drivercore: fix a corner case for deferred probe
PATCH 1/1 drivercore: fix a corner case for deferred probe
There is one corner case in deferred probe which will lead a device in "dream" in the deferred_probe_pending_list.
Suppose we have three devices, Tom, Jerry and Spike. Tom and Jerry have a close relationship, Tom could be up until Jerry is up. Spike is an independent device.
Device probe sequence: Tom -> Spike -> Jerry
1. Tom probes, fails for deferred probe
adds itself to pending list
2. Spike probes, succeed
move devices in pending list to active list
trigger deferred probe
3. Tom is taken off from active list
probes and still fails, scheduled out
not added to pending list this time
(Tom is not in pending list neither in active list)
4. Jerry probes, succeed
move devices in pending list to active list(but Tom is not there) trigger deferred probe
go through the active list
5. Tom add itself to pending list
and wait
Tom will be trapped in the pending list until someone else help it out.
This patch adds a counter of success probe. Every time a driver probe succeeds, this is increased by 1. In the deferred_probe_work_func, when probe fails and returns EPROBE_DEFER, it checks this counter. If some driver succeed to probe during this period, it adds itself to active list again.
--- drivers/base/base.h | 2 +- drivers/base/bus.c | 3 ++- drivers/base/dd.c | 14 +++++++++++++- 3 files changed, 16 insertions(+), 3 deletions(-)
drivers/base/base.h +1 -1
@@ -105,7 +105,7 @@ extern void container_dev_init(void);struct kobject *virtual_device_parent (struct device *dev);extern int bus_add_device (struct device *dev); -extern void bus_probe_device (struct device *dev); +extern int bus_probe_device (struct device *dev);extern void bus_remove_device (struct device *dev);extern int bus_add_driver (struct device_driver *drv);
drivers/base/bus.c +2 -1
@@ -543,7 +543,7 @@ out_put:* * - Automatically probe for a driver if the bus allows it. */ -void bus_probe_device (struct device *dev) +int bus_probe_device (struct device *dev) {struct bus_type *bus = dev->bus;struct subsys_interface *sif;
@@ -562,6 +562,7 @@ void bus_probe_device(struct device *dev)if (sif->add_dev) sif->add_dev (dev, sif);mutex_unlock (&bus->p->mutex); +return ret; }/**
drivers/base/dd.c +13 -1
@@ -52,6 +52,7 @@ static DEFINE_MUTEX(deferred_probe_mutex);static LIST_HEAD (deferred_probe_pending_list);static LIST_HEAD (deferred_probe_active_list);static struct workqueue_struct *deferred_wq; +static u32 success_probe;/** * deferred_probe_work_func() - Retry probing devices in the active list.
@@ -60,6 +61,8 @@ static void deferred_probe_work_func(struct work_struct *work) {struct device *dev;struct device_private *private; +u32 old_success; +int ret =0 ;/* * This block processes every device in the deferred 'active' list. * Each device is removed from the active list and passed to
@@ -80,6 +83,7 @@ static void deferred_probe_work_func(struct work_struct *work)list_del_init (&private->deferred_probe);get_device (dev); + old_success =ACCESS_ONCE (success_probe);/* * Drop the mutex while probing each device; the probe path may
@@ -98,7 +102,14 @@ static void deferred_probe_work_func(struct work_struct *work)device_pm_unlock ();dev_dbg (dev,"Retrying from deferred list\n" ); -bus_probe_device (dev); + ret =bus_probe_device (dev); +if (ret == -EPROBE_DEFER ) { +mutex_lock (&deferred_probe_mutex); +if (old_success != success_probe) +list_move (&private->deferred_probe, + &deferred_probe_active_list); +mutex_unlock (&deferred_probe_mutex); + }mutex_lock (&deferred_probe_mutex);
@@ -147,6 +158,7 @@ static void driver_deferred_probe_trigger(void)* into the active list so they can be retried by the workqueue */ mutex_lock (&deferred_probe_mutex); + success_probe++;list_splice_tail_init (&deferred_probe_pending_list, &deferred_probe_active_list);mutex_unlock (&deferred_probe_mutex);
Signature · 1 line
1.7.9.5
Hi, all
Anyone has some comment on this?
Quoted 128 lines · Wei Yang
There is one corner case in deferred probe which will lead a device in "dream" in the
deferred_probe_pending_list.Suppose we have three devices, Tom, Jerry and Spike. Tom and Jerry have a close relationship, Tom could be up until Jerry is up. Spike is an independent device.
Device probe sequence: Tom -> Spike -> Jerry
1. Tom probes, fails for deferred probe
adds itself to pending list
2. Spike probes, succeed
move devices in pending list to active list
trigger deferred probe
3. Tom is taken off from active list
probes and still fails, scheduled out
not added to pending list this time
(Tom is not in pending list neither in active list)
4. Jerry probes, succeed
move devices in pending list to active list(but Tom is not there) trigger deferred probe
go through the active list
5. Tom add itself to pending list
and waitTom will be trapped in the pending list until someone else help it out.
