Freezer Patches.
Subsystems (27)
Hi Pavel.
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
Regards,
Nigel
drivers/acpi/osl.c +1 -1
@@ -95,7 +95,7 @@ acpi_os_initialize1(void) return AE_NULL_ENTRY; }#endif - kacpid_wq =create_singlethread_workqueue ("kacpid" ); + kacpid_wq =create_singlethread_workqueue ("kacpid" ,PF_NOFREEZE );BUG_ON (!kacpid_wq);return AE_OK ;
drivers/block/ll_rw_blk.c +1 -1
@@ -3266,7 +3266,7 @@ EXPORT_SYMBOL(kblockd_flush);int __init blk_dev_init (void ) { - kblockd_workqueue =create_workqueue ("kblockd" ); + kblockd_workqueue =create_workqueue ("kblockd" ,PF_NOFREEZE );if (!kblockd_workqueue)panic ("Failed to create kblockd\n" );
drivers/block/pktcdvd.c +1 -1
@@ -2366,7 +2366,7 @@ static int pkt_new_dev(struct pktcdvd_depkt_init_queue (pd);atomic_set (&pd->cdrw.pending_bios,0 ); - pd->cdrw.thread =kthread_run (kcdrwd, pd,"%s" , pd->name); + pd->cdrw.thread =kthread_run (kcdrwd, pd,0 ,"%s" , pd->name);if (IS_ERR (pd->cdrw.thread)) {printk ("pktcdvd: can't start kernel thread\n" ); ret = -ENOMEM ;
drivers/char/hvc_console.c +1 -1
@@ -709,7 +709,7 @@ int __init hvc_init(void)/* Always start the kthread because there can be hotplug vty adapters * added later. */ - hvc_task =kthread_run (khvcd,NULL ,"khvcd" ); + hvc_task =kthread_run (khvcd,NULL ,PF_NOFREEZE ,"khvcd" );if (IS_ERR (hvc_task)) {panic ("Couldn't create kthread for console.\n" );put_tty_driver (hvc_driver);
drivers/char/hvcs.c +1 -1
@@ -1403,7 +1403,7 @@ static int __init hvcs_module_init(void)return -ENOMEM ; } - hvcs_task =kthread_run (khvcsd,NULL ,"khvcsd" ); + hvcs_task =kthread_run (khvcsd,NULL ,PF_NOFREEZE ,"khvcsd" );if (IS_ERR (hvcs_task)) {printk (KERN_ERR "HVCS: khvcsd creation failed. Driver not loaded.\n" );kfree (hvcs_pi_buff);
drivers/infiniband/core/fmr_pool.c +1 -0
@@ -266,6 +266,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(s pool->thread =kthread_create (ib_fmr_cleanup_thread, pool, +0 ,"ib_fmr(%s)" , device->name);if (IS_ERR (pool->thread)) {
drivers/infiniband/core/mad.c +1 -1
@@ -2502,7 +2502,7 @@ static int ib_mad_port_open(struct ib_de goto error7;snprintf (name,sizeof name,"ib_mad%d" , port_num); - port_priv->wq =create_singlethread_workqueue (name); + port_priv->wq =create_singlethread_workqueue (name,0 );if (!port_priv->wq) { ret = -ENOMEM ; goto error8;
drivers/infiniband/ulp/ipoib/ipoib_main.c +1 -1
@@ -1070,7 +1070,7 @@ static int __init ipoib_init_module(void* so flush_scheduled_work() can deadlock during device * removal. */ - ipoib_workqueue =create_singlethread_workqueue ("ipoib" ); + ipoib_workqueue =create_singlethread_workqueue ("ipoib" ,0 );if (!ipoib_workqueue) { ret = -ENOMEM ; goto err_fs;
drivers/macintosh/therm_adt746x.c +1 -1
@@ -438,7 +438,7 @@ static int attach_one_thermostat(structwrite_both_fan_speed (th,-1 ); } - thread_therm =kthread_run (monitor_task, th,"kfand" ); + thread_therm =kthread_run (monitor_task, th,0 ,"kfand" );if (thread_therm ==ERR_PTR (-ENOMEM )) {printk (KERN_INFO "adt746x: Kthread creation failed\n" );
drivers/md/dm-crypt.c +1 -1
@@ -936,7 +936,7 @@ static int __init dm_crypt_init(void)if (!_crypt_io_pool)return -ENOMEM ; - _kcryptd_workqueue =create_workqueue ("kcryptd" ); + _kcryptd_workqueue =create_workqueue ("kcryptd" ,PF_NOFREEZE );if (!_kcryptd_workqueue) { r = -ENOMEM ;DMERR (PFX "couldn't create kcryptd" );
drivers/md/dm-raid1.c +1 -1
@@ -1229,7 +1229,7 @@ static int __init dm_mirror_init(void)if (r)return r; - _kmirrord_wq =create_workqueue ("kmirrord" ); + _kmirrord_wq =create_workqueue ("kmirrord" ,PF_SYNCTHREAD );if (!_kmirrord_wq) {DMERR ("couldn't start kmirrord" );dm_dirty_log_exit ();
drivers/md/kcopyd.c +1 -1
@@ -609,7 +609,7 @@ static int kcopyd_init(void)return r; } - _kcopyd_wq =create_singlethread_workqueue ("kcopyd" ); + _kcopyd_wq =create_singlethread_workqueue ("kcopyd" ,PF_SYNCTHREAD );if (!_kcopyd_wq) {jobs_exit ();up (&kcopyd_init_lock);
drivers/media/video/cx88/cx88-video.c +1 -1
@@ -2121,7 +2121,7 @@ static int __devinit cx8800_initdev(stru/* start tvaudio thread */ if (core->tuner_type !=TUNER_ABSENT ) - core->kthread =kthread_run (cx88_audio_thread, core,"cx88 tvaudio" ); + core->kthread =kthread_run (cx88_audio_thread, core,0 ,"cx88 tvaudio" );return 0 ; fail_unreg:
drivers/media/video/msp3400.c +1 -1
