workqueue: another assorted flush_scheduled_work() removals
Subsystems (9)
PATCH 1/1 workqueue: another assorted flush_scheduled_work() removals
Hello,
This is another set of patches to remove flush_scheduled_work() users.
0001-init-don-t-call-flush_scheduled_work-from-do_initcal.patch 0002-ioc4-use-static-work_struct-for-ioc4_load_modules.patch 0003-media-video-explicitly-flush-request_module-work.patch 0004-media-video-don-t-use-flush_scheduled_work.patch 0005-speedtch-don-t-abuse-struct-delayed_work.patch 0006-usb-don-t-use-flush_scheduled_work.patch
Andrew, implementing proper work function in init text test requires tracking an additional state for each worker. I don't think it'll be worth the effort or complexity as removal of init sections itself will make the bugs, if exists, quite visible anyway, so I just updated patch description of 0001. What do you think?
0002 updates the ioc4 driver to use static work_struct and removes flush_scheduled_work() call in the process.
0003-0004 convert the video drivers and 0005-0006 usb.
These patches are available in the following git tree.
git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq.git net-dev-kill-flush_scheduled_work
Note that the branch is on top of the previous set of flush_scheduled_work() removal patches; however, they're independent and this series cleanly applies on top of 2.6.37-rc5 (6313e3c21743cc88bb5bd8aa72948ee1e83937b6).
diffstat follows. Thanks.
drivers/media/video/bt8xx/bttv-driver.c | 9 ++++++++ drivers/media/video/bt8xx/bttv-input.c | 5 ---- drivers/media/video/cx18/cx18-driver.c | 8 +++++++ drivers/media/video/cx231xx/cx231xx-cards.c | 8 +++++++ drivers/media/video/cx23885/cx23885-input.c | 2 - drivers/media/video/cx88/cx88-mpeg.c | 8 +++++++ drivers/media/video/em28xx/em28xx-cards.c | 8 +++++++ drivers/media/video/omap24xxcam.c | 6 ++--- drivers/media/video/saa7134/saa7134-core.c | 11 ++++++++- drivers/media/video/saa7134/saa7134-empress.c | 2 - drivers/misc/ioc4.c | 29 ++++++++------------------ drivers/usb/atm/speedtch.c | 27 ++++++++++++------------ drivers/usb/gadget/u_ether.c | 4 --- drivers/usb/host/ohci-hcd.c | 3 +- drivers/usb/otg/isp1301_omap.c | 2 - drivers/usb/serial/oti6858.c | 5 +--- drivers/video/fb_defio.c | 5 +--- init/main.c | 3 -- 18 files changed, 86 insertions(+), 59 deletions(-)
--
tejun
PATCH 1/6 init: don't call flush_scheduled_work() from do_initcalls()
The call to flush_scheduled_work() in do_initcalls() is there to make sure all works queued to system_wq by initcalls finish before the init sections are dropped.
However, the call doesn't make much sense at this point - there already are multiple different workqueues and different subsystems are free to create and use their own. Ordering requirements are and should be expressed explicitly.
Drop the call to prepare for the deprecation and removal of flush_scheduled_work().
Andrew suggested adding sanity check where the workqueue code checks whether any pending or running work has the work function in the init text section. However, checking this for running works requires the worker to keep track of the current function being executed, and checking only the pending works will miss most cases. As a violation will almost always be caught by the usual page fault mechanism, I don't think it would be worthwhile to make the workqueue code track extra state just for this.
---
This is part of a series to remove flush_scheduled_work() usage to prepare for deprecation of flush_scheduled_work(). Patches in this series are self contained and mostly straight-forward.
Please feel free to take it into the appropriate tree, or just ack it. In the latter case, I'll merge the patch through the workqueue tree during the next merge window.
Thank you.
init/main.c | 3 --- 1 files changed, 0 insertions(+), 3 deletions(-)
init/main.c +0 -3
@@ -775,9 +775,6 @@ static void __init do_initcalls(void)for (fn = __early_initcall_end; fn < __initcall_end; fn++)do_one_initcall (*fn); - -/* Make sure there is no pending stuff from the initcall sequence */ -flush_scheduled_work (); }/*
Signature · 1 line
1.7.1
PATCH 2/6 ioc4: use static work_struct for ioc4_load_modules()
In reply to Tejun HeoThere is no reason to dynamically allocate work_struct for ioc4_load_modules(). It makes the code more complex and makes it impossible to flush the work directly. Use static work ioc4_load_modules_work instead and flush it directly on exit.
This removes the use of flush_scheduled_work() which is being deprecated.
---
This is part of a series to remove flush_scheduled_work() usage to prepare for deprecation of flush_scheduled_work(). Patches in this series are self contained and mostly straight-forward.
Please feel free to take it into the appropriate tree, or just ack it. In the latter case, I'll merge the patch through the workqueue tree during the next merge window.
