Freezer Upgrade Patches
Subsystems (39)
cover Freezer Upgrade Patches
Hi everyone.
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).
- The use of bdev freezing to ensure filesystems are properly frozen, thereby increasing the integrity of on-disk data in the case where a resume doesn't occur. This is also helpful in the case of Suspend2, where we don't atomically copy all memory, instead writing LRU pages separately.
--
Nigel Cunningham nigel at suspend2 dot net http://suspend2.net
PATCH 1/23 Make workqueues freezeable.
Prior to this patch, kernel threads and workqueues are unconditionally unfreezeable. This patch reverses that behaviour, making the default for kernel processes to be frozen. New variations of the routines for starting kernel threads and workqueues (containing _nofreeze_) allow threads that need to run during suspend to be made nofreeze again.
block/ll_rw_blk.c | 2 - drivers/acpi/osl.c | 2 - drivers/char/hvc_console.c | 2 - drivers/char/hvcs.c | 2 - drivers/input/serio/serio.c | 2 - drivers/md/dm-crypt.c | 2 - drivers/scsi/hosts.c | 2 - drivers/scsi/lpfc/lpfc_init.c | 2 - drivers/usb/net/pegasus.c | 2 - include/linux/kthread.h | 27 ++++++++---- include/linux/workqueue.h | 9 +++- kernel/audit.c | 3 + kernel/kthread.c | 94 ++++++++++++++++++++++++++++++++++++++--- kernel/sched.c | 1 kernel/softirq.c | 3 - kernel/workqueue.c | 37 +++++++++++----- 16 files changed, 149 insertions(+), 43 deletions(-)
block/ll_rw_blk.c +1 -1
@@ -3435,7 +3435,7 @@ int __init blk_dev_init(void) {int i; - kblockd_workqueue =create_workqueue ("kblockd" ); + kblockd_workqueue =create_nofreeze_workqueue ("kblockd" );if (!kblockd_workqueue)panic ("Failed to create kblockd\n" );
drivers/acpi/osl.c +1 -1
@@ -91,7 +91,7 @@ acpi_status acpi_os_initialize1(void) "Access to PCI configuration space unavailable\n");return AE_NULL_ENTRY ; } - kacpid_wq =create_singlethread_workqueue ("kacpid" ); + kacpid_wq =create_nofreeze_singlethread_workqueue ("kacpid" );BUG_ON (!kacpid_wq);return AE_OK ;
drivers/char/hvc_console.c +1 -1
@@ -839,7 +839,7 @@ 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_nofreeze_run (khvcd,NULL ,"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
@@ -1404,7 +1404,7 @@ static int __init hvcs_module_init(void)return -ENOMEM ; } - hvcs_task =kthread_run (khvcsd,NULL ,"khvcsd" ); + hvcs_task =kthread_nofreeze_run (khvcsd,NULL ,"khvcsd" );if (IS_ERR (hvcs_task)) {printk (KERN_ERR "HVCS: khvcsd creation failed. Driver not loaded.\n" );kfree (hvcs_pi_buff);
drivers/input/serio/serio.c +1 -1
@@ -901,7 +901,7 @@ irqreturn_t serio_interrupt(struct seriostatic int __init serio_init (void ) { - serio_task =kthread_run (serio_thread,NULL ,"kseriod" ); + serio_task =kthread_nofreeze_run (serio_thread,NULL ,"kseriod" );if (IS_ERR (serio_task)) {printk (KERN_ERR "serio: Failed to start kseriod\n" );return PTR_ERR (serio_task);
