From: Greg KH <gregkh@suse.de>
To: linux-kernel@vger.kernel.org
Cc: stern@rowland.harvard.edu
Subject: [PATCH] driver core: Fix races in driver_detach()
Date: Mon, 20 Jun 2005 15:59:27 -0700 [thread overview]
Message-ID: <11193083672789@kroah.com> (raw)
In-Reply-To: <11193083672422@kroah.com>
[PATCH] driver core: Fix races in driver_detach()
This patch is intended for your "driver" tree. It fixes several subtle
races in driver_detach() and device_release_driver() in the driver-model
core.
The major change is to use klist_remove() rather than klist_del() when
taking a device off its driver's list. There's no other way to guarantee
that the list pointers will be updated before some other driver binds to
the device. For this to work driver_detach() can't use a klist iterator,
so the loop over the devices must be written out in full. In addition the
patch protects against the possibility that, when a driver and a device
are unregistered at the same time, one may be unloaded from memory before
the other is finished using it.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
---
commit c95a6b057b108c2b7add35cba1354f9af921349e
tree 5a312f634b0aec295201a93020ba025d840e5f21
parent 6623415687eaffef49429292ab062bb046ee3311
author Alan Stern <stern@rowland.harvard.edu> Fri, 06 May 2005 15:38:33 -0400
committer Greg Kroah-Hartman <gregkh@suse.de> Mon, 20 Jun 2005 15:15:28 -0700
drivers/base/dd.c | 51 ++++++++++++++++++++++++++++++++++++++-------------
1 files changed, 38 insertions(+), 13 deletions(-)
diff --git a/drivers/base/dd.c b/drivers/base/dd.c
--- a/drivers/base/dd.c
+++ b/drivers/base/dd.c
@@ -177,41 +177,66 @@ void driver_attach(struct device_driver
* @dev: device.
*
* Manually detach device from driver.
- * Note that this is called without incrementing the bus
- * reference count nor taking the bus's rwsem. Be sure that
- * those are accounted for before calling this function.
+ *
+ * __device_release_driver() must be called with @dev->sem held.
*/
-void device_release_driver(struct device * dev)
+
+static void __device_release_driver(struct device * dev)
{
struct device_driver * drv;
- down(&dev->sem);
- if (dev->driver) {
- drv = dev->driver;
+ drv = dev->driver;
+ if (drv) {
+ get_driver(drv);
sysfs_remove_link(&drv->kobj, kobject_name(&dev->kobj));
sysfs_remove_link(&dev->kobj, "driver");
- klist_del(&dev->knode_driver);
+ klist_remove(&dev->knode_driver);
if (drv->remove)
drv->remove(dev);
dev->driver = NULL;
+ put_driver(drv);
}
- up(&dev->sem);
}
-static int __remove_driver(struct device * dev, void * unused)
+void device_release_driver(struct device * dev)
{
- device_release_driver(dev);
- return 0;
+ /*
+ * If anyone calls device_release_driver() recursively from
+ * within their ->remove callback for the same device, they
+ * will deadlock right here.
+ */
+ down(&dev->sem);
+ __device_release_driver(dev);
+ up(&dev->sem);
}
+
/**
* driver_detach - detach driver from all devices it controls.
* @drv: driver.