Thank you.
drivers/misc/ioc4.c | 29 +++++++++-------------------- 1 files changed, 9 insertions(+), 20 deletions(-)
drivers/misc/ioc4.c +9 -20
@@ -273,13 +273,11 @@ ioc4_variant(struct ioc4_driver_data *idd)static void __devinit ioc4_load_modules (struct work_struct *work) { -/* arg just has to be freed */ -request_module ("sgiioc4" ); - -kfree (work); } +static DECLARE_WORK(ioc4_load_modules_work, ioc4_load_modules); +/* Adds a new instance of an IOC4 card */ static int __devinit ioc4_probe (struct pci_dev *pdev,const struct pci_device_id *pci_id)
@@ -396,21 +394,12 @@ ioc4_probe(struct pci_dev *pdev, const struct pci_device_id *pci_id)* PCI device. */ if (idd->idd_variant !=IOC4_VARIANT_PCI_RT ) { -struct work_struct *work; - work =kzalloc (sizeof (struct work_struct ),GFP_KERNEL ); -if (!work) { -printk (KERN_WARNING -"%s: IOC4 unable to allocate memory for " -"load of sub-modules.\n" , __func__); - }else { -/* Request the module from a work procedure as the -* modprobe goes out to a userland helper and that -* will hang if done directly from ioc4_probe(). -*/ -printk (KERN_INFO "IOC4 loading sgiioc4 submodule\n" ); -INIT_WORK (work, ioc4_load_modules); -schedule_work (work); - } +/* Request the module from a work procedure as the modprobe +* goes out to a userland helper and that will hang if done +* directly from ioc4_probe(). +*/ +printk (KERN_INFO "IOC4 loading sgiioc4 submodule\n" ); +schedule_work (&ioc4_load_modules_work); }return 0 ;
@@ -498,7 +487,7 @@ static void __exitioc4_exit (void) {/* Ensure ioc4_load_modules() has completed before exiting */ -flush_scheduled_work (); +flush_work_sync (&ioc4_load_modules_work);pci_unregister_driver (&ioc4_driver); }
Signature · 1 line
1.7.1
PATCH 3/6 media/video: explicitly flush request_module work
In reply to Tejun HeoVideo drivers request submodules using a work during probe and calls flush_scheduled_work() on exit to make sure the work is complete before being unloaded. This patch makes these drivers flush the work directly instead of using flush_scheduled_work().
While at it, relocate request_submodules() call in saa7134_initdev() right right before successful return as in other drivers to avoid failing after the work is scheduled and returning failure without the work still active.
This is in preparation for the deprecation of flush_scheduled_work().
---
This is part of a series to remove flush_scheduled_work() usage to prepare for deprecation of flush_scheduled_work(). Patches in this series are self contained and mostly straight-forward.
Please feel free to take it into the appropriate tree, or just ack it. In the latter case, I'll merge the patch through the workqueue tree during the next merge window.
Thank you.
drivers/media/video/bt8xx/bttv-driver.c | 9 +++++++++ drivers/media/video/cx18/cx18-driver.c | 8 ++++++++ drivers/media/video/cx231xx/cx231xx-cards.c | 8 ++++++++ drivers/media/video/cx88/cx88-mpeg.c | 8 ++++++++ drivers/media/video/em28xx/em28xx-cards.c | 8 ++++++++ drivers/media/video/saa7134/saa7134-core.c | 11 +++++++++-- 6 files changed, 50 insertions(+), 2 deletions(-)
drivers/media/video/bt8xx/bttv-driver.c +9 -0
@@ -189,8 +189,14 @@ static void request_modules(struct bttv *dev)INIT_WORK (&dev->request_module_wk, request_module_async); schedule_work(&dev->request_module_wk); } + +static void flush_request_modules (struct bttv *dev) +{ +flush_work_sync (&dev->request_module_wk); +}#else #define request_modules (dev) +#define flush_request_modules (dev)#endif /* CONFIG_MODULES */
@@ -4573,6 +4579,9 @@ static void __devexit bttv_remove(struct pci_dev *pci_dev)if (bttv_verbose)printk ("bttv%d: unloading\n" ,btv->c.nr); +if (bttv_tvcards[btv->c.type].has_dvb) +flush_request_modules (btv); +/* shutdown everything (DMA+IRQs) */ btand (~15 ,BT848_GPIO_DMA_CTL );btwrite (0 ,BT848_INT_MASK );
drivers/media/video/cx18/cx18-driver.c +8 -0
@@ -266,8 +266,14 @@ static void request_modules(struct cx18 *dev)INIT_WORK (&dev->request_module_wk, request_module_async); schedule_work(&dev->request_module_wk); } + +static void flush_request_modules (struct cx18 *dev) +{ +flush_work_sync (&dev->request_module_wk); +}#else #define request_modules (dev) +#define flush_request_modules (dev)#endif /* CONFIG_MODULES */ /* Generic utility functions */