drivers/md/dm-crypt.c +1 -1
@@ -928,7 +928,7 @@ static int __init dm_crypt_init(void)if (!_crypt_io_pool)return -ENOMEM ; - _kcryptd_workqueue =create_workqueue ("kcryptd" ); + _kcryptd_workqueue =create_nofreeze_workqueue ("kcryptd" );if (!_kcryptd_workqueue) { r = -ENOMEM ;DMERR (PFX "couldn't create kcryptd" );
drivers/scsi/hosts.c +1 -1
@@ -227,7 +227,7 @@ int scsi_add_host(struct Scsi_Host *shosif (shost->transportt->create_work_queue) {snprintf (shost->work_q_name,KOBJ_NAME_LEN ,"scsi_wq_%d" , shost->host_no); - shost->work_q =create_singlethread_workqueue ( + shost->work_q =create_nofreeze_singlethread_workqueue ( shost->work_q_name);if (!shost->work_q) goto out_free_shost_data;
drivers/scsi/lpfc/lpfc_init.c +1 -1
@@ -1521,7 +1521,7 @@ lpfc_pci_probe_one(struct pci_dev *pdev, phba->work_ha_mask |= (HA_RXMASK << (LPFC_ELS_RING *4 ));/* Startup the kernel thread for this host adapter. */ - phba->worker_thread =kthread_run (lpfc_do_work, phba, + phba->worker_thread =kthread_nofreeze_run (lpfc_do_work, phba,"lpfc_worker_%d" , phba->brd_no);if (IS_ERR (phba->worker_thread)) { error =PTR_ERR (phba->worker_thread);
drivers/usb/net/pegasus.c +1 -1
@@ -1451,7 +1451,7 @@ static int __init pegasus_init(void)pr_info ("%s: %s, " DRIVER_DESC "\n" , driver_name,DRIVER_VERSION );if (devid)parse_id (devid); - pegasus_workqueue =create_singlethread_workqueue ("pegasus" ); + pegasus_workqueue =create_nofreeze_singlethread_workqueue ("pegasus" );if (!pegasus_workqueue)return -ENOMEM ;return usb_register (&pegasus_driver);
include/linux/kthread.h +19 -8
@@ -23,10 +23,20 @@* * Returns a task_struct or ERR_PTR(-ENOMEM). */ +struct task_struct *__kthread_create (int (*threadfn)(void *data), +void *data, +unsigned long freezer_flags, +const char namefmt[], +va_list * args); +struct task_struct *kthread_create (int (*threadfn)(void *data),void *data,const char namefmt[], ...); +struct task_struct *kthread_nofreeze_create (int (*threadfn)(void *data), +void *data, +const char namefmt[], ...); +/** * kthread_run: create and wake a thread. * @threadfn: the function to run until signal_pending(current).
@@ -35,14 +45,15 @@ struct task_struct *kthread_create(int (* * Description: Convenient wrapper for kthread_create() followed by * wake_up_process(). Returns the kthread, or ERR_PTR(-ENOMEM). */ -#define kthread_run (threadfn, data, namefmt, ...) \ -({ \ - struct task_struct *__k \ - = kthread_create(threadfn, data, namefmt, ## __VA_ARGS__); \ - if (!IS_ERR(__k)) \ - wake_up_process(__k); \ - __k; \ -}) + +extern struct task_struct *kthread_run (int (*threadfn)(void *data), +void *data, +const char namefmt[], ...); + +extern struct task_struct *kthread_nofreeze_run (int (*threadfn)(void *data), +void *data, +const char namefmt[], ...); +/** * kthread_bind: bind a just-created kthread to a cpu.
