From: Alan Stern <[email protected]> There is a potential deadlock in the driver core. It boils down to the fact that bus_remove_device() calls klist_remove() instead of klist_del(), thereby waiting until the reference count of the klist_node in the bus's klist of devices drops to 0. The refcount can't reach 0 so long as a modprobe process is trying to bind a new driver to the device being removed, by calling __driver_attach(). The problem is that __driver_attach() tries to acquire the device's parent's semaphore, but the caller of bus_remove_device() is quite likely to own that semaphore already. It isn't sufficient just to replace klist_remove() with klist_del(). Doing so runs the risk that the device would remain on the bus's klist of devices for some time, and so could be bound to another driver even after it was unregistered. What's needed is a new way to distinguish whether or not a device is registered, based on a criterion other than whether its klist_node is linked into the bus's klist of devices. That way driver binding can fail when the device is unregistered, even if it is still linked into the klist. This patch (as782) implements the solution, by adding a new bitflag to indiate when a struct device is registered, by testing the flag before allowing a driver to bind a device, and by changing the definition of the device_is_registered() inline. Signed-off-by: Alan Stern <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]> --- drivers/base/bus.c | 8 ++++++-- drivers/base/dd.c | 2 ++ include/linux/device.h | 3 ++- 3 files changed, 10 insertions(+), 3 deletions(-) diff --git a/drivers/base/bus.c b/drivers/base/bus.c index aa685a2..636af53 100644 --- a/drivers/base/bus.c +++ b/drivers/base/bus.c @@ -392,6 +392,7 @@ out: * bus_attach_device - add device to bus * @dev: device tried to attach to a driver * + * - Add device to bus's list of devices. * - Try to attach to driver. */ int bus_attach_device(struct device * dev) @@ -400,11 +401,13 @@ int bus_attach_device(struct device * de int ret = 0; if (bus) { + dev->is_registered = 1; ret = device_attach(dev); if (ret >= 0) { klist_add_tail(&dev->knode_bus, &bus->klist_devices); ret = 0; - } + } else + dev->is_registered = 0; } return ret; } @@ -425,7 +428,8 @@ void bus_remove_device(struct device * d sysfs_remove_link(&dev->kobj, "bus"); sysfs_remove_link(&dev->bus->devices.kobj, dev->bus_id); device_remove_attrs(dev->bus, dev); - klist_remove(&dev->knode_bus); + dev->is_registered = 0; + klist_del(&dev->knode_bus); pr_debug("bus %s: remove device %s\n", dev->bus->name, dev->bus_id); device_release_driver(dev); put_bus(dev->bus); diff --git a/drivers/base/dd.c b/drivers/base/dd.c index 319a73b..b5f43c3 100644 --- a/drivers/base/dd.c +++ b/drivers/base/dd.c @@ -162,6 +162,8 @@ int driver_probe_device(struct device_dr struct task_struct *probe_task; int ret = 0; + if (!device_is_registered(dev)) + return -ENODEV; if (drv->bus->match && !drv->bus->match(dev, drv)) goto done; diff --git a/include/linux/device.h b/include/linux/device.h index 74246ef..662e6a1 100644 --- a/include/linux/device.h +++ b/include/linux/device.h @@ -329,6 +329,7 @@ struct device { struct kobject kobj; char bus_id[BUS_ID_SIZE]; /* position on parent bus */ + unsigned is_registered:1; struct device_attribute uevent_attr; struct device_attribute *devt_attr; @@ -381,7 +382,7 @@ dev_set_drvdata (struct device *dev, voi static inline int device_is_registered(struct device *dev) { - return klist_node_attached(&dev->knode_bus); + return dev->is_registered; } /* -- 1.4.2.1 - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
- Follow-Ups:
- References:
- [GIT PATCH] Driver Core patches for 2.6.18
- From: Greg KH <[email protected]>
- [PATCH 1/47] Documentation/ABI: devfs is not obsolete, but removed!