@@ -1555,7 +1555,7 @@ static int msp_attach(struct i2c_adapter/* startup control thread if needed */ if (thread_func) { - msp->kthread =kthread_run (thread_func, c,"msp34xx" ); + msp->kthread =kthread_run (thread_func, c,0 ,"msp34xx" );if (NULL == msp->kthread)printk (KERN_WARNING "msp34xx: kernel_thread() failed\n" );msp_wake_thread (c);
drivers/media/video/video-buf-dvb.c +1 -1
@@ -104,7 +104,7 @@ static int videobuf_dvb_start_feed(strucif (NULL != dvb->thread) goto out; dvb->thread =kthread_run (videobuf_dvb_thread, - dvb,"%s dvb" , dvb->name); + dvb,0 ,"%s dvb" , dvb->name);if (IS_ERR (dvb->thread)) { rc =PTR_ERR (dvb->thread); dvb->thread =NULL ;
drivers/message/i2o/driver.c +1 -1
@@ -79,7 +79,7 @@ int i2o_driver_register(struct i2o_drivepr_debug ("i2o: Register driver %s\n" , drv->name);if (drv->event) { - drv->event_queue =create_workqueue (drv->name); + drv->event_queue =create_workqueue (drv->name,0 );if (!drv->event_queue) {printk (KERN_ERR "i2o: Could not initialize event queue " "for driver %s\n" , drv->name);
drivers/net/wan/sdlamain.c +1 -1
@@ -240,7 +240,7 @@ static int __init wanpipe_init(void)printk (KERN_INFO "%s v%u.%u %s\n" , fullname,DRV_VERSION ,DRV_RELEASE , copyright); - wanpipe_wq =create_workqueue ("wanpipe_wq" ); + wanpipe_wq =create_workqueue ("wanpipe_wq" ,0 );if (!wanpipe_wq)return -ENOMEM ;
drivers/s390/cio/device.c +4 -3
@@ -148,15 +148,16 @@ init_ccw_bus_type (void)init_waitqueue_head (&ccw_device_init_wq);atomic_set (&ccw_device_init_count,0 ); - ccw_device_work =create_singlethread_workqueue ("cio" ); + ccw_device_work =create_singlethread_workqueue ("cio" ,0 );if (!ccw_device_work)return -ENOMEM ;/* FIXME: better errno ? */ - ccw_device_notify_work =create_singlethread_workqueue ("cio_notify" ); + ccw_device_notify_work =create_singlethread_workqueue ("cio_notify" , +0 );if (!ccw_device_notify_work) { ret = -ENOMEM ;/* FIXME: better errno ? */ goto out_err; } - slow_path_wq =create_singlethread_workqueue ("kslowcrw" ); + slow_path_wq =create_singlethread_workqueue ("kslowcrw" ,0 );if (!slow_path_wq) { ret = -ENOMEM ;/* FIXME: better errno ? */ goto out_err;
drivers/scsi/hosts.c +1 -1
@@ -133,7 +133,7 @@ int scsi_add_host(struct Scsi_Host *shossnprintf (shost->work_q_name,KOBJ_NAME_LEN ,"scsi_wq_%d" , shost->host_no); shost->work_q =create_singlethread_workqueue ( - shost->work_q_name); + shost->work_q_name,PF_NOFREEZE );if (!shost->work_q) goto out_free_shost_data; }
drivers/scsi/libata-core.c +1 -1
@@ -3951,7 +3951,7 @@ int pci_test_config_bits(struct pci_devstatic int __init ata_init (void ) { - ata_wq =create_workqueue ("ata" ); + ata_wq =create_workqueue ("ata" ,0 );if (!ata_wq)return -ENOMEM ;
drivers/usb/net/pegasus.c +1 -1
@@ -1412,7 +1412,7 @@ static struct usb_driver pegasus_driverstatic int __init pegasus_init (void ) {pr_info ("%s: %s, " DRIVER_DESC "\n" , driver_name,DRIVER_VERSION ); - pegasus_workqueue =create_singlethread_workqueue ("pegasus" ); + pegasus_workqueue =create_singlethread_workqueue ("pegasus" ,PF_NOFREEZE );if (!pegasus_workqueue)return -ENOMEM ;return usb_register (&pegasus_driver);
fs/aio.c +1 -1
@@ -73,7 +73,7 @@ static int __init aio_setup(void) kioctx_cachep =kmem_cache_create ("kioctx" ,sizeof (struct kioctx ),0 ,SLAB_HWCACHE_ALIGN |SLAB_PANIC ,NULL ,NULL ); - aio_wq =create_workqueue ("aio" ); + aio_wq =create_workqueue ("aio" ,0 );pr_debug ("aio_setup: sizeof(struct page) = %d\n" , (int )sizeof (struct page ));
fs/nfsd/nfs4state.c +1 -1
@@ -1356,7 +1356,7 @@ void nfsd_break_deleg_cb(struct file_loc/* XXX need to merge NFSD_LEASE_TIME with fs/locks.c:lease_break_time */ fl->fl_break_time = jiffies +NFSD_LEASE_TIME *HZ ; - t =kthread_run (do_recall, dp,"%s" ,"nfs4_cb_recall" ); + t =kthread_run (do_recall, dp,0 ,"%s" ,"nfs4_cb_recall" );if (IS_ERR (t)) {struct nfs4_client *clp = dp->dl_client;
fs/reiserfs/journal.c +1 -1
@@ -2512,7 +2512,7 @@ int journal_init(struct super_block *p_s reiserfs_mounted_fs_count++ ;if (reiserfs_mounted_fs_count <=1 ) - commit_wq =create_workqueue ("reiserfs" ); + commit_wq =create_workqueue ("reiserfs" ,PF_SYNCTHREAD );INIT_WORK (&journal->j_work, flush_async_commits, p_s_sb);return 0 ;
fs/xfs/linux-2.6/xfs_buf.c +2 -2
@@ -1895,11 +1895,11 @@ pagebuf_daemon_start(void) {int rval; - pagebuf_logio_workqueue =create_workqueue ("xfslogd" ); + pagebuf_logio_workqueue =create_workqueue ("xfslogd" ,PF_SYNCTHREAD );if (!pagebuf_logio_workqueue)return -ENOMEM ; - pagebuf_dataio_workqueue =create_workqueue ("xfsdatad" ); + pagebuf_dataio_workqueue =create_workqueue ("xfsdatad" ,PF_SYNCTHREAD );if (!pagebuf_dataio_workqueue) {destroy_workqueue (pagebuf_logio_workqueue);return -ENOMEM ;