include/linux/workqueue.h +6 -3
@@ -51,9 +51,12 @@ 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) __create_workqueue((name), 0, 0) +#define create_nofreeze_workqueue (name) __create_workqueue((name), 0, PF_NOFREEZE) +#define create_singlethread_workqueue (name) __create_workqueue((name), 1, 0) +#define create_nofreeze_singlethread_workqueue (name) __create_workqueue((name), 1, PF_NOFREEZE)extern void destroy_workqueue (struct workqueue_struct *wq);
kernel/audit.c +3 -0
@@ -288,6 +288,9 @@ static int kauditd_thread(void *dummy) } }else {DECLARE_WAITQUEUE (wait, current); + +try_to_freeze (); +set_current_state (TASK_INTERRUPTIBLE );add_wait_queue (&kauditd_wait, &wait);
kernel/kthread.c +86 -8
132 lines not on this page · raw message
kernel/sched.c +0 -1
@@ -4805,7 +4805,6 @@ static int migration_call(struct notifie p =kthread_create (migration_thread, hcpu,"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 +1 -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 );
@@ -456,7 +455,7 @@ 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_nofreeze_create (ksoftirqd, hcpu,"ksoftirqd/%d" , hotcpu);if (IS_ERR (p)) {printk ("ksoftirqd for %i failed\n" , hotcpu);return NOTIFY_BAD ;
kernel/workqueue.c +25 -12
@@ -191,8 +191,6 @@ 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 */
@@ -213,6 +211,7 @@ static int worker_thread(void *__cwq)schedule (); else__set_current_state (TASK_RUNNING ); +try_to_freeze ();remove_wait_queue (&cwq->more_work, &wait);if (!list_empty (&cwq->worklist))
@@ -282,7 +281,8 @@ 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 =per_cpu_ptr (wq->cpu_wq, cpu);struct task_struct *p;
@@ -296,10 +296,21 @@ static struct task_struct *create_workquinit_waitqueue_head (&cwq->more_work);init_waitqueue_head (&cwq->work_done); -if (is_single_threaded (wq)) - p =kthread_create (worker_thread, cwq,"%s" , wq->name); -else - p =kthread_create (worker_thread, cwq,"%s/%d" , wq->name, cpu); +if (is_single_threaded (wq)) { +if (freezer_flags) + p =kthread_nofreeze_create (worker_thread, cwq, +"%s" , wq->name); +else + p =kthread_create (worker_thread, cwq, +"%s" , wq->name); + }else { +if (freezer_flags) + p =kthread_nofreeze_create (worker_thread, cwq, +"%s/%d" , wq->name, cpu); +else + p =kthread_create (worker_thread, cwq, +"%s/%d" , wq->name, cpu); + }if (IS_ERR (p))return NULL ; cwq->thread = p;
@@ -307,7 +318,8 @@ 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;
@@ -328,7 +340,8 @@ struct workqueue_struct *__create_workqulock_cpu_hotplug ();if (singlethread) {INIT_LIST_HEAD (&wq->list); - p =create_workqueue_thread (wq, singlethread_cpu); + p =create_workqueue_thread (wq, singlethread_cpu, + freezer_flags);if (!p) destroy =1 ;else
@@ -338,7 +351,7 @@ 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);
@@ -529,7 +542,7 @@ static int __devinit workqueue_cpu_callbcase CPU_UP_PREPARE :/* Create a new workqueue thread for it. */ list_for_each_entry (wq, &workqueues, list) { -if (!create_workqueue_thread (wq, hotcpu)) { +if (!create_workqueue_thread (wq, hotcpu,0 )) {printk ("workqueue for %i failed\n" , hotcpu);return NOTIFY_BAD ; }
@@ -572,7 +585,7 @@ void init_workqueues(void) { singlethread_cpu =first_cpu (cpu_possible_map);hotcpu_notifier (workqueue_cpu_callback,0 ); - keventd_wq =create_workqueue ("events" ); + keventd_wq =create_nofreeze_workqueue ("events" );BUG_ON (!keventd_wq); }
--
Nigel Cunningham nigel at suspend2 dot net
Re: [ 01/23] [Suspend2] Make workqueues freezeable.
Hi,
Quoted 6 lines · Nigel Cunningham
Prior to this patch, kernel threads and workqueues are unconditionally unfreezeable. This patch reverses that behaviour, making the default for kernel processes to be frozen. New variations of the routines for starting kernel threads and workqueues (containing _nofreeze_) allow threads that need to run during suspend to be made nofreeze again.
This looks like "let's make everything freezable and hunt for things that must not be frozen" kind of approach, but isn't it error-prone? I mean, for example, if someone creates a kernel thread that in fact must not be frozen, but forgets to use the _nofreeze_ call, things will break for some people and the problem will be worse than the current one, it seems.