@@ -1226,6 +1232,8 @@ static void cx18_remove(struct pci_dev *pci_dev)CX18_DEBUG_INFO ("Removing Card\n" ); +flush_request_modules (cx); +/* Stop all captures */ CX18_DEBUG_INFO ("Stopping all streams\n" );if (atomic_read (&cx->tot_capturing) >0 )
drivers/media/video/cx231xx/cx231xx-cards.c +8 -0
@@ -762,8 +762,14 @@ static void request_modules(struct cx231xx *dev)INIT_WORK (&dev->request_module_wk, request_module_async); schedule_work(&dev->request_module_wk); } + +static void flush_request_modules (struct cx231xx *dev) +{ +flush_work_sync (&dev->request_module_wk); +}#else #define request_modules (dev) +#define flush_request_modules (dev)#endif /* CONFIG_MODULES */ /*
@@ -1096,6 +1102,8 @@ static void cx231xx_usb_disconnect(struct usb_interface *interface)if (!dev->udev)return ; +flush_request_modules (dev); +/* delete v4l2 device */ v4l2_device_unregister (&dev->v4l2_dev);
drivers/media/video/cx88/cx88-mpeg.c +8 -0
@@ -66,8 +66,14 @@ static void request_modules(struct cx8802_dev *dev)INIT_WORK (&dev->request_module_wk, request_module_async); schedule_work(&dev->request_module_wk); } + +static void flush_request_modules (struct cx8802_dev *dev) +{ +flush_work_sync (&dev->request_module_wk); +}#else #define request_modules (dev) +#define flush_request_modules (dev)#endif /* CONFIG_MODULES */
@@ -819,6 +825,8 @@ static void __devexit cx8802_remove(struct pci_dev *pci_dev)dprintk (1 ,"%s\n" , __func__); +flush_request_modules (dev); +if (!list_empty (&dev->drvlist)) {struct cx8802_driver *drv, *tmp;int err;
drivers/media/video/em28xx/em28xx-cards.c +8 -0
@@ -2632,8 +2632,14 @@ static void request_modules(struct em28xx *dev)INIT_WORK (&dev->request_module_wk, request_module_async); schedule_work(&dev->request_module_wk); } + +static void flush_request_modules (struct em28xx *dev) +{ +flush_work_sync (&dev->request_module_wk); +}#else #define request_modules (dev) +#define flush_request_modules (dev)#endif /* CONFIG_MODULES */ /*
@@ -3060,6 +3066,8 @@ static void em28xx_usb_disconnect(struct usb_interface *interface)em28xx_info ("disconnecting %s\n" , dev->vdev->name); +flush_request_modules (dev); +/* wait until all current v4l2 io is finished then deallocate resources */ mutex_lock (&dev->lock);
drivers/media/video/saa7134/saa7134-core.c +9 -2
@@ -166,8 +166,14 @@ static void request_submodules(struct saa7134_dev *dev) schedule_work(&dev->request_module_wk); } +static void flush_request_submodules (struct saa7134_dev *dev) +{ +flush_work_sync (&dev->request_module_wk); +} +#else #define request_submodules (dev) +#define flush_request_submodules (dev)#endif /* CONFIG_MODULES */ /* ------------------------------------------------------------------ */
@@ -1010,8 +1016,6 @@ static int __devinit saa7134_initdev(struct pci_dev *pci_dev, } } -request_submodules (dev); -v4l2_prio_init (&dev->prio);mutex_lock (&saa7134_devlist_lock);
@@ -1066,6 +1070,7 @@ static int __devinit saa7134_initdev(struct pci_dev *pci_dev,if (saa7134_dmasound_init && !dev->dmasound.priv_data)saa7134_dmasound_init (dev); +request_submodules (dev);return 0 ; fail4:
@@ -1091,6 +1096,8 @@ static void __devexit saa7134_finidev(struct pci_dev *pci_dev) struct saa7134_dev *dev = container_of(v4l2_dev, struct saa7134_dev, v4l2_dev);struct saa7134_mpeg_ops *mops; +flush_request_submodules (dev); +/* Release DMA sound modules if present */ if (saa7134_dmasound_exit && dev->dmasound.priv_data) {saa7134_dmasound_exit (dev);
Signature · 1 line
1.7.1
PATCH 4/6 media/video: don't use flush_scheduled_work()
In reply to Tejun HeoThis patch converts the remaining users of flush_scheduled_work() in media/video.
* bttv-input.c and cx23885-input.c don't use workqueue at all. No need to flush.
* Make omap24xxcam.c and saa7134-empress.c flush the used work directly.
* In fd_defio.c, replace cancel_delayed_work() + flush_scheduled_work() with cancel_delayed_work_sync(). While at it, replace the deprecated cancel_rearming_delayed_work() with cancel_delayed_work_sync().
---
This is part of a series to remove flush_scheduled_work() usage to prepare for deprecation of flush_scheduled_work(). Patches in this series are self contained and mostly straight-forward.