*/
void driver_detach(struct device_driver * drv)
{
- driver_for_each_device(drv, NULL, NULL, __remove_driver);
+ struct device * dev;
+
+ for (;;) {
+ spin_lock_irq(&drv->klist_devices.k_lock);
+ if (list_empty(&drv->klist_devices.k_list)) {
+ spin_unlock_irq(&drv->klist_devices.k_lock);
+ break;
+ }
+ dev = list_entry(drv->klist_devices.k_list.prev,
+ struct device, knode_driver.n_node);
+ get_device(dev);
+ spin_unlock_irq(&drv->klist_devices.k_lock);
+
+ down(&dev->sem);
+ if (dev->driver == drv)
+ __device_release_driver(dev);
+ up(&dev->sem);
+ put_device(dev);
+ }
}
next prev parent reply other threads:[~2005-06-21 2:27 UTC|newest]
Thread overview: 92+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-06-20 22:55 [GIT PATCH] Driver core changes for 2.6.12 Greg KH
2005-06-20 22:59 ` [PATCH] sysfs_{create|remove}_link should take const char * Greg KH
2005-06-20 22:59 ` [PATCH] kobject_hotplug() should use kobject_name() Greg KH
2005-06-20 22:59 ` [PATCH] Make kobject's name be const char * Greg KH
2005-06-20 22:59 ` [PATCH] kset_hotplug_ops->name shoudl return " Greg KH
2005-06-20 22:59 ` [PATCH] make driver's name be " Greg KH
2005-06-20 22:59 ` [PATCH] Make attributes names " Greg KH
2005-06-20 22:59 ` [PATCH] sysfs: (driver/base) if show/store is missing return -EIO Greg KH
2005-06-20 22:59 ` [PATCH] sysfs: " Greg KH
2005-06-20 22:59 ` [PATCH] sysfs: (driver/pci) " Greg KH
2005-06-20 22:59 ` [PATCH] sysfs: (driver/block) " Greg KH
2005-06-20 22:59 ` [PATCH] sysfs: (rest) " Greg KH
2005-06-20 22:59 ` [PATCH] INPUT: move to use the new class code, instead of class_simple Greg KH
2005-06-20 22:59 ` [PATCH] tty: " Greg KH
2005-06-20 22:59 ` [PATCH] CLASS: move a "simple" class logic into the class core Greg KH
2005-06-20 22:59 ` [PATCH] class: convert sound/* to use the new class api instead of class_simple Greg KH
2005-06-20 22:59 ` [PATCH] USB: move the usb hcd code to use the new class code Greg KH
2005-06-20 22:59 ` [PATCH] class: convert drivers/block/* to use the new class api instead of class_simple Greg KH
2005-06-20 22:59 ` [PATCH] class: convert drivers/ieee1394/* " Greg KH
2005-06-20 22:59 ` [PATCH] class: convert drivers/char/* " Greg KH
2005-06-20 22:59 ` [PATCH] class: convert drivers/scsi/* " Greg KH
2005-06-20 22:59 ` [PATCH] class: convert drivers/* " Greg KH
2005-06-20 22:59 ` [PATCH] USB: trivial error path fix Greg KH
2005-06-20 22:59 ` [PATCH] class: convert arch/* to use the new class api instead of class_simple Greg KH
2005-06-20 22:59 ` [PATCH] class: convert the remaining class_simple users in the kernel to usee the new class api Greg KH
2005-06-20 22:59 ` [PATCH] class: add kerneldoc for the new class functions Greg KH
2005-06-20 22:59 ` [PATCH] class: remove class_simple code, as no one in the tree is using it anymore Greg KH
2005-06-20 22:59 ` [PATCH] fix "make mandocs" after class_simple.c removal Greg KH
2005-06-20 22:59 ` [PATCH] Add a semaphore to struct device to synchronize calls to its driver Greg KH
2005-06-20 22:59 ` [PATCH] fix up ipmi code after class_simple.c removal Greg KH
2005-06-20 22:59 ` [PATCH] Move device/driver code to drivers/base/dd.c Greg KH
2005-06-20 22:59 ` [PATCH] Use driver_for_each_device() instead of manually walking list Greg KH
2005-06-20 22:59 ` [PATCH] Use driver_for_each_device() in drivers/pnp/driver.c " Greg KH
2005-06-20 22:59 ` [PATCH] Add driver_for_each_device() Greg KH
2005-06-20 22:59 ` [PATCH] Add a klist to struct bus_type for its drivers Greg KH
2005-06-20 22:59 ` [PATCH] Add a klist to struct bus_type for its devices Greg KH
2005-06-20 22:59 ` [PATCH] Add initial implementation of klist helpers Greg KH
2005-06-20 22:59 ` [PATCH] Add a klist to struct device_driver for the devices bound to it Greg KH
2005-06-20 22:59 ` [PATCH] Remove the unused device_find() Greg KH
2005-06-20 22:59 ` [PATCH] Use bus_for_each_{dev,drv} for driver binding Greg KH
2005-06-20 22:59 ` [PATCH] add klist_node_attached() to determine if a node is on a list or not Greg KH
2005-06-20 22:59 ` [PATCH] Fix up USB to use klist_node_attached() instead of list_empty() on lists that will go away Greg KH
2005-06-20 22:59 ` [PATCH] Remove struct device::driver_list Greg KH