include/linux/kthread.h +8 -2
@@ -25,20 +25,26 @@*/ struct task_struct *kthread_create (int (*threadfn)(void *data),void *data, +unsigned long freezer_flags,const char namefmt[], ...);/** * kthread_run: create and wake a thread. * @threadfn: the function to run until signal_pending(current). * @data: data ptr for @threadfn. +* @freezer_flags: process flags that should be used for freezing. +* PF_SYNCTHREAD if needed for syncing data to disk. +* PF_NOFREEZE if also needed for writing the image. +* 0 otherwise. * @namefmt: printf-style name for the thread. * * Description: Convenient wrapper for kthread_create() followed by * wake_up_process(). Returns the kthread, or ERR_PTR(-ENOMEM). */ -#define kthread_run (threadfn, data, namefmt, ...) \ +#define kthread_run (threadfn, data, freezer_flags, namefmt, ...) \ ({ \ struct task_struct *__k \ - = kthread_create(threadfn, data, namefmt, ## __VA_ARGS__); \ + = kthread_create(threadfn, data, freezer_flags, \ + namefmt, ## __VA_ARGS__); \ if (!IS_ERR(__k)) \ wake_up_process(__k); \ __k; \
include/linux/workqueue.h +4 -3
@@ -51,9 +51,10 @@ struct work_struct { }while (0 ) extern struct workqueue_struct *__create_workqueue(const char *name, -int singlethread); -#define create_workqueue (name) __create_workqueue((name), 0) -#define create_singlethread_workqueue (name) __create_workqueue((name), 1) +int singlethread, +unsigned long freezer_flag); +#define create_workqueue (name, flags) __create_workqueue((name), 0, flags) +#define create_singlethread_workqueue (name, flags) __create_workqueue((name), 1, flags)extern void destroy_workqueue (struct workqueue_struct *wq);
kernel/kmod.c +1 -1
@@ -251,6 +251,6 @@ EXPORT_SYMBOL(call_usermodehelper);void __init usermodehelper_init (void ) { - khelper_wq =create_singlethread_workqueue ("khelper" ); + khelper_wq =create_singlethread_workqueue ("khelper" ,0 );BUG_ON (!khelper_wq); }
kernel/kthread.c +8 -1
@@ -25,6 +25,7 @@ struct kthread_create_info/* Information passed to kthread() from keventd. */ int (*threadfn)(void *data);void *data; +unsigned long freezer_flags;struct completion started;/* Result passed back to kthread_create() from keventd. */
@@ -86,6 +87,10 @@ static int kthread(void *_create)/* By default we can run anywhere, unlike keventd. */ set_cpus_allowed (current,CPU_MASK_ALL ); +/* Set our freezer flags */ + current->flags &= ~(PF_SYNCTHREAD |PF_NOFREEZE ); + current->flags |= create->freezer_flags; +/* OK, tell user we're spawned, wait for stop or wakeup */ __set_current_state (TASK_INTERRUPTIBLE );complete (&create->started);
@@ -121,6 +126,7 @@ static void keventd_create_kthread(voidstruct task_struct *kthread_create (int (*threadfn)(void *data),void *data, +unsigned long freezer_flags,const char namefmt[], ...) {
@@ -129,6 +135,7 @@ struct task_struct *kthread_create(int ( create.threadfn = threadfn; create.data = data; + create.freezer_flags = freezer_flags;init_completion (&create.started);init_completion (&create.done);
@@ -193,7 +200,7 @@ EXPORT_SYMBOL(kthread_stop);static __init int helper_init (void ) { - helper_wq =create_singlethread_workqueue ("kthread" ); + helper_wq =create_singlethread_workqueue ("kthread" ,0 );BUG_ON (!helper_wq);return 0 ;
kernel/sched.c +2 -2
@@ -4379,10 +4379,10 @@ static int migration_call(struct notifieswitch (action) {case CPU_UP_PREPARE : - p =kthread_create (migration_thread, hcpu,"migration/%d" ,cpu); + p =kthread_create (migration_thread, hcpu,0 , +"migration/%d" ,cpu);if (IS_ERR (p))return NOTIFY_BAD ; - p->flags |=PF_NOFREEZE ;kthread_bind (p, cpu);/* Must be high prio: stop_machine expects to yield to it. */ rq =task_rq_lock (p, &flags);
kernel/softirq.c +3 -2
@@ -350,7 +350,6 @@ void __init softirq_init(void)static int ksoftirqd (void * __bind_cpu) {set_user_nice (current,19 ); - current->flags |=PF_NOFREEZE ;set_current_state (TASK_INTERRUPTIBLE );
@@ -364,6 +363,8 @@ static int ksoftirqd(void * __bind_cpu)__set_current_state (TASK_RUNNING ); +try_to_freeze (PF_FREEZE ); +while (local_softirq_pending ()) {/* Preempt disable stops cpu going offline. If already offline, we'll be on wrong CPU:
@@ -456,7 +457,7 @@ static int __devinit cpu_callback(structcase CPU_UP_PREPARE :BUG_ON (per_cpu (tasklet_vec, hotcpu).list);BUG_ON (per_cpu (tasklet_hi_vec, hotcpu).list); - p =kthread_create (ksoftirqd, hcpu,"ksoftirqd/%d" , hotcpu); + p =kthread_create (ksoftirqd, hcpu,PF_NOFREEZE ,"ksoftirqd/%d" , hotcpu);if (IS_ERR (p)) {printk ("ksoftirqd for %i failed\n" , hotcpu);return NOTIFY_BAD ;
kernel/stop_machine.c +1 -1
@@ -184,7 +184,7 @@ struct task_struct *__stop_machine_run(iif (cpu ==NR_CPUS ) cpu =_smp_processor_id (); - p =kthread_create (do_stop, &smdata,"kstopmachine" ); + p =kthread_create (do_stop, &smdata,0 ,"kstopmachine" );if (!IS_ERR (p)) {kthread_bind (p, cpu);wake_up_process (p);
kernel/workqueue.c +12 -9
@@ -186,8 +186,6 @@ static int worker_thread(void *__cwq)struct k_sigaction sa;sigset_t blocked; - current->flags |=PF_NOFREEZE ; -set_user_nice (current,-5 );/* Block and flush all signals */
@@ -208,6 +206,7 @@ static int worker_thread(void *__cwq)schedule (); else__set_current_state (TASK_RUNNING ); +try_to_freeze (PF_FREEZE );remove_wait_queue (&cwq->more_work, &wait);if (!list_empty (&cwq->worklist))
@@ -277,7 +276,8 @@ void fastcall flush_workqueue(struct wor }static struct task_struct *create_workqueue_thread (struct workqueue_struct *wq, -int cpu) +int cpu, +unsigned long freezer_flags) {struct cpu_workqueue_struct *cwq = wq->cpu_wq + cpu;struct task_struct *p;
@@ -292,9 +292,11 @@ static struct task_struct *create_workquinit_waitqueue_head (&cwq->work_done);if (is_single_threaded (wq)) - p =kthread_create (worker_thread, cwq,"%s" , wq->name); + p =kthread_create (worker_thread, cwq, freezer_flags, +"%s" , wq->name);else - p =kthread_create (worker_thread, cwq,"%s/%d" , wq->name, cpu); + p =kthread_create (worker_thread, cwq, freezer_flags, +"%s/%d" , wq->name, cpu);if (IS_ERR (p))return NULL ; cwq->thread = p;
@@ -302,7 +304,8 @@ static struct task_struct *create_workqu }struct workqueue_struct *__create_workqueue (const char *name, -int singlethread) +int singlethread, +unsigned long freezer_flags) {int cpu, destroy =0 ;struct workqueue_struct *wq;
@@ -320,7 +323,7 @@ struct workqueue_struct *__create_workqulock_cpu_hotplug ();if (singlethread) {INIT_LIST_HEAD (&wq->list); - p =create_workqueue_thread (wq,0 ); + p =create_workqueue_thread (wq,0 , freezer_flags);if (!p) destroy =1 ;else
@@ -330,7 +333,7 @@ struct workqueue_struct *__create_workqulist_add (&wq->list, &workqueues);spin_unlock (&workqueue_lock);for_each_online_cpu (cpu) { - p =create_workqueue_thread (wq, cpu); + p =create_workqueue_thread (wq, cpu, freezer_flags);if (p) {kthread_bind (p, cpu);wake_up_process (p);
@@ -540,7 +543,7 @@ static int __devinit workqueue_cpu_callbvoid init_workqueues (void ) {hotcpu_notifier (workqueue_cpu_callback,0 ); - keventd_wq =create_workqueue ("events" ); + keventd_wq =create_workqueue ("events" ,PF_NOFREEZE );BUG_ON (!keventd_wq); }
mm/pdflush.c +1 -1
@@ -213,7 +213,7 @@ int pdflush_operation(void (*fn)(unsignestatic void start_one_pdflush_thread (void ) { -kthread_run (pdflush,NULL ,"pdflush" ); +kthread_run (pdflush,NULL ,PF_SYNCTHREAD ,"pdflush" ); }static int __init pdflush_init (void )
net/sunrpc/sched.c +1 -1
@@ -1003,7 +1003,7 @@ rpciod_up(void)* Create the rpciod thread and wait for it to start. */ error = -ENOMEM ; - wq =create_workqueue ("rpciod" ); + wq =create_workqueue ("rpciod" ,0 );if (wq ==NULL ) {printk (KERN_WARNING "rpciod_up: create workqueue failed, error=%d\n" , error); rpciod_users--;
include/linux/sched.h +2 -0
@@ -1248,6 +1248,7 @@ extern void normalize_rt_tasks(void);extern void refrigerator (unsigned long );extern int freeze_processes (void );extern void thaw_processes (void ); +extern unsigned int refrigerator_on;static inline int try_to_freeze (unsigned long refrigerator_flags) {
@@ -1261,6 +1262,7 @@ static inline int try_to_freeze(unsignedstatic inline void refrigerator (unsigned long flag) {}static inline int freeze_processes (void ) {BUG ();return 0 ; }static inline void thaw_processes (void ) {} +#define refrigerator_on (0)static inline int try_to_freeze (unsigned long refrigerator_flags) {
kernel/exit.c +2 -0
@@ -789,6 +789,8 @@ fastcall NORET_TYPE void do_exit(long copanic ("Attempted to kill init!" );if (tsk->io_context)exit_io_context (); +if (unlikely (refrigerator_on)) +refrigerator (0 );if (unlikely (current->ptrace &PT_TRACE_EXIT )) { current->ptrace_message = code;
kernel/fork.c +3 -0