Please feel free to take it into the appropriate tree, or just ack it. In the latter case, I'll merge the patch through the workqueue tree during the next merge window.
Thank you.
drivers/media/video/bt8xx/bttv-input.c | 5 +---- drivers/media/video/cx23885/cx23885-input.c | 2 -- drivers/media/video/omap24xxcam.c | 6 +++--- drivers/media/video/saa7134/saa7134-empress.c | 2 +- drivers/video/fb_defio.c | 5 ++--- 5 files changed, 7 insertions(+), 13 deletions(-)
drivers/media/video/bt8xx/bttv-input.c +1 -4
@@ -229,16 +229,13 @@ static void bttv_ir_start(struct bttv *btv, struct card_ir *ir)static void bttv_ir_stop (struct bttv *btv) { -if (btv->remote->polling) { +if (btv->remote->polling)del_timer_sync (&btv->remote->timer); -flush_scheduled_work (); - }if (btv->remote->rc5_gpio) {u32 gpio;del_timer_sync (&btv->remote->timer_end); -flush_scheduled_work (); gpio =bttv_gpio_read (&btv->c);bttv_gpio_write (&btv->c, gpio & ~(1 <<4 ));
drivers/media/video/cx23885/cx23885-input.c +0 -2
@@ -230,8 +230,6 @@ static void cx23885_input_ir_stop(struct cx23885_dev *dev)v4l2_subdev_call (dev->sd_ir, ir, rx_s_parameters, ¶ms);v4l2_subdev_call (dev->sd_ir, ir, rx_g_parameters, ¶ms); } - -flush_scheduled_work (); }static void cx23885_input_ir_close (void *priv)
drivers/media/video/omap24xxcam.c +3 -3
@@ -1198,7 +1198,7 @@ static int vidioc_streamoff(struct file *file, void *fh, enum v4l2_buf_type i)atomic_inc (&cam->reset_disable); -flush_scheduled_work (); +flush_work_sync (&cam->sensor_reset_work); rval =videobuf_streamoff (q);if (!rval) {
@@ -1512,7 +1512,7 @@ static int omap24xxcam_release(struct file *file)atomic_inc (&cam->reset_disable); -flush_scheduled_work (); +flush_work_sync (&cam->sensor_reset_work);/* stop streaming capture */ videobuf_streamoff (&fh->vbq);
@@ -1536,7 +1536,7 @@ static int omap24xxcam_release(struct file *file)* not be scheduled anymore since streaming is already * disabled.) */ -flush_scheduled_work (); +flush_work_sync (&cam->sensor_reset_work);mutex_lock (&cam->mutex);if (atomic_dec_return (&cam->users) ==0 ) {
drivers/media/video/saa7134/saa7134-empress.c +1 -1
@@ -553,7 +553,7 @@ static int empress_fini(struct saa7134_dev *dev)if (NULL == dev->empress_dev)return 0 ; -flush_scheduled_work (); +flush_work_sync (&dev->empress_workqueue);video_unregister_device (dev->empress_dev); dev->empress_dev =NULL ;return 0 ;
drivers/video/fb_defio.c +2 -3
@@ -75,7 +75,7 @@ int fb_deferred_io_fsync(struct file *file, int datasync)return 0 ;/* Kill off the delayed work */ -cancel_rearming_delayed_work (&info->deferred_work); +cancel_delayed_work_sync (&info->deferred_work);/* Run it immediately */ return schedule_delayed_work (&info->deferred_work,0 );
@@ -216,8 +216,7 @@ void fb_deferred_io_cleanup(struct fb_info *info)int i;BUG_ON (!fbdefio); -cancel_delayed_work (&info->deferred_work); -flush_scheduled_work (); +cancel_delayed_work_sync (&info->deferred_work);/* clear out the mapping that we setup */ for (i =0 ; i < info->fix.smem_len; i +=PAGE_SIZE ) {
Signature · 1 line
1.7.1
PATCH 5/6 speedtch: don't abuse struct delayed_work
In reply to Tejun Heospeedtch directly uses the internal timer and work members of a struct delayed_work. Use a separate work item and timer instead.
---
This is part of a series to remove flush_scheduled_work() usage to prepare for deprecation of flush_scheduled_work(). Patches in this series are self contained and mostly straight-forward.
Please feel free to take it into the appropriate tree, or just ack it. In the latter case, I'll merge the patch through the workqueue tree during the next merge window.