- From: Greg KH <[email protected]>
- [PATCH 2/47] deprecate PHYSDEV* keys
- From: Greg KH <[email protected]>
- [PATCH 3/47] class_device_create(): make fmt argument 'const char *'
- From: Greg KH <[email protected]>
- [PATCH 4/47] device_create(): make fmt argument 'const char *'
- From: Greg KH <[email protected]>
- [PATCH 5/47] Driver core: add const to class_create
- From: Greg KH <[email protected]>
- [PATCH 6/47] sysfs: Make poll behaviour consistent
- From: Greg KH <[email protected]>
- [PATCH 7/47] Debugfs: kernel-doc fixes for debugfs
- From: Greg KH <[email protected]>
- [PATCH 8/47] SYSFS: allow sysfs_create_link to create symlinks in the root of sysfs
- From: Greg KH <[email protected]>
- [PATCH 9/47] Suspend infrastructure cleanup and extension
- From: Greg KH <[email protected]>
- [PATCH 10/47] Suspend changes for PCI core
- From: Greg KH <[email protected]>
- [PATCH 11/47] make suspend quieter
- From: Greg KH <[email protected]>
- [PATCH 12/47] fix broken/dubious driver suspend() methods
- From: Greg KH <[email protected]>
- [PATCH 13/47] PM: define PM_EVENT_PRETHAW
- From: Greg KH <[email protected]>
- [PATCH 14/47] PM: PCI and IDE handle PM_EVENT_PRETHAW
- From: Greg KH <[email protected]>
- [PATCH 15/47] PM: video drivers and PM_EVENT_PRETHAW
- From: Greg KH <[email protected]>
- [PATCH 16/47] PM: USB HCDs use PM_EVENT_PRETHAW
- From: Greg KH <[email protected]>
- [PATCH 17/47] PM: issue PM_EVENT_PRETHAW
- From: Greg KH <[email protected]>
- [PATCH 18/47] updated Documentation/power/devices.txt
- From: Greg KH <[email protected]>
- [PATCH 19/47] PM: update docs for writing .../power/state
- From: Greg KH <[email protected]>
- [PATCH 20/47] PM: add kconfig option for deprecated .../power/state files
- From: Greg KH <[email protected]>
- [PATCH 22/47] PM: no suspend_prepare() phase
- From: Greg KH <[email protected]>
- [PATCH 23/47] PM: add /sys/power documentation to Documentation/ABI
- From: Greg KH <[email protected]>
- [PATCH 24/47] PM: device_suspend/resume may sleep
- From: Greg KH <[email protected]>
- [PATCH 25/47] PM: platform_bus and late_suspend/early_resume
- From: Greg KH <[email protected]>
- [PATCH 26/47] Driver core: add groups support to struct device
- From: Greg KH <[email protected]>
- [PATCH 27/47] Driver core: allow devices in classes to have no parent
- From: Greg KH <[email protected]>
- [PATCH 28/47] Driver core: add ability for classes to handle devices properly
- From: Greg KH <[email protected]>
- [PATCH 29/47] Driver core: add device_rename function
- From: Greg KH <[email protected]>
- [PATCH 30/47] Driver core: create devices/virtual/ tree
- From: Greg KH <[email protected]>
- [PATCH 31/47] Class: add support for class interfaces for devices
- From: Greg KH <[email protected]>
- [PATCH 32/47] Driver core: add ability for devices to create and remove bin files
- From: Greg KH <[email protected]>
- [PATCH 33/47] kobject: must_check fixes
- From: Greg KH <[email protected]>
- [PATCH 34/47] sysfs_remove_bin_file: no return value, dump_stack on error
- From: Greg KH <[email protected]>
- [PATCH 35/47] Driver core: fix comments in drivers/base/power/resume.c
- From: Greg KH <[email protected]>
- [PATCH 36/47] Driver core: fixed add_bind_files() definition
- From: Greg KH <[email protected]>
- [PATCH 37/47] add __must_check to device management code
- From: Greg KH <[email protected]>
- [PATCH 38/47] add CONFIG_ENABLE_MUST_CHECK
- From: Greg KH <[email protected]>
- [PATCH 39/47] v4l-dev2: handle __must_check
- From: Greg KH <[email protected]>
- [PATCH 40/47] drivers/base: Platform notify needs to occur before drivers attach to the device
- From: Greg KH <[email protected]>
- [PATCH 41/47] drivers/base: check errors
- From: Greg KH <[email protected]>
- [PATCH 43/47] Driver Core: add ability for drivers to do a threaded probe
- From: Greg KH <[email protected]>
- [PATCH 44/47] PCI: enable driver multi-threaded probe
- From: Greg KH <[email protected]>
- [GIT PATCH] Driver Core patches for 2.6.18
- Prev by Date: [PATCH 44/47] PCI: enable driver multi-threaded probe
- Next by Date: [PATCH 47/47] Driver core: Don't call put methods while holding a spinlock
- Previous by thread: [PATCH 44/47] PCI: enable driver multi-threaded probe
- Next by thread: [PATCH 46/47] Driver core: Remove unneeded routines from driver core
- Index(es):