@@ -1200,6 +1200,9 @@ long do_fork(unsigned long clone_flags,int trace =0 ;long pid =alloc_pidmap (); +if (unlikely (refrigerator_on)) +refrigerator (0 ); +if (pid <0 )return -EAGAIN ;if (unlikely (current->ptrace)) {
kernel/kmod.c +1 -1
@@ -237,7 +237,7 @@ int call_usermodehelper(char *path, char };DECLARE_WORK (work, __call_usermodehelper, &sub_info); -if (!khelper_wq) +if ((!khelper_wq) || (refrigerator_on))return -EBUSY ;if (path[0 ] =='\0' )
kernel/power/process.c +266 -70
387 lines not on this page · raw message
fs/buffer.c +30 -2
@@ -171,6 +171,10 @@ EXPORT_SYMBOL(sync_blockdev);*/ int fsync_super (struct super_block *sb) { +int ret; + + current->flags |=PF_SYNCTHREAD ; +sync_inodes_sb (sb,0 );DQUOT_SYNC (sb);lock_super (sb);
@@ -182,7 +186,10 @@ int fsync_super(struct super_block *sb)sync_blockdev (sb->s_bdev);sync_inodes_sb (sb,1 ); -return sync_blockdev (sb->s_bdev); + ret =sync_blockdev (sb->s_bdev); + + current->flags &= ~PF_SYNCTHREAD ; +return ret; }/*
@@ -193,12 +200,19 @@ int fsync_super(struct super_block *sb)int fsync_bdev (struct block_device *bdev) {struct super_block *sb =get_super (bdev); +int ret; + + current->flags |=PF_SYNCTHREAD ; +if (sb) {int res =fsync_super (sb);drop_super (sb); + current->flags &= ~PF_SYNCTHREAD ;return res; } -return sync_blockdev (bdev); + ret =sync_blockdev (bdev); + current->flags &= ~PF_SYNCTHREAD ; +return ret; }/**
@@ -278,6 +292,8 @@ EXPORT_SYMBOL(thaw_bdev);*/ static void do_sync (unsigned long wait) { + current->flags |=PF_SYNCTHREAD ; +wakeup_bdflush (0 );sync_inodes (0 );/* All mappings, inodes and their blockdevs */ DQUOT_SYNC (NULL );
@@ -289,6 +305,8 @@ static void do_sync(unsigned long wait)printk ("Emergency Sync complete\n" );if (unlikely (laptop_mode))laptop_sync_completion (); + + current->flags &= ~PF_SYNCTHREAD ; }asmlinkage long sys_sync (void )
@@ -314,6 +332,8 @@ int file_fsync(struct file *filp, structstruct super_block * sb;int ret, err; + current->flags |=PF_SYNCTHREAD ; +/* sync the inode to buffers */ ret =write_inode_now (inode,0 );
@@ -328,6 +348,8 @@ int file_fsync(struct file *filp, struct err =sync_blockdev (sb->s_bdev);if (!ret) ret = err; + + current->flags &= ~PF_SYNCTHREAD ;return ret; }
@@ -337,6 +359,8 @@ asmlinkage long sys_fsync(unsigned int fstruct address_space *mapping;int ret, err; + current->flags |=PF_SYNCTHREAD ; + ret = -EBADF ; file =fget (fd);if (!file)
@@ -370,6 +394,7 @@ asmlinkage long sys_fsync(unsigned int f out_putf:fput (file); out: + current->flags &= ~PF_SYNCTHREAD ;return ret; }
@@ -379,6 +404,8 @@ asmlinkage long sys_fdatasync(unsigned istruct address_space *mapping;int ret, err; + current->flags |=PF_SYNCTHREAD ; + ret = -EBADF ; file =fget (fd);if (!file)
@@ -405,6 +432,7 @@ asmlinkage long sys_fdatasync(unsigned i out_putf:fput (file); out: + current->flags &= ~PF_SYNCTHREAD ;return ret; }
fs/jbd/journal.c +1 -0
@@ -130,6 +130,7 @@ int kjournald(void *arg) current_journal = journal;daemonize ("kjournald" ); + current->flags |=PF_SYNCTHREAD ;/* Set up an interval timer which can be used to trigger a commit wakeup after the commit interval expires */
fs/jffs/intrep.c +4 -0
@@ -3364,6 +3364,7 @@ jffs_garbage_collect_thread(void *ptr) D1(int i = 1);daemonize ("jffs_gcd" ); + current->flags |=PF_SYNCTHREAD ; c->gc_task = current;
@@ -3399,6 +3400,9 @@ jffs_garbage_collect_thread(void *ptr)siginfo_t info;unsigned long signr =0 ; +if (try_to_freeze (PF_FREEZE )) +continue ; +spin_lock_irq (¤t->sighand->siglock); signr =dequeue_signal (current, ¤t->blocked, &info);spin_unlock_irq (¤t->sighand->siglock);
fs/jfs/jfs_logmgr.c +1 -0
@@ -2321,6 +2321,7 @@ int jfsIOWait(void *arg)struct lbuf *bp;daemonize ("jfsIO" ); + current->flags |=PF_SYNCTHREAD ;complete (&jfsIOwait);
fs/jfs/jfs_txnmgr.c +1 -1
@@ -47,7 +47,6 @@#include <linux/vmalloc.h> #include <linux/smp_lock.h> #include <linux/completion.h> -#include <linux/suspend.h> #include <linux/module.h> #include <linux/moduleparam.h> #include "jfs_incore.h"
@@ -2750,6 +2749,7 @@ int jfs_lazycommit(void *arg)struct jfs_sb_info *sbi;daemonize ("jfsCommit" ); + current->flags |=PF_SYNCTHREAD ;complete (&jfsIOwait);
fs/lockd/clntlock.c +2 -0
@@ -200,6 +200,7 @@ reclaimer(void *ptr)struct inode *inode;daemonize ("%s-reclaim" , host->h_name); + current->flags |=PF_SYNCTHREAD ;allow_signal (SIGKILL );/* This one ensures that our parent doesn't terminate while the