This patch adds a counter of success probe. Every time a driver probe succeeds, this is increased by 1. In the
deferred_probe_work_func, when probe fails and returnsEPROBE_DEFER, it checks this counter. If some driver succeed to probe during this period, it adds itself to active list again.Signed-off-by Wei Yang <[email protected]>---
drivers/base/base.h | 2 +- drivers/base/bus.c | 3 ++- drivers/base/dd.c | 14 +++++++++++++- 3 files changed, 16 insertions(+), 3 deletions(-)drivers/base/base.h +1
-1@@ -105,7 +105,7 @@ extern void container_dev_init(void);struct kobject *virtual_device_parent (struct device *dev);extern int bus_add_device (struct device *dev); -extern void bus_probe_device (struct device *dev); +extern int bus_probe_device (struct device *dev);extern void bus_remove_device (struct device *dev);extern int bus_add_driver (struct device_driver *drv);drivers/base/bus.c +2
-1@@ -543,7 +543,7 @@ out_put:* * - Automatically probe for a driver if the bus allows it. */ -void bus_probe_device (struct device *dev) +int bus_probe_device (struct device *dev) {struct bus_type *bus = dev->bus;struct subsys_interface *sif;@@ -562,6 +562,7 @@ void bus_probe_device(struct device *dev)if (sif->add_dev) sif->add_dev (dev, sif);mutex_unlock (&bus->p->mutex); +return ret; }/** drivers/base/dd.c +13
-1@@ -52,6 +52,7 @@ static DEFINE_MUTEX(deferred_probe_mutex);static LIST_HEAD (deferred_probe_pending_list);static LIST_HEAD (deferred_probe_active_list);static struct workqueue_struct *deferred_wq; +static u32 success_probe;/** * deferred_probe_work_func() - Retry probing devices in the active list. @@ -60,6 +61,8 @@ static void deferred_probe_work_func(struct work_struct *work) {struct device *dev;struct device_private *private; +u32 old_success; +int ret =0 ;/* * This block processes every device in the deferred 'active' list. * Each device is removed from the active list and passed to @@ -80,6 +83,7 @@ static void deferred_probe_work_func(struct work_struct *work)list_del_init (&private->deferred_probe);get_device (dev); + old_success =ACCESS_ONCE (success_probe);/* * Drop the mutex while probing each device; the probe path may @@ -98,7 +102,14 @@ static void deferred_probe_work_func(struct work_struct *work)device_pm_unlock ();dev_dbg (dev,"Retrying from deferred list\n" ); -bus_probe_device (dev); + ret =bus_probe_device (dev); +if (ret == -EPROBE_DEFER ) { +mutex_lock (&deferred_probe_mutex); +if (old_success != success_probe) +list_move (&private->deferred_probe, + &deferred_probe_active_list); +mutex_unlock (&deferred_probe_mutex); + }mutex_lock (&deferred_probe_mutex);@@ -147,6 +158,7 @@ static void driver_deferred_probe_trigger(void)* into the active list so they can be retried by the workqueue */ mutex_lock (&deferred_probe_mutex); + success_probe++;list_splice_tail_init (&deferred_probe_pending_list, &deferred_probe_active_list);mutex_unlock (&deferred_probe_mutex);Signature · 1 line
1.7.9.5
Signature · 2 lines
Richard Yang Help you, Help me
Greg KH
maintainerQuoted 3 lines · Wei Yang
Hi, all
Anyone has some comment on this?
Did you miss the patch from Grant that is now in Linus's tree that should resolve this issue?
thanks,
greg k-h
Quoted 7 lines · Greg KH
Quoted 3 lines · Wei Yang
Hi, all
Anyone has some comment on this?
Did you miss the patch from Grant that is now in Linus's tree that should resolve this issue?
Hi, Greg,
Nice to hear from you.
Would you mind telling me the exact tree and commit? I searched in the git.kernel.org, and see several Linus's tree, including Linus Torvalds and Linus Walleij. So, not sure which you are pointing to. With the exact commit and tree, I could take a more deep look into it :-)
BTW, I am using this tree:git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
based on commit "455c6fd Linux 3.14".
Quoted 4 lines
thanks,
greg k-h
Signature · 2 lines
Richard Yang Help you, Help me
Greg KH
maintainerQuoted 9 lines · Wei Yang
Quoted 7 lines · Greg KH
Quoted 3 lines · Wei Yang
Hi, all
Anyone has some comment on this?
Did you miss the patch from Grant that is now in Linus's tree that should resolve this issue?
Quoted 8 lines
Hi, Greg,
Nice to hear from you.
Would you mind telling me the exact tree and commit? I searched in the git.kernel.org, and see several Linus's tree, including Linus Torvalds and Linus Walleij. So, not sure which you are pointing to. With the exact commit and tree, I could take a more deep look into it :-)
Use Linus Torvalds's tree, and look at commit 58b116bce13612e5aa6fcd49ecbd4cf8bb59e835 "drivercore: deferral race condition fix"
Quoted 3 lines
BTW, I am using this tree:
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
based on commit "455c6fd Linux 3.14".
That is the right tree, but a very old commit, please update your tree.
thanks,
greg k-h