Thank you.
drivers/usb/atm/speedtch.c | 25 +++++++++++++------------ 1 files changed, 13 insertions(+), 12 deletions(-)
drivers/usb/atm/speedtch.c +13 -12
@@ -139,7 +139,8 @@ struct speedtch_instance_data {struct speedtch_params params;/* set in probe, constant afterwards */ -struct delayed_work status_checker; +struct timer_list status_check_timer; +struct work_struct status_check_work;unsigned char last_status;
@@ -498,7 +499,7 @@ static void speedtch_check_status(struct work_struct *work) {struct speedtch_instance_data *instance = container_of(work, struct speedtch_instance_data, - status_checker.work); + status_check_work);struct usbatm_data *usbatm = instance->usbatm;struct atm_dev *atm_dev = usbatm->atm_dev;unsigned char *buf = instance->scratch_buffer;
@@ -575,11 +576,11 @@ static void speedtch_status_poll(unsigned long data) {struct speedtch_instance_data *instance = (void *)data; -schedule_delayed_work (&instance->status_checker,0 ); +schedule_work (&instance->status_check_work);/* The following check is racy, but the race is harmless */ if (instance->poll_delay <MAX_POLL_DELAY ) -mod_timer (&instance->status_checker.timer, jiffies +msecs_to_jiffies (instance->poll_delay)); +mod_timer (&instance->status_check_timer, jiffies +msecs_to_jiffies (instance->poll_delay));else atm_warn (instance->usbatm,"Too many failures - disabling line status polling\n" ); }
@@ -595,7 +596,7 @@ static void speedtch_resubmit_int(unsigned long data)if (int_urb) { ret =usb_submit_urb (int_urb,GFP_ATOMIC );if (!ret) -schedule_delayed_work (&instance->status_checker,0 ); +schedule_work (&instance->status_check_work);else {atm_dbg (instance->usbatm,"%s: usb_submit_urb failed with result %d\n" , __func__, ret);mod_timer (&instance->resubmit_timer, jiffies +msecs_to_jiffies (RESUBMIT_DELAY ));
@@ -624,7 +625,7 @@ static void speedtch_handle_int(struct urb *int_urb) }if ((count ==6 ) && !memcmp (up_int, instance->int_data,6 )) { -del_timer (&instance->status_checker.timer); +del_timer (&instance->status_check_timer);atm_info (usbatm,"DSL line goes up\n" ); }else if ((count ==6 ) && !memcmp (down_int, instance->int_data,6 )) {atm_info (usbatm,"DSL line goes down\n" );
@@ -640,7 +641,7 @@ static void speedtch_handle_int(struct urb *int_urb)if ((int_urb = instance->int_urb)) { ret =usb_submit_urb (int_urb,GFP_ATOMIC ); -schedule_delayed_work (&instance->status_checker,0 ); +schedule_work (&instance->status_check_work);if (ret <0 ) {atm_dbg (usbatm,"%s: usb_submit_urb failed with result %d\n" , __func__, ret); goto fail;
@@ -686,7 +687,7 @@ static int speedtch_atm_start(struct usbatm_data *usbatm, struct atm_dev *atm_de }/* Start status polling */ -mod_timer (&instance->status_checker.timer, jiffies +msecs_to_jiffies (1000 )); +mod_timer (&instance->status_check_timer, jiffies +msecs_to_jiffies (1000 ));return 0 ; }
@@ -698,7 +699,7 @@ static void speedtch_atm_stop(struct usbatm_data *usbatm, struct atm_dev *atm_deatm_dbg (usbatm,"%s entered\n" , __func__); -del_timer_sync (&instance->status_checker.timer); +del_timer_sync (&instance->status_check_timer);/* * Since resubmit_timer and int_urb can schedule themselves and
@@ -869,10 +870,10 @@ static int speedtch_bind(struct usbatm_data *usbatm, usbatm->flags |= (use_isoc ?UDSL_USE_ISOC :0 ); -INIT_DELAYED_WORK (&instance->status_checker, speedtch_check_status); +INIT_WORK (&instance->status_check_work, speedtch_check_status); - instance->status_checker.timer.function = speedtch_status_poll; - instance->status_checker.timer.data = (unsigned long )instance; + instance->status_check_timer.function = speedtch_status_poll; + instance->status_check_timer.data = (unsigned long )instance; instance->last_status =0xff ; instance->poll_delay =MIN_POLL_DELAY ;
Signature · 1 line
1.7.1
Re: [PATCH 5/6] speedtch: don't abuse struct delayed_work
Quoted 6 lines · Tejun Heo
speedtch directly uses the internal timer and work members of a struct
delayed_work. Use a separate work item and timer instead.
Quoted 8 lines
---
This is part of a series to removeflush_scheduled_work()usage to prepare for deprecation offlush_scheduled_work(). Patches in this series are self contained and mostly straight-forward.Please feel free to take it into the appropriate tree, or just ack it. In the latter case, I'll merge the patch through the workqueue tree during the next merge window.
You forgot to add a "Signed-off-by:" on these patches, so I can't take them in my tree :(
But feel free to add:
to both 5/6 and 6/6 of this series and take them through your tree.
thanks,
greg k-h
Re: [PATCH 5/6] speedtch: don't abuse struct delayed_work
Hello,
Quoted 2 lines · Greg KH
You forgot to add a "Signed-off-by:" on these patches, so I can't take them in my tree :(
Oops, sorry about that. Where did they go? :-)
Quoted 4 lines
But feel free to add:
to both 5/6 and 6/6 of this series and take them through your tree.