Greetings,
Rafael
Re: [ 01/23] [Suspend2] Make workqueues freezeable.
Hi Rafael.
Quoted 9 lines · Rafael J. Wysocki
Hi,
Quoted 5 lines · Nigel Cunningham
Prior to this patch, kernel threads and workqueues are unconditionally unfreezeable. This patch reverses that behaviour, making the default for kernel processes to be frozen. New variations of the routines for starting kernel threads and workqueues (containing _nofreeze_) allow threads that need to run during suspend to be made nofreeze again.
Quoted 6 lines
This looks like "let's make everything freezable and hunt for things that must not be frozen" kind of approach, but isn't it error-prone? I mean, for example, if someone creates a kernel thread that in fact must not be frozen, but forgets to use the _nofreeze_ call, things will break for some people and the problem will be worse than the current one, it seems.
You're right that the default is now to freeze threads. This is mostly because the majority of kernel threads can and should be frozen because they're not needed. If a thread is frozen which shouldn't be, the user (which will hopefully be the developer testing their work) will experience a deadlock when they try to suspend. If, on the other hand, we let threads run that should be frozen, well - the outcome depends on what is done when the thread should be frozen. Obviously it doesn't hurt too much to let kernel threads run, or we'd see problems at the moment. For this reason, I consider this patch more to be a nicety than a necessity. Nice because it makes kernel threads behave in a manner consistent with what happens to userspace (hence defaulting to freezing, rather than not).
Regards,
Nigel
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
PATCH 2/23 Add new include/linux/freezer.h header.
In reply to Nigel CunninghamThis patch implements a new freezer header file, separated from suspend.h and sched.h so that 95 percent of the kernel doesn't need to be recompiled if/when we modify the freezer implementation.
include/linux/freezer.h | 28 ++++++++++++++++++++++++++++ 1 files changed, 28 insertions(+), 0 deletions(-)
include/linux/freezer.h new file +28 -0
@@ -0,0 +1,28 @@ +/* Freezer declarations */ + +#define FREEZER_ON 0 +#define ABORT_FREEZING 1 + +#define FREEZER_KERNEL_THREADS 0 +#define FREEZER_ALL_THREADS 1 + +#ifdef CONFIG_PM +extern unsigned long freezer_state; + +#define test_freezer_state (bit) test_bit(bit, &freezer_state) +#define set_freezer_state (bit) set_bit(bit, &freezer_state) +#define clear_freezer_state (bit) clear_bit(bit, &freezer_state) + +#define freezer_is_on () (test_freezer_state(FREEZER_ON)) + +extern void do_freeze_process (struct notifier_block *nl); + +#else + +#define test_freezer_state (bit) (0) +#define set_freezer_state (bit) do { } while(0) +#define clear_freezer_state (bit) do { } while(0) + +#define freezer_is_on () (0) + +#endif
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 3/23 Allow a notifier to remove itself from the notifier list.
In reply to Nigel CunninghamThis is Christoph Lameter's original patch, posted to LKML some time ago, but as yet unmerged.
Allow a notifier to remove itself from the notifier list.
If the next pointer is retrieved before notifier_call then the notifier can remove itself from the chain.
kernel/sys.c | 7 +++++-- 1 files changed, 5 insertions(+), 2 deletions(-)
kernel/sys.c +5 -2
@@ -174,15 +174,18 @@ int __kprobes notifier_call_chain(struct {int ret=NOTIFY_DONE ;struct notifier_block *nb = *n; +struct notifier_block *next;while (nb) { - ret=nb->notifier_call (nb,val,v); +/* Determining next here allows the notifier to unregister itself */ + next = nb->next; + ret = nb->notifier_call (nb,val,v);if (ret&NOTIFY_STOP_MASK ) {return ret; } - nb=nb->next; + nb = next; }return ret; }
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 4/23 Todo notifier for processes.