@@ -222,6 +223,7 @@ restart: fl->fl_u.nfs_fl.flags &= ~NFS_LCK_RECLAIM ;nlmclnt_reclaim (host, fl); +try_to_freeze (PF_FREEZE );if (signalled ())break ; goto restart;
fs/lockd/svc.c +3 -0
@@ -115,6 +115,7 @@ lockd(struct svc_rqst *rqstp)up (&lockd_start);daemonize ("lockd" ); + current->flags |=PF_SYNCTHREAD ;/* Process request with signals blocked, but allow SIGKILL. */ allow_signal (SIGKILL );
@@ -138,6 +139,8 @@ lockd(struct svc_rqst *rqstp)while ((nlmsvc_users || !signalled ()) && nlmsvc_pid == current->pid) {long timeout =MAX_SCHEDULE_TIMEOUT ; +try_to_freeze (PF_SYNCTHREAD ); +if (signalled ()) {flush_signals (current);if (nlmsvc_ops) {
fs/nfsd/nfssvc.c +1 -0
@@ -182,6 +182,7 @@ nfsd(struct svc_rqst *rqstp)/* Lock module and set up kernel thread */ lock_kernel ();daemonize ("nfsd" ); + current->flags |=PF_SYNCTHREAD ;/* After daemonize() this kernel thread shares current->fs * with the init process. We need to create files with a
fs/xfs/linux-2.6/xfs_buf.c +1 -1
@@ -1770,7 +1770,7 @@ pagebuf_daemon(/* Set up the thread */ daemonize ("xfsbufd" ); - current->flags |=PF_MEMALLOC ; + current->flags |=PF_MEMALLOC |PF_SYNCTHREAD ; pagebuf_daemon_task = current; pagebuf_daemon_active =1 ;
fs/xfs/linux-2.6/xfs_super.c +1 -0
@@ -471,6 +471,7 @@ xfssyncd(struct vfs_sync_work *work, *n;daemonize ("xfssyncd" ); + current->flags |=PF_SYNCTHREAD ; vfsp->vfs_sync_work.w_vfs = vfsp; vfsp->vfs_sync_work.w_syncer = vfs_sync_worker;
include/linux/sched.h +2 -0
@@ -788,6 +788,8 @@ do { if (atomic_dec_and_test(&(tsk)->usa#define PF_SYNCWRITE 0x00200000/* I am doing a sync write */ #define PF_BORROWED_MM 0x00400000/* I am a kthread doing use_mm */ #define PF_RANDOMIZE 0x00800000/* randomize virtual address space */ +#define PF_SYNCTHREAD 0x01000000/* this thread can start activity during the ++ early part of freezing processes */ /* * Only the _current_ task can read/write to tsk->flags, but other
mm/pdflush.c +1 -1
@@ -89,7 +89,7 @@ struct pdflush_work {static int __pdflush (struct pdflush_work *my_work) { - current->flags |=PF_FLUSHER ; + current->flags |= (PF_FLUSHER |PF_SYNCTHREAD ); my_work->fn =NULL ; my_work->who = current;INIT_LIST_HEAD (&my_work->list);
Hi!
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
304 seems ugly and completely useless for mainline 300: stopping softirqd seems dangerous to me... are you sure? 301: patching exit should not be neccessary. Why do you need it?
Pavel
Signature · 1 line
64 bytes from 195.113.31.123: icmp_seq=28 ttl=51 time=448769.1 ms
Quoted 7 lines · Pavel Machek
Hi!
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
300: stopping softirqd seems dangerous to me... are you sure?
That sounds bogus indeed.
Hi.
Quoted 10 lines · Benjamin Herrenschmidt
Quoted 7 lines · Pavel Machek
Hi!
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
300: stopping softirqd seems dangerous to me... are you sure?
That sounds bogus indeed.
Ok. Bogus but harmless. I'll try removing it.
Regards,
Nigel
Quoted 10 lines · Nigel Cunningham
Hi.
Quoted 7 lines · Benjamin Herrenschmidt
Quoted 5 lines · Pavel Machek
Hi!
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
Quoted 5 lines
Quoted 3 lines
Quoted 1 line
300: stopping softirqd seems dangerous to me... are you sure?
That sounds bogus indeed.
Ok. Bogus but harmless. I'll try removing it.
Not really harmless, you happen to be lucky :)
Ben.
Morning.
Quoted 7 lines · Pavel Machek
Hi!
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
304 seems ugly and completely useless for mainline
That's because you don't understand what it's doing.
The new refrigerator implementation works like this:
Userspace processes that begin a sys_*sync gain the process flag PF_SYNCTHREAD for the duration of their syscall.
We then freeze processes in three distinct groups.
1. Userspace processes (p->mm) that don't have PF_SYNCTHREAD set are signalled first. This causes them to enter the refrigerator, thereby stopping them from submitting further I/O that could cause the syncing threads to potentially spin for ever.
2. Syncing threads are signalled so that when they finish syncing (if they haven't already), they also enter the refrigerator. Any processes that were in group 2 but are now in group 1 are also signalled.
It is possible that no process was syncing data. We thus now call sys_sync ourselves, to flush remaining dirty data to disk.
