From: Jonathan Nieder <jrnieder@gmail.com>
To: linux-media@vger.kernel.org
Cc: Huber Andreas <hobrom@corax.at>,
Mauro Carvalho Chehab <mchehab@redhat.com>,
Hans Verkuil <hverkuil@xs4all.nl>,
linux-kernel@vger.kernel.org, andrew.walker27@ntlworld.com,
Ben Hutchings <ben@decadent.org.uk>,
Trent Piepho <xyzzy@speakeasy.org>
Subject: [PATCH 1/3] [media] cx88: protect per-device driver list with device lock
Date: Sat, 2 Apr 2011 04:41:15 -0500 [thread overview]
Message-ID: <20110402094115.GB17015@elie> (raw)
In-Reply-To: <20110402093856.GA17015@elie>
The BKL conversion of this family of drivers seems to have gone wrong.
Opening cx88-blackbird will deadlock. Various other uses of the
sub-device and driver lists appear to be subject to race conditions.
For example: various functions access drvlist without a relevant
lock held, which will race against removal of drivers. Let's
start with that --- clean up by consistently protecting dev->drvlist
with dev->core->lock, noting driver functions that require the device
lock to be held or not held.
The only goal is to make the semantics clearer in preparation for
other changes. There are still some relevant races (noted in
comments) and the deadlock noticed by Andi remains; later patches
will address that.
Based-on-patch-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Cc: Andi Huber <hobrom@gmx.at>
Cc: stable@kernel.org
---
drivers/media/video/cx88/cx88-blackbird.c | 8 +++++++-
drivers/media/video/cx88/cx88-dvb.c | 8 ++++++++
drivers/media/video/cx88/cx88-mpeg.c | 11 +++++++----
drivers/media/video/cx88/cx88.h | 9 ++++++++-
4 files changed, 30 insertions(+), 6 deletions(-)
diff --git a/drivers/media/video/cx88/cx88-blackbird.c b/drivers/media/video/cx88/cx88-blackbird.c
index bca307e..85910c6 100644
--- a/drivers/media/video/cx88/cx88-blackbird.c
+++ b/drivers/media/video/cx88/cx88-blackbird.c
@@ -1122,10 +1122,16 @@ static int mpeg_release(struct file *file)
mutex_lock(&dev->core->lock);
file->private_data = NULL;
kfree(fh);
- mutex_unlock(&dev->core->lock);
/* Make sure we release the hardware */
drv = cx8802_get_driver(dev, CX88_MPEG_BLACKBIRD);
+ mutex_unlock(&dev->core->lock);
+
+ /*
+ * NEEDSWORK: the driver can be yanked from under our feet.
+ * The following really ought to be protected with core->lock.
+ */
+
if (drv)
drv->request_release(drv);
diff --git a/drivers/media/video/cx88/cx88-dvb.c b/drivers/media/video/cx88/cx88-dvb.c
index 7b8c9d3..5d0f947 100644
--- a/drivers/media/video/cx88/cx88-dvb.c
+++ b/drivers/media/video/cx88/cx88-dvb.c
@@ -133,7 +133,15 @@ static int cx88_dvb_bus_ctrl(struct dvb_frontend* fe, int acquire)
return -EINVAL;
}
+ mutex_lock(&dev->core->lock);
drv = cx8802_get_driver(dev, CX88_MPEG_DVB);
+ mutex_unlock(&dev->core->lock);
+
+ /*
+ * NEEDSWORK: The driver can be yanked from under our feet now.
+ * We ought to keep holding core->lock during the below.
+ */
+
if (drv) {
if (acquire){
dev->frontends.active_fe_id = fe_id;
diff --git a/drivers/media/video/cx88/cx88-mpeg.c b/drivers/media/video/cx88/cx88-mpeg.c
index addf954..918172b 100644
--- a/drivers/media/video/cx88/cx88-mpeg.c
+++ b/drivers/media/video/cx88/cx88-mpeg.c
@@ -748,6 +748,8 @@ int cx8802_unregister_driver(struct cx8802_driver *drv)
dev->pci->subsystem_device, dev->core->board.name,
dev->core->boardnr);
+ mutex_lock(&dev->core->lock);
+
list_for_each_entry_safe(d, dtmp, &dev->drvlist, drvlist) {
/* only unregister the correct driver type */
if (d->type_id != drv->type_id)
@@ -755,15 +757,14 @@ int cx8802_unregister_driver(struct cx8802_driver *drv)
err = d->remove(d);
if (err == 0) {
- mutex_lock(&drv->core->lock);
list_del(&d->drvlist);
- mutex_unlock(&drv->core->lock);
kfree(d);
} else
printk(KERN_ERR "%s/2: cx8802 driver remove "
"failed (%d)\n", dev->core->name, err);
}
+ mutex_unlock(&dev->core->lock);
}
return err;
@@ -827,6 +828,8 @@ static void __devexit cx8802_remove(struct pci_dev *pci_dev)
flush_request_modules(dev);
+ mutex_lock(&dev->core->lock);
