/29] videodev2.h: add V4L2_CTRL_FLAG_IN_REQUEST Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 10/29] v4l2-ctrls: v4l2_ctrl_add_handler: add from_other_dev Hans Verkuil
2018-04-10 11:42 ` Mauro Carvalho Chehab
2018-04-12 13:35 ` Sakari Ailus
2018-04-09 14:20 ` [RFCv11 PATCH 11/29] v4l2-ctrls: prepare internal structs for request API Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 12/29] v4l2-ctrls: alloc memory for p_req Hans Verkuil
2018-04-10 9:32 ` Tomasz Figa
2018-04-10 13:57 ` Mauro Carvalho Chehab
2018-04-23 11:39 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 13/29] v4l2-ctrls: use ref in helper instead of ctrl Hans Verkuil
2018-04-10 12:47 ` Mauro Carvalho Chehab
2018-04-09 14:20 ` [RFCv11 PATCH 14/29] v4l2-ctrls: add core request support Hans Verkuil
2018-04-11 8:37 ` Paul Kocialkowski
2018-04-11 9:43 ` Tomasz Figa
2018-04-23 9:23 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 15/29] v4l2-ctrls: support g/s_ext_ctrls for requests Hans Verkuil
2018-04-10 13:00 ` Mauro Carvalho Chehab
2018-04-09 14:20 ` [RFCv11 PATCH 16/29] videodev2.h: Add request_fd field to v4l2_buffer Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 17/29] vb2: store userspace data in vb2_v4l2_buffer Hans Verkuil
2018-04-09 15:11 ` Kieran Bingham
2018-04-23 9:53 ` Hans Verkuil
2018-04-10 13:32 ` Mauro Carvalho Chehab
2018-04-09 14:20 ` [RFCv11 PATCH 18/29] videobuf2-core: embed media_request_object Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 19/29] videobuf2-core: integrate with media requests Hans Verkuil
2018-04-12 8:13 ` Tomasz Figa
2018-04-23 12:49 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 20/29] videobuf2-v4l2: " Hans Verkuil
2018-04-10 13:52 ` Mauro Carvalho Chehab
2018-04-17 4:36 ` Alexandre Courbot
2018-04-23 14:53 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 21/29] videobuf2-core: add vb2_core_request_has_buffers Hans Verkuil
2018-04-10 13:53 ` Mauro Carvalho Chehab
2018-04-09 14:20 ` [RFCv11 PATCH 22/29] videobuf2-v4l2: add vb2_request_queue helper Hans Verkuil
2018-04-12 8:29 ` Tomasz Figa
2018-04-23 12:51 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 23/29] videobuf2-v4l2: export request_fd Hans Verkuil
2018-04-10 13:59 ` Mauro Carvalho Chehab
2018-04-09 14:20 ` [RFCv11 PATCH 24/29] Documentation: v4l: document request API Hans Verkuil
2018-04-10 14:19 ` Mauro Carvalho Chehab
2018-04-12 8:51 ` Tomasz Figa
2018-04-09 14:20 ` [RFCv11 PATCH 25/29] media: vim2m: add media device Hans Verkuil
2018-04-12 8:54 ` Tomasz Figa
2018-04-23 13:23 ` Hans Verkuil
2018-04-17 4:37 ` Alexandre Courbot
2018-04-09 14:20 ` [RFCv11 PATCH 26/29] vim2m: use workqueue Hans Verkuil
2018-04-11 14:06 ` Paul Kocialkowski
2018-04-30 14:51 ` Hans Verkuil
2018-04-12 9:02 ` Tomasz Figa
2018-04-30 14:58 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 27/29] vim2m: support requests Hans Verkuil
2018-04-11 14:01 ` Paul Kocialkowski
2018-04-12 9:15 ` Tomasz Figa
2018-04-23 13:30 ` Hans Verkuil
2018-04-17 4:37 ` Alexandre Courbot
2018-04-23 15:06 ` Hans Verkuil
2018-04-17 11:42 ` Paul Kocialkowski
2018-04-23 15:10 ` Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 28/29] vivid: add mc Hans Verkuil
2018-04-09 14:20 ` [RFCv11 PATCH 29/29] vivid: add request support Hans Verkuil
2018-04-10 14:31 ` [RFCv11 PATCH 00/29] Request API Mauro Carvalho Chehab
2018-04-17 4:33 ` Alexandre Courbot
2018-04-17 6:12 ` Hans Verkuil
2018-04-17 6:17 ` Alexandre Courbot
2018-04-17 11:40 ` Paul Kocialkowski
2018-04-18 2:06 ` Alexandre Courbot
2018-04-19 7:48 ` Paul Kocialkowski
2018-04-23 15:10 ` Hans Verkuil
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=20180410075756.3975ed22@vento.lan \
--to=mchehab@s-opensource.com \
--cc=hans.verkuil@cisco.com \
--cc=hverkuil@xs4all.nl \
--cc=linux-media@vger.kernel.org \
/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