In reply to Nigel Cunningham(This is Christoph Lameter's patch, previously posted to LKML. This revision only updates it to apply against the current codebase).
Introduce a todo notifier in the task_struct so that a task can be told to do certain things. Abuse the suspend hooks try_to_freeze, freezing and refrigerator to establish checkpoints where the todo list is processed.
This will break software suspend (next patch fixes and cleans up software suspend).
This patch has been updated so as to apply to current git. Since that's the only change, I've kept Christoph's original Signed-off-by.
arch/arm/kernel/signal.c | 2 - arch/frv/kernel/signal.c | 2 - arch/h8300/kernel/signal.c | 2 - arch/i386/kernel/io_apic.c | 2 - arch/i386/kernel/signal.c | 2 - arch/m32r/kernel/signal.c | 2 - arch/mips/kernel/irixsig.c | 2 - arch/mips/kernel/signal32.c | 2 - arch/powerpc/kernel/signal_32.c | 2 - arch/sh/kernel/signal.c | 2 - arch/sh64/kernel/signal.c | 2 - arch/x86_64/kernel/signal.c | 2 - drivers/block/pktcdvd.c | 3 - drivers/ieee1394/nodemgr.c | 2 - drivers/input/gameport/gameport.c | 2 - drivers/input/serio/serio.c | 8 +++ drivers/macintosh/therm_adt746x.c | 2 - drivers/media/dvb/dvb-core/dvb_frontend.c | 2 - drivers/media/video/msp3400-driver.c | 2 - drivers/media/video/tvaudio.c | 2 - drivers/media/video/video-buf-dvb.c | 2 - drivers/net/irda/stir4200.c | 4 + drivers/net/wireless/airo.c | 2 - drivers/pcmcia/cs.c | 2 - drivers/pnp/pnpbios/core.c | 2 - drivers/usb/core/hub.c | 2 - drivers/usb/gadget/file_storage.c | 2 - drivers/usb/storage/usb.c | 2 - drivers/w1/w1.c | 4 + fs/afs/kafsasyncd.c | 2 - fs/afs/kafstimod.c | 2 - fs/jbd/journal.c | 4 + fs/jffs/intrep.c | 2 - fs/jffs2/background.c | 2 - fs/jfs/jfs_logmgr.c | 4 + fs/jfs/jfs_txnmgr.c | 8 +-- fs/lockd/clntlock.c | 1 fs/lockd/clntproc.c | 2 - fs/lockd/svc.c | 2 + fs/xfs/linux-2.6/xfs_buf.c | 4 + fs/xfs/linux-2.6/xfs_super.c | 2 - include/linux/sched.h | 80 ++++++++--------------------- kernel/audit.c | 2 - kernel/sched.c | 2 - kernel/signal.c | 4 + kernel/workqueue.c | 2 - mm/page_alloc.c | 1 mm/pdflush.c | 2 - mm/vmscan.c | 2 - net/rxrpc/krxiod.c | 2 - net/rxrpc/krxsecd.c | 2 - net/rxrpc/krxtimod.c | 2 - net/sunrpc/sched.c | 4 + net/sunrpc/svcsock.c | 4 + 54 files changed, 94 insertions(+), 119 deletions(-)
arch/arm/kernel/signal.c +1 -1
@@ -644,7 +644,7 @@ static int do_signal(sigset_t *oldset, sif (!user_mode (regs))return 0 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (current->ptrace &PT_SINGLESTEP )
arch/frv/kernel/signal.c +1 -1
@@ -500,7 +500,7 @@ static void do_signal(void)if (!user_mode (__frame))return ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (test_thread_flag (TIF_RESTORE_SIGMASK ))
arch/h8300/kernel/signal.c +1 -1
@@ -516,7 +516,7 @@ asmlinkage int do_signal(struct pt_regsif ((regs->ccr &0x10 ))return 1 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal; current->thread.esp0 = (unsigned long ) regs;
arch/i386/kernel/io_apic.c +1 -1