+
if (!list_empty(&dev->drvlist)) {
struct cx8802_driver *drv, *tmp;
int err;
@@ -838,9 +841,7 @@ static void __devexit cx8802_remove(struct pci_dev *pci_dev)
list_for_each_entry_safe(drv, tmp, &dev->drvlist, drvlist) {
err = drv->remove(drv);
if (err == 0) {
- mutex_lock(&drv->core->lock);
list_del(&drv->drvlist);
- mutex_unlock(&drv->core->lock);
} else
printk(KERN_ERR "%s/2: cx8802 driver remove "
"failed (%d)\n", dev->core->name, err);
@@ -848,6 +849,8 @@ static void __devexit cx8802_remove(struct pci_dev *pci_dev)
}
}
+ mutex_unlock(&dev->core->lock);
+
/* Destroy any 8802 reference. */
dev->core->dvbdev = NULL;
diff --git a/drivers/media/video/cx88/cx88.h b/drivers/media/video/cx88/cx88.h
index 9b3742a..e3d56c2 100644
--- a/drivers/media/video/cx88/cx88.h
+++ b/drivers/media/video/cx88/cx88.h
@@ -506,7 +506,11 @@ struct cx8802_driver {
int (*resume)(struct pci_dev *pci_dev);
/* MPEG 8802 -> mini driver - Driver probe and configuration */
+
+ /* Caller must _not_ hold core->lock */
int (*probe)(struct cx8802_driver *drv);
+
+ /* Caller must hold core->lock */
int (*remove)(struct cx8802_driver *drv);
/* MPEG 8802 -> mini driver - Access for hardware control */
@@ -561,8 +565,9 @@ struct cx8802_dev {
/* for switching modulation types */
unsigned char ts_gen_cntrl;
- /* List of attached drivers */
+ /* List of attached drivers; must hold core->lock to access */
struct list_head drvlist;
+
struct work_struct request_module_wk;
};
@@ -685,6 +690,8 @@ int cx88_audio_thread(void *data);
int cx8802_register_driver(struct cx8802_driver *drv);
int cx8802_unregister_driver(struct cx8802_driver *drv);
+
+/* Caller must hold core->lock */
struct cx8802_driver * cx8802_get_driver(struct cx8802_dev *dev, enum cx88_board_type btype);
/* ----------------------------------------------------------- */
--
1.7.5.rc0
next prev parent reply other threads:[~2011-04-02 9:41 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20110327150610.4029.95961.reportbug@xen.corax.at>
2011-03-27 15:28 ` [linux-dvb] cx88-blackbird broken (since 2.6.37) Jonathan Nieder
2011-04-02 9:38 ` [RFC/PATCH 0/3] locking fixes for cx88 Jonathan Nieder
2011-04-02 9:41 ` Jonathan Nieder [this message]
2011-04-02 9:41 ` [PATCH 2/3] [media] cx88: fix locking of sub-driver operations Jonathan Nieder
2011-04-02 9:44 ` [PATCH 3/3] [media] cx88: use a mutex to protect cx8802_devlist Jonathan Nieder
[not found] ` <4D971B8D.4040305@corax.at>
2011-04-02 19:29 ` Jonathan Nieder
2011-04-02 20:13 ` Andreas Huber
2011-04-04 2:12 ` Andreas Huber
2011-04-05 1:16 ` Jonathan Nieder
2011-04-02 15:19 ` [RFC/PATCH 0/3] locking fixes for cx88 Ben Hutchings
2011-04-02 18:30 ` Jonathan Nieder
2011-04-05 3:20 ` [RFC/PATCH v2 0/7] " Jonathan Nieder
2011-04-05 3:22 ` [PATCH 1/7] [media] cx88: protect per-device driver list with device lock Jonathan Nieder
2011-04-05 3:25 ` [PATCH 2/7] [media] cx88: fix locking of sub-driver operations Jonathan Nieder
2011-04-05 3:27 ` [PATCH 3/7] [media] cx88: hold device lock during sub-driver initialization Jonathan Nieder
2011-04-05 3:27 ` [PATCH 4/7] [media] cx88: use a mutex to protect cx8802_devlist Jonathan Nieder
2011-04-05 3:29 ` [PATCH 5/7] [media] cx88: gracefully reject attempts to use unregistered cx88-blackbird driver Jonathan Nieder
2011-04-05 3:30 ` [PATCH 6/7] [media] cx88: don't use atomic_t for core->mpeg_users Jonathan Nieder
2011-04-05 3:31 ` [PATCH 7/7] [mpeg] cx88: don't use atomic_t for core->users Jonathan Nieder
2011-04-05 3:41 ` Jonathan Nieder
2011-04-05 18:17 ` [RFC/PATCH v2 0/7] locking fixes for cx88 Jonathan Nieder
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=20110402094115.GB17015@elie \
--to=jrnieder@gmail.com \
--cc=andrew.walker27@ntlworld.com \
--cc=ben@decadent.org.uk \
--cc=hobrom@corax.at \
--cc=hverkuil@xs4all.nl \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-media@vger.kernel.org \
--cc=mchehab@redhat.com \
--cc=xyzzy@speakeasy.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