Yeap, will do so.
Thanks.
Signature · 1 line
tejun
PATCH 6/6 usb: don't use flush_scheduled_work()
In reply to Tejun Heoflush_scheduled_work() is being deprecated. Directly flush or cancel work items instead.
* u_ether, isp1301_omap, speedtch conversions are straight-forward.
* ochi-hcd should only flush when quirk_nec() is true as otherwise the work wouldn't have been initialized.
* In oti6858, cancel_delayed_work() + flush_scheduled_work() -> cancel_delayed_work_sync().
---
This is part of a series to remove flush_scheduled_work() usage to prepare for deprecation of flush_scheduled_work(). Patches in this series are self contained and mostly straight-forward.
Please feel free to take it into the appropriate tree, or just ack it. In the latter case, I'll merge the patch through the workqueue tree during the next merge window.
Thank you.
drivers/usb/atm/speedtch.c | 2 +- drivers/usb/gadget/u_ether.c | 4 +--- drivers/usb/host/ohci-hcd.c | 3 ++- drivers/usb/otg/isp1301_omap.c | 2 +- drivers/usb/serial/oti6858.c | 5 ++--- 5 files changed, 7 insertions(+), 9 deletions(-)
drivers/usb/atm/speedtch.c +1 -1
@@ -718,7 +718,7 @@ static void speedtch_atm_stop(struct usbatm_data *usbatm, struct atm_dev *atm_dedel_timer_sync (&instance->resubmit_timer);usb_free_urb (int_urb); -flush_scheduled_work (); +flush_work_sync (&instance->status_check_work); }static int speedtch_pre_reset (struct usb_interface *intf)
drivers/usb/gadget/u_ether.c +1 -3
@@ -829,11 +829,9 @@ void gether_cleanup(void)return ;unregister_netdev (the_dev->net); +flush_work_sync (&the_dev->work);free_netdev (the_dev->net); -/* assuming we used keventd, it must quiesce too */ -flush_scheduled_work (); - the_dev =NULL ; }
drivers/usb/host/ohci-hcd.c +2 -1
@@ -901,7 +901,8 @@ static void ohci_stop (struct usb_hcd *hcd)ohci_dump (ohci,1 ); -flush_scheduled_work (); +if (quirk_nec (ohci)) +flush_work_sync (&ohci->nec_work);ohci_usb_reset (ohci);ohci_writel (ohci,OHCI_INTR_MIE , &ohci->regs->intrdisable);
drivers/usb/otg/isp1301_omap.c +1 -1
@@ -1247,7 +1247,7 @@ static int __exit isp1301_remove(struct i2c_client *i2c) isp->timer.data =0 ;set_bit (WORK_STOP , &isp->todo);del_timer_sync (&isp->timer); -flush_scheduled_work (); +flush_work_sync (&isp->work);put_device (&i2c->dev); the_transceiver =NULL ;
drivers/usb/serial/oti6858.c +2 -3
@@ -613,9 +613,8 @@ static void oti6858_close(struct usb_serial_port *port)dbg ("%s(): after buf_clear()" , __func__);/* cancel scheduled setup */ -cancel_delayed_work (&priv->delayed_setup_work); -cancel_delayed_work (&priv->delayed_write_work); -flush_scheduled_work (); +cancel_delayed_work_sync (&priv->delayed_setup_work); +cancel_delayed_work_sync (&priv->delayed_write_work);/* shutdown our urbs */ dbg ("%s(): shutting down urbs" , __func__);
Signature · 1 line
1.7.1
Re: speedtch: don't abuse struct delayed_work
In reply to Tejun HeoQuoted 7 lines · Tejun Heo
Quoted 6 lines
speedtch directly uses the internal timer and work members of a struct
delayed_work. Use a separate work item and timer instead.This is part of a series to remove
flush_scheduled_work()usage to prepare for deprecation offlush_scheduled_work(). Patches in this series are self contained and mostly straight-forward
I tested this patch with 2.6.37-rc6. Not sure whether I'm doing something wrong, but I failed to start the PPP daemon with my speedtouch usb:
Dec 17 14:17:24 absol pppd[4853]: Plugin pppoatm.so loaded. Dec 17 14:17:24 absol pppd[4853]: PPPoATM plugin_init Dec 17 14:17:24 absol pppd[4853]: PPPoATM setdevname_pppoatm - SUCCESS:8.48 Dec 17 14:17:24 absol pppd[4854]: pppd 2.4.4 started by root, uid 0 Dec 17 14:17:24 absol pppd[4854]: connect(8.48): No such device Dec 17 14:17:24 absol pppd[4854]: Exit.