@@ -578,7 +578,7 @@ static int balanced_irq(void *unused)for ( ; ; ) { time_remaining =schedule_timeout_interruptible (time_remaining); -try_to_freeze (); +try_todo_list ();if (time_after (jiffies, prev_balance_time+balanced_irq_interval)) {preempt_disable ();
arch/i386/kernel/signal.c +1 -1
@@ -582,7 +582,7 @@ static void fastcall do_signal(struct ptif (!user_mode (regs))return ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (test_thread_flag (TIF_RESTORE_SIGMASK ))
arch/m32r/kernel/signal.c +1 -1
@@ -370,7 +370,7 @@ int do_signal(struct pt_regs *regs, sigsif (!user_mode (regs))return 1 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (!oldset)
arch/mips/kernel/irixsig.c +1 -1
@@ -185,7 +185,7 @@ asmlinkage int do_irix_signal(sigset_t *if (!user_mode (regs))return 1 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (!oldset)
arch/mips/kernel/signal32.c +1 -1
@@ -822,7 +822,7 @@ int do_signal32(sigset_t *oldset, structif (!user_mode (regs))return 1 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (!oldset)
arch/powerpc/kernel/signal_32.c +1 -1
@@ -1132,7 +1132,7 @@ int do_signal(sigset_t *oldset, struct pint signr, ret;#ifdef CONFIG_PPC32 -if (try_to_freeze ()) { +if (try_todo_list ()) { signr =0 ;if (!signal_pending (current)) goto no_signal;
arch/sh/kernel/signal.c +1 -1
@@ -578,7 +578,7 @@ int do_signal(struct pt_regs *regs, sigsif (!user_mode (regs))return 1 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (!oldset)
arch/sh64/kernel/signal.c +1 -1
@@ -696,7 +696,7 @@ int do_signal(struct pt_regs *regs, sigsif (!user_mode (regs))return 1 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (!oldset)
arch/x86_64/kernel/signal.c +1 -1
@@ -443,7 +443,7 @@ int do_signal(struct pt_regs *regs, sigsif (!user_mode (regs))return 1 ; -if (try_to_freeze ()) +if (try_todo_list ()) goto no_signal;if (!oldset)
drivers/block/pktcdvd.c +1 -2
@@ -1255,8 +1255,7 @@ static int kcdrwd(void *foobar) residue =schedule_timeout (min_sleep_time);VPRINTK ("kcdrwd: wake up\n" ); -/* make swsusp happy with our thread */ -try_to_freeze (); +try_todo_list ();list_for_each_entry (pkt, &pd->cdrw.pkt_active_list, list) {if (!pkt->sleep_time)
drivers/ieee1394/nodemgr.c +1 -1
@@ -1581,7 +1581,7 @@ static int nodemgr_host_thread(void *__hif (down_interruptible (&hi->reset_sem) ||down_interruptible (&nodemgr_serialize)) { -if (try_to_freeze ()) +if (try_todo_list ())continue ;printk ("NodeMgr: received unexpected signal?!\n" );break ;
drivers/input/gameport/gameport.c +1 -1
@@ -440,7 +440,7 @@ static int gameport_thread(void *nothinggameport_handle_event ();wait_event_interruptible (gameport_wait,kthread_should_stop () || !list_empty (&gameport_event_list)); -try_to_freeze (); +try_todo_list (); }while (!kthread_should_stop ());printk (KERN_DEBUG "gameport: kgameportd exiting\n" );
drivers/input/serio/serio.c +7 -1
@@ -312,6 +312,12 @@ static void serio_handle_event(void)serio_remove_duplicate_events (event);serio_free_event (event); + +if (unlikely (todo_list_active ())) { +up (&serio_sem); +try_todo_list (); +down (&serio_sem); + } }up (&serio_sem);