2005-06-20 22:59 ` [PATCH] Fix up bus code and remove use of rwsem Greg KH
2005-06-20 22:59 ` [PATCH] Remove struct device::bus_list Greg KH
2005-06-20 22:59 ` [PATCH] Don't reference NULL klist pointer in klist_remove() Greg KH
2005-06-20 22:59 ` [PATCH] Call klist_del() instead of klist_remove() Greg KH
2005-06-20 22:59 ` [PATCH] Use device_for_each_child() to unregister devices in scsi_remove_target() Greg KH
2005-06-20 22:59 ` [PATCH] use device_for_each_child() to properly access child devices Greg KH
2005-06-20 22:59 ` [PATCH] Use a klist for device child lists Greg KH
2005-06-20 22:59 ` [PATCH] Fix up bogus comment Greg KH
2005-06-20 22:59 ` [PATCH] driver core: change export symbol for driver_for_each_device() Greg KH
2005-06-20 22:59 ` [PATCH] Use device_for_each_child() to unregister devices in nodemgr_remove_host_dev() Greg KH
2005-06-20 22:59 ` [PATCH] use device_for_each_child() to properly access child devices Greg KH
2005-06-20 22:59 ` [PATCH] USB: fix build warning in usb core as pointed out by Andrew Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: fix bk-driver-core kills ppc64 Greg KH
2005-06-20 22:59 ` [PATCH] Fix typo in scdrv_init() Greg KH
2005-06-20 22:59 ` [PATCH] Driver core: Fix up the driver and device iterators to be quieter Greg KH
2005-06-20 22:59 ` [PATCH] usb: klist_node_attached() fix Greg KH
2005-06-20 22:59 ` [PATCH] sn: fixes due to driver core changes Greg KH
2005-06-20 22:59 ` Greg KH [this message]
2005-06-20 22:59 ` [PATCH] Driver Core: driver model doc update Greg KH
2005-06-20 22:59 ` [PATCH] Driver core: unregister_node() for hotplug use Greg KH
2005-06-20 22:59 ` [PATCH] usbcore: Don't call device_release_driver recursively Greg KH
2005-06-20 22:59 ` [PATCH] libfs: add simple attribute files Greg KH
2005-06-20 22:59 ` [PATCH] Driver core: change device_attribute callbacks Greg KH
2005-06-20 22:59 ` [PATCH] driver core: fix error handling in bus_add_device Greg KH
2005-06-20 22:59 ` [PATCH] Driver core: Documentation: update device attribute callbacks Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/base - drivers/i2c/chips/adm1026.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: arch: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/i2c/chips/adm1031.c - lm75.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/i2c/chips/lm77.c - max1619.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/i2c/chips/pc87360.c - w83627hf.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/char/raw3270.c - drivers/net/netiucv.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/i2c/chips/w83781d.c - drivers/s390/block/dcssblk.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/usb/input/aiptek.c - drivers/zorro/zorro-sysfs.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: drivers/s390/net/qeth_sys.c - drivers/usb/gadget/pxa2xx_udc.c: " Greg KH
2005-06-20 22:59 ` [PATCH] Driver Core: include: " Greg KH
2005-06-20 22:59 ` [PATCH] I2C: drivers/i2c/chips/adm1026.c: use dynamic sysfs callbacks Greg KH
2005-06-20 22:59 ` [PATCH] I2C: add i2c sensor_device_attribute and macros Greg KH
2005-06-20 22:59 ` [PATCH] sysfs-iattr: attach sysfs_dirent before new inode Greg KH
2005-06-20 22:59 ` [PATCH] Driver core: Don't "lose" devices on suspend on failure Greg KH
2005-06-20 22:59 ` [PATCH] sysfs-iattr: set inode attributes Greg KH
2005-06-20 22:59 ` [PATCH] sysfs-iattr: add sysfs_setattr Greg KH
2005-06-20 22:59 ` [PATCH] SYSFS: fix PAGE_SIZE check Greg KH
2005-06-20 22:59 ` [PATCH] USB: fix show_modalias() function due to attribute change Greg KH
2005-06-20 22:59 ` [PATCH] PCI: " Greg KH
2005-07-06 0:38 ` [PATCH] Use device_for_each_child() to unregister devices in scsi_remove_target() Patrick Mansfield
2005-07-12 0:20 ` Greg KH
2005-07-12 0:56 ` Patrick Mansfield
2005-06-21 11:42 ` [PATCH] Add initial implementation of klist helpers Rik van Riel
2005-06-21 23:13 ` Greg KH
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=11193083672789@kroah.com \
--to=gregkh@suse.de \
--cc=greg@kroah.com \
--cc=linux-kernel@vger.kernel.org \
--cc=stern@rowland.harvard.edu \
/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