Dec 17 14:17:26 absol rc-scripts: Failed to start the PPP daemon
Without the patch:
Dec 17 14:19:19 absol pppd[4847]: Plugin pppoatm.so loaded. Dec 17 14:19:19 absol pppd[4847]: PPPoATM plugin_init Dec 17 14:19:19 absol pppd[4847]: PPPoATM setdevname_pppoatm - SUCCESS:8.48 Dec 17 14:19:19 absol pppd[4848]: pppd 2.4.4 started by root, uid 0 Dec 17 14:19:19 absol pppd[4848]: using channel 1 Dec 17 14:19:19 absol pppd[4848]: Using interface ppp0 Dec 17 14:19:19 absol pppd[4848]: Connect: ppp0 <--> 8.48
Best regards,
Nicolas Kaiser
Re: speedtch: don't abuse struct delayed_work
* Nicolas Kaiser <[email protected]>:
Quoted 11 lines
Quoted 7 lines · Tejun Heo
Quoted 6 lines
speedtch directly uses the internal timer and work members of a struct
delayed_work. Use a separate work item and timer instead.This is part of a series to remove
flush_scheduled_work()usage to prepare for deprecation offlush_scheduled_work(). Patches in this series are self contained and mostly straight-forwardI tested this patch with 2.6.37-rc6. Not sure whether I'm doing something wrong, but I failed to start the PPP daemon with my speedtouch usb:
I narrowed down my problem, and noticed that speedtch_atm_start never returns, but gets stuck trying to start status polling.
Now when I delete this mod_timer, everything appears to work nicely at my place. With additional debug output:
Dec 18 12:24:40 absol kernel: ATM dev 0: speedtch_atm_start entered Dec 18 12:24:40 absol kernel: ATM dev 0: speedtch_start_synchro entered Dec 18 12:24:40 absol kernel: ATM dev 0: speedtch_start_synchro: modem prodded. 2 bytes returned: 00 00 Dec 18 12:24:40 absol kernel: return from speedtch_start_synchro Dec 18 12:24:40 absol kernel: return from speedtch_atm_start
Dec 18 12:24:49 absol pppd[4844]: Plugin pppoatm.so loaded. Dec 18 12:24:49 absol pppd[4844]: PPPoATM plugin_init Dec 18 12:24:49 absol pppd[4844]: PPPoATM setdevname_pppoatm - SUCCESS:8.48 Dec 18 12:24:49 absol pppd[4845]: pppd 2.4.4 started by root, uid 0 Dec 18 12:24:49 absol pppd[4845]: using channel 1 Dec 18 12:24:49 absol pppd[4845]: Using interface ppp0 Dec 18 12:24:49 absol pppd[4845]: Connect: ppp0 <--> 8.48
Dec 18 12:24:54 absol kernel: ATM dev 0: speedtch_handle_int entered Dec 18 12:24:54 absol kernel: ATM dev 0: DSL line goes up Dec 18 12:24:54 absol kernel: return from speedtch_handle_int Dec 18 12:24:54 absol kernel: entered speedtch_check_status Dec 18 12:24:54 absol kernel: entered speedtch_read_status Dec 18 12:24:54 absol kernel: return from speedtch_read_status Dec 18 12:24:54 absol kernel: ATM dev 0: speedtch_check_status: line state 0x20 Dec 18 12:24:54 absol kernel: ATM dev 0: ADSL line is up (6176 kb/s down | 512 kb/s up) Dec 18 12:24:54 absol kernel: return from speedtch_check_status
Best regards,
Nicolas Kaiser
--- drivers/usb/atm/speedtch.c | 3 --- 1 files changed, 0 insertions(+), 3 deletions(-)
drivers/usb/atm/speedtch.c +0 -3
@@ -686,9 +686,6 @@ static int speedtch_atm_start(struct usbatm_data *usbatm, struct atm_dev *atm_de } } -/* Start status polling */ -mod_timer (&instance->status_check_timer, jiffies +msecs_to_jiffies (1000 )); -return 0 ; }
Signature · 1 line
1.7.2.2
PATCH 5/6 speedtch: don't abuse struct delayed_work
In reply to Nicolas Kaiserspeedtch directly uses the internal timer and work members of a struct delayed_work. Use a separate work item and timer instead.
* Nicolas Kaiser discovered that timer init was missing. Fixed.