@@ -375,7 +381,7 @@ static int serio_thread(void *nothing)serio_handle_event ();wait_event_interruptible (serio_wait,kthread_should_stop () || !list_empty (&serio_event_list)); -try_to_freeze (); +try_todo_list (); }while (!kthread_should_stop ());printk (KERN_DEBUG "serio: kseriod exiting\n" );
drivers/macintosh/therm_adt746x.c +1 -1
@@ -337,7 +337,7 @@ static int monitor_task(void *arg)struct thermostat * th = arg;while (!kthread_should_stop ()) { -try_to_freeze (); +try_todo_list ();msleep_interruptible (2000 );#ifndef DEBUG
drivers/media/dvb/dvb-core/dvb_frontend.c +1 -1
@@ -520,7 +520,7 @@ static int dvb_frontend_thread(void *datbreak ; } -try_to_freeze (); +try_todo_list ();if (down_interruptible (&fepriv->sem))break ;
drivers/media/video/msp3400-driver.c +1 -1
@@ -398,7 +398,7 @@ int msp_sleep(struct msp_state *state, i }remove_wait_queue (&state->wq, &wait); -try_to_freeze (); +try_todo_list ();return state->restart; }
drivers/media/video/tvaudio.c +1 -1
@@ -283,7 +283,7 @@ static int chip_thread(void *data)schedule (); }remove_wait_queue (&chip->wq, &wait); -try_to_freeze (); +try_todo_list ();if (chip->done ||signal_pending (current))break ;v4l_dbg (1 , debug, &chip->c,"%s: thread wakeup\n" , chip->c.name);
drivers/media/video/video-buf-dvb.c +1 -1
@@ -62,7 +62,7 @@ static int videobuf_dvb_thread(void *datbreak ;if (kthread_should_stop ())break ; -try_to_freeze (); +try_todo_list ();/* feed buffer data to demux */ if (buf->state ==STATE_DONE )
drivers/net/irda/stir4200.c +2 -2
@@ -762,7 +762,7 @@ static int stir_transmit_thread(void *ar {#ifdef CONFIG_PM /* if suspending, then power off and wait */ -if (unlikely (freezing (current))) { +if (unlikely (todo_list_active ())) {if (stir->receiving)receive_stop (stir);else
@@ -770,7 +770,7 @@ static int stir_transmit_thread(void *arwrite_reg (stir,REG_CTRL1 ,CTRL1_TXPWD |CTRL1_RXPWD ); -refrigerator (); +run_todo_list ();if (change_speed (stir, stir->speed))break ;
drivers/net/wireless/airo.c +1 -1
@@ -2910,7 +2910,7 @@ static int airo_thread(void *data) {flush_signals (current);/* make swsusp happy with our thread */ -try_to_freeze (); +try_todo_list ();if (test_bit (JOB_DIE , &ai->flags))break ;
drivers/pcmcia/cs.c +1 -1
@@ -681,7 +681,7 @@ static int pccardd(void *__skt)break ;schedule (); -try_to_freeze (); +try_todo_list (); }/* make sure we are running before we exit */ set_current_state (TASK_RUNNING );
drivers/pnp/pnpbios/core.c +1 -1
@@ -170,7 +170,7 @@ static int pnp_dock_thread(void * unusedmsleep_interruptible (2000 );if (signal_pending (current)) { -if (try_to_freeze ()) +if (try_todo_list ())continue ;break ; }
drivers/usb/core/hub.c +1 -1
@@ -2870,7 +2870,7 @@ static int hub_thread(void *__unused)wait_event_interruptible (khubd_wait, !list_empty (&hub_event_list) ||kthread_should_stop ()); -try_to_freeze (); +try_todo_list (); }while (!kthread_should_stop () || !list_empty (&hub_event_list));pr_debug ("%s: khubd exiting\n" , usbcore_name);
drivers/usb/gadget/file_storage.c +1 -1