3. Kernel threads not needed during suspend (kjournald, eg) are frozen.
Quoted 1 line
300: stopping softirqd seems dangerous to me... are you sure?
It's been that way for ages without any reports of deadlocks. I'm happy to change it if it really was dangerous for some reason.
Quoted 1 line
301: patching exit should not be neccessary. Why do you need it?
Maybe you're right. I think it's a left over from the old implementation. I'll try removing the call and see how the freezer survives some stress testing.
Regards,
Nigel
Hi!
(Well, it is just after midnight here :-).
Quoted 4 lines
Quoted 4 lines
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
Quoted 8 lines
Quoted 1 line
304 seems ugly and completely useless for mainline
That's because you don't understand what it's doing.
The new refrigerator implementation works like this:
Userspace processes that begin a sys_*sync gain the process flag
PF_SYNCTHREADfor the duration of their syscall.
swsusp1 should not need any special casing of sync, right? We can simply do sys_sync(), then freeze, or something like that. We could even remove sys_sync() completely; it is not needed for correctness.
Pavel
Quoted 12 lines · Pavel Machek
Hi!
(Well, it is just after midnight here :-).
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
304 seems ugly and completely useless for mainline
That's because you don't understand what it's doing.
Quoted 8 lines
Quoted 4 lines
The new refrigerator implementation works like this:
Userspace processes that begin a sys_*sync gain the process flag
PF_SYNCTHREADfor the duration of their syscall.swsusp1 should not need any special casing of sync, right? We can simply do
sys_sync(), then freeze, or something like that. We could even removesys_sync()completely; it is not needed for correctness.
It's still quite nice to have ... I put it in my pre-freeze callback in fact for both STR and STD :) We really want it for STD but I think it doesn't work properly after freeze.
Ben.
Quoted 18 lines · Benjamin Herrenschmidt
Quoted 17 lines · Pavel Machek
Hi!
(Well, it is just after midnight here :-).
Quoted 12 lines
Quoted 5 lines
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
304 seems ugly and completely useless for mainline
That's because you don't understand what it's doing.
The new refrigerator implementation works like this:
Userspace processes that begin a sys_*sync gain the process flag
PF_SYNCTHREADfor the duration of their syscall.
Quoted 7 lines
Quoted 3 lines
swsusp1 should not need any special casing of sync, right? We can simply do
sys_sync(), then freeze, or something like that. We could even removesys_sync()completely; it is not needed for correctness.It's still quite nice to have ... I put it in my pre-freeze callback in fact for both STR and STD :) We really want it for STD but I think it doesn't work properly after freeze.
I agree that sync() is nice to have, but I'm not going to slow down fork/exit for it. And besides, sys_sync() just before suspend works just fine.
Pavel
Quoted 4 lines
I agree that
sync()is nice to have, but I'm not going to slow down fork/exit for it. And besides,sys_sync()just before suspend works just fine.
Yes, that's why I put it in my pre-freeze :)
I'll see about making my pre/post freeze stuff (APM emu + sync basically now) generic to avoid the callbacks.
Ben.
Hi.
Quoted 8 lines · Benjamin Herrenschmidt
Quoted 3 lines
I agree that
sync()is nice to have, but I'm not going to slow down fork/exit for it. And besides,sys_sync()just before suspend works just fine.Yes, that's why I put it in my pre-freeze :)
I'll see about making my pre/post freeze stuff (APM emu + sync basically now) generic to avoid the callbacks.
Ummm. Don't do it if kthreads are already frozen. You'll deadlock against kjournald.
Regards,
Nigel
Quoted 7 lines · Nigel Cunningham
Hi.
Quoted 4 lines · Benjamin Herrenschmidt
Quoted 3 lines
I agree that
sync()is nice to have, but I'm not going to slow down fork/exit for it. And besides,sys_sync()just before suspend works just fine.
Quoted 6 lines
Quoted 4 lines
Yes, that's why I put it in my pre-freeze :)
I'll see about making my pre/post freeze stuff (APM emu + sync basically now) generic to avoid the callbacks.
Ummm. Don't do it if kthreads are already frozen. You'll deadlock against kjournald.
Which is why I do it pre-freeze.
With a 2 stage freezing though, it would make sense to first freeze userland, then sync, then freeze kthreads. That would make the sync faster (as it won't be "blocked" by pending userland IOs as I experience regulary here) and more reliable.
Ben.
Hi.
Quoted 14 lines · Pavel Machek
Quoted 13 lines · Benjamin Herrenschmidt
Quoted 12 lines · Pavel Machek
Hi!
(Well, it is just after midnight here :-).
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Here are the freezer patches. They were prepared against rc3, but I think they still apply fine against rc5. (Ben, these are the same ones I sent you the other day).
304 seems ugly and completely useless for mainline
That's because you don't understand what it's doing.
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
The new refrigerator implementation works like this:
Userspace processes that begin a sys_*sync gain the process flag
PF_SYNCTHREADfor the duration of their syscall.swsusp1 should not need any special casing of sync, right? We can simply do
sys_sync(), then freeze, or something like that. We could even removesys_sync()completely; it is not needed for correctness.
Wrong. I guess you're only trying it on a machine that isn't actually doing anything :). I've forgotten whether it was this freezer implementation or the last, but we've been testing freezing processes when the load average exceeds 100. If you have a thread that is syncing and another that's submitting I/O continually (think dd, for example), you want this.
Quoted 7 lines
Quoted 3 lines
It's still quite nice to have ... I put it in my pre-freeze callback in fact for both STR and STD :) We really want it for STD but I think it doesn't work properly after freeze.
I agree that
sync()is nice to have, but I'm not going to slow down fork/exit for it. And besides,sys_sync()just before suspend works just fine.
Again, try it under load... and stop talking about fork and exit like they're real hot paths. (Unless you regularly submit your machines to fork bombs :)).