---
Can you please test this one? Thanks.
drivers/usb/atm/speedtch.c | 26 ++++++++++++++------------ 1 file changed, 14 insertions(+), 12 deletions(-)
drivers/usb/atm/speedtch.c +14 -12
@@ -139,7 +139,8 @@ struct speedtch_instance_data {struct speedtch_params params;/* set in probe, constant afterwards */ -struct delayed_work status_checker; +struct timer_list status_check_timer; +struct work_struct status_check_work;unsigned char last_status;
@@ -498,7 +499,7 @@ static void speedtch_check_status(struct {struct speedtch_instance_data *instance = container_of(work, struct speedtch_instance_data, - status_checker.work); + status_check_work);struct usbatm_data *usbatm = instance->usbatm;struct atm_dev *atm_dev = usbatm->atm_dev;unsigned char *buf = instance->scratch_buffer;
@@ -575,11 +576,11 @@ static void speedtch_status_poll(unsigne {struct speedtch_instance_data *instance = (void *)data; -schedule_delayed_work (&instance->status_checker,0 ); +schedule_work (&instance->status_check_work);/* The following check is racy, but the race is harmless */ if (instance->poll_delay <MAX_POLL_DELAY ) -mod_timer (&instance->status_checker.timer, jiffies +msecs_to_jiffies (instance->poll_delay)); +mod_timer (&instance->status_check_timer, jiffies +msecs_to_jiffies (instance->poll_delay));else atm_warn (instance->usbatm,"Too many failures - disabling line status polling\n" ); }
@@ -595,7 +596,7 @@ static void speedtch_resubmit_int(unsignif (int_urb) { ret =usb_submit_urb (int_urb,GFP_ATOMIC );if (!ret) -schedule_delayed_work (&instance->status_checker,0 ); +schedule_work (&instance->status_check_work);else {atm_dbg (instance->usbatm,"%s: usb_submit_urb failed with result %d\n" , __func__, ret);mod_timer (&instance->resubmit_timer, jiffies +msecs_to_jiffies (RESUBMIT_DELAY ));
@@ -624,7 +625,7 @@ static void speedtch_handle_int(struct u }if ((count ==6 ) && !memcmp (up_int, instance->int_data,6 )) { -del_timer (&instance->status_checker.timer); +del_timer (&instance->status_check_timer);atm_info (usbatm,"DSL line goes up\n" ); }else if ((count ==6 ) && !memcmp (down_int, instance->int_data,6 )) {atm_info (usbatm,"DSL line goes down\n" );
@@ -640,7 +641,7 @@ static void speedtch_handle_int(struct uif ((int_urb = instance->int_urb)) { ret =usb_submit_urb (int_urb,GFP_ATOMIC ); -schedule_delayed_work (&instance->status_checker,0 ); +schedule_work (&instance->status_check_work);if (ret <0 ) {atm_dbg (usbatm,"%s: usb_submit_urb failed with result %d\n" , __func__, ret); goto fail;
@@ -686,7 +687,7 @@ static int speedtch_atm_start(struct usb }/* Start status polling */ -mod_timer (&instance->status_checker.timer, jiffies +msecs_to_jiffies (1000 )); +mod_timer (&instance->status_check_timer, jiffies +msecs_to_jiffies (1000 ));return 0 ; }
@@ -698,7 +699,7 @@ static void speedtch_atm_stop(struct usbatm_dbg (usbatm,"%s entered\n" , __func__); -del_timer_sync (&instance->status_checker.timer); +del_timer_sync (&instance->status_check_timer);/* * Since resubmit_timer and int_urb can schedule themselves and
@@ -869,10 +870,11 @@ static int speedtch_bind(struct usbatm_d usbatm->flags |= (use_isoc ?UDSL_USE_ISOC :0 ); -INIT_DELAYED_WORK (&instance->status_checker, speedtch_check_status); +INIT_WORK (&instance->status_check_work, speedtch_check_status); +init_timer (&instance->status_check_timer); - instance->status_checker.timer.function = speedtch_status_poll; - instance->status_checker.timer.data = (unsigned long )instance; + instance->status_check_timer.function = speedtch_status_poll; + instance->status_check_timer.data = (unsigned long )instance; instance->last_status =0xff ; instance->poll_delay =MIN_POLL_DELAY ;
Re: [PATCH UPDATED 5/6] speedtch: don't abuse struct delayed_work
* Tejun Heo <[email protected]>:
Quoted 6 lines
speedtch directly uses the internal timer and work members of a struct
delayed_work. Use a separate work item and timer instead.* Nicolas Kaiser discovered that timer init was missing. Fixed.
Can you please test this one? Thanks.
Works for me, thanks!
I'm about to replace this device, and will have to return it to my ISP in about a fortnight. Until then I can perform further tests if needed.
Best regards,
Nicolas Kaiser
Tejun Heo
AppliedQuoted 10 lines · Tejun Heo
Hello,
This is another set of patches to remove
flush_scheduled_work()users.0001-init-don-t-call-flush_scheduled_work-from-do_initcal.patch 0002-ioc4-use-static-work_struct-for-ioc4_load_modules.patch 0003-media-video-explicitly-flush-request_module-work.patch 0004-media-video-don-t-use-flush_scheduled_work.patch 0005-speedtch-don-t-abuse-struct-delayed_work.patch 0006-usb-don-t-use-flush_scheduled_work.patch
Merged into wq#for-next. Thanks.
Signature · 1 line
tejun