Regards,
Nigel
Quoted 4 lines
Again, try it under load... and stop talking about fork and exit like they're real hot paths. (Unless you regularly submit your machines to fork bombs :)).
They are real hot path
Ben.
Hi!
Quoted 10 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
swsusp1 should not need any special casing of sync, right? We can simply do
sys_sync(), then freeze, or something like that. We could even removesys_sync()completely; it is not needed for correctness.Wrong. I guess you're only trying it on a machine that isn't actually doing anything :). I've forgotten whether it was this freezer implementation or the last, but we've been testing freezing processes when the load average exceeds 100. If you have a thread that is syncing and another that's submitting I/O continually (think dd, for example), you want this.
If sys_sync() is not working, *fix sys_sync()*. [BTW I see that problem before and I think it is being worked on.] I'm *not* going to work around it in refrigerator.
Pavel
Hi.
Quoted 13 lines · Pavel Machek
Hi!
Quoted 10 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
swsusp1 should not need any special casing of sync, right? We can simply do
sys_sync(), then freeze, or something like that. We could even removesys_sync()completely; it is not needed for correctness.Wrong. I guess you're only trying it on a machine that isn't actually doing anything :). I've forgotten whether it was this freezer implementation or the last, but we've been testing freezing processes when the load average exceeds 100. If you have a thread that is syncing and another that's submitting I/O continually (think dd, for example), you want this.
Quoted 3 lines
If
sys_sync()is not working, *fix sys_sync()*. [BTW I see that problem before and I think it is being worked on.] I'm *not* going to work around it in refrigerator.
I'm not saying sys_sync is broken. I _am_ saying that if you have processes submitting I/O while you're trying to sync, syncing will take longer and you may well still end up with dirty buffers at the end. On top of this, you may think freezing has failed because processes don't enter the refrigerator within your timelimit (assuming you have one).
The simple, logic solution is to stop threads that are submitting I/O before you stop threads that are syncing I/O.
Regards,
Nigel
Hi!
Quoted 10 lines
Quoted 3 lines
If
sys_sync()is not working, *fix sys_sync()*. [BTW I see that problem before and I think it is being worked on.] I'm *not* going to work around it in refrigerator.I'm not saying
sys_syncis broken. I _am_ saying that if you have processes submitting I/O while you're trying to sync, syncing will take longer and you may well still end up with dirty buffers at the end. On top of this, you may think freezing has failed because processes don't enter the refrigerator within your timelimit (assuming you have one).
Then simple launch sys_sync(), let it finish, *then* do refrigeration. That way sys_sync() does not count to the timelimit.
Now, sys_sync() takes too long on some setups. That needs to be fixed, anyway; users don't like to wait for 15 minutes after typing "sync". Do not work around it in refrigerator.
Pavel
Quoted 13 lines · Pavel Machek
Hi!
Quoted 10 lines
Quoted 3 lines
If
sys_sync()is not working, *fix sys_sync()*. [BTW I see that problem before and I think it is being worked on.] I'm *not* going to work around it in refrigerator.I'm not saying
sys_syncis broken. I _am_ saying that if you have processes submitting I/O while you're trying to sync, syncing will take longer and you may well still end up with dirty buffers at the end. On top of this, you may think freezing has failed because processes don't enter the refrigerator within your timelimit (assuming you have one).
Quoted 6 lines
Then simple launch
sys_sync(), let it finish, *then* do refrigeration. That waysys_sync()does not count to the timelimit.Now,
sys_sync()takes too long on some setups. That needs to be fixed, anyway; users don't like to wait for 15 minutes after typing "sync". Do not work around it in refrigerator.
Whatever you guys decide to do (I actually do sys_sync() before freezing on pmac and yes, it takes sometimes way too long), to be uber-safe, we could/should _also_ do sync after freezing userland processes and before freezing kernel threads (that is, splitting here). In fact, that would help also avoid deadlocks where a frozen kernel thread is holding a semaphore preventing a process from freezing.
That way, if we sys_sync() once processes are sleeping and before kernel threads are, we pretty-much make sure no new dirty buffer will appear.
Anyway, that's mostly food for thoughts at this point
Ben.
Hi Ben.
Quoted 11 lines · Benjamin Herrenschmidt
Whatever you guys decide to do (I actually do
sys_sync()before freezing on pmac and yes, it takes sometimes way too long), to be uber-safe, we could/should _also_ do sync after freezing userland processes and before freezing kernel threads (that is, splitting here). In fact, that would help also avoid deadlocks where a frozen kernel thread is holding a semaphore preventing a process from freezing.That way, if we
sys_sync()once processes are sleeping and before kernel threads are, we pretty-much make sure no new dirty buffer will appear.Anyway, that's mostly food for thoughts at this point
I fully agree, and that's what I'm already doing. Nothing I can do about Pavel not seeing logic, so guess I just have to hope everyone else does :)
Regards,
Nigel
Hi!
Quoted 7 lines
Quoted 7 lines
Whatever you guys decide to do (I actually do
sys_sync()before freezing on pmac and yes, it takes sometimes way too long), to be uber-safe, we could/should _also_ do sync after freezing userland processes and before freezing kernel threads (that is, splitting here). In fact, that would help also avoid deadlocks where a frozen kernel thread is holding a semaphore preventing a process from freezing.
Quoted 8 lines
Quoted 4 lines
That way, if we
sys_sync()once processes are sleeping and before kernel threads are, we pretty-much make sure no new dirty buffer will appear.Anyway, that's mostly food for thoughts at this point
I fully agree, and that's what I'm already doing. Nothing I can do about Pavel not seeing logic, so guess I just have to hope everyone else does :)
I think I see your logic, and it is wrong. Anyway, just do not touch fork/exit hot paths.
Pavel