@@ -1557,7 +1557,7 @@ static int sleep_thread(struct fsg_dev */* Wait until a signal arrives or we are woken up */ for (;;) { -try_to_freeze (); +try_todo_list ();set_current_state (TASK_INTERRUPTIBLE );if (signal_pending (current)) { rc = -EINTR ;
drivers/usb/storage/usb.c +1 -1
@@ -860,7 +860,7 @@ retry:wait_event_interruptible_timeout (us->delay_wait,test_bit (US_FLIDX_DISCONNECTING , &us->flags), delay_use *HZ ); -if (try_to_freeze ()) +if (try_todo_list ()) goto retry; }
drivers/w1/w1.c +2 -2
@@ -720,7 +720,7 @@ static int w1_control(void *data)while (!control_needs_exit || have_to_wait) { have_to_wait =0 ; -try_to_freeze (); +try_todo_list ();msleep_interruptible (w1_control_timeout *1000 );if (signal_pending (current))
@@ -796,7 +796,7 @@ int w1_process(void *data)allow_signal (SIGTERM );while (!test_bit (W1_MASTER_NEED_EXIT , &dev->flags)) { -try_to_freeze (); +try_todo_list ();msleep_interruptible (w1_timeout *1000 );if (signal_pending (current))
fs/afs/kafsasyncd.c +1 -1
@@ -116,7 +116,7 @@ static int kafsasyncd(void *arg)remove_wait_queue (&kafsasyncd_sleepq, &myself);set_current_state (TASK_RUNNING ); -try_to_freeze (); +try_todo_list ();/* discard pending signals */ afs_discard_my_signals ();
fs/afs/kafstimod.c +1 -1
@@ -91,7 +91,7 @@ static int kafstimod(void *arg)complete_and_exit (&kafstimod_dead,0 ); } -try_to_freeze (); +try_todo_list ();/* discard pending signals */ afs_discard_my_signals ();
fs/jbd/journal.c +2 -2
@@ -153,7 +153,7 @@ loop: }wake_up (&journal->j_wait_done_commit); -if (freezing (current)) { +if (todo_list_active ()) {/* * The simpler the better. Flushing journal isn't a * good idea, because that depends on threads that may
@@ -161,7 +161,7 @@ loop:*/ jbd_debug (1 ,"Now suspending kjournald\n" );spin_unlock (&journal->j_state_lock); -refrigerator (); +run_todo_list ();spin_lock (&journal->j_state_lock); }else {/*
fs/jffs/intrep.c +1 -1
@@ -3391,7 +3391,7 @@ jffs_garbage_collect_thread(void *ptr)siginfo_t info;unsigned long signr =0 ; -if (try_to_freeze ()) +if (try_todo_list ())continue ;spin_lock_irq (¤t->sighand->siglock);
fs/jffs2/background.c +1 -1
@@ -96,7 +96,7 @@ static int jffs2_garbage_collect_thread(schedule (); } -if (try_to_freeze ()) +if (try_todo_list ())continue ;cond_resched ();
fs/jfs/jfs_logmgr.c +2 -2
@@ -2362,9 +2362,9 @@ int jfsIOWait(void *arg)lbmStartIO (bp);spin_lock_irq (&log_redrive_lock); } -if (freezing (current)) { +if (todo_list_active ()) {spin_unlock_irq (&log_redrive_lock); -refrigerator (); +run_todo_list (); }else {add_wait_queue (&jfs_IO_thread_wait, &wq);set_current_state (TASK_INTERRUPTIBLE );
fs/jfs/jfs_txnmgr.c +4 -4
@@ -2793,9 +2793,9 @@ int jfs_lazycommit(void *arg)/* In case a wakeup came while all threads were active */ jfs_commit_thread_waking =0 ; -if (freezing (current)) { +if (todo_list_active ()) {LAZY_UNLOCK (flags); -refrigerator (); +run_todo_list (); }else {DECLARE_WAITQUEUE (wq, current);
@@ -2992,9 +2992,9 @@ int jfs_sync(void *arg)/* Add anon_list2 back to anon_list */ list_splice_init (&TxAnchor.anon_list2, &TxAnchor.anon_list); -if (freezing (current)) { +if (todo_list_active ()) {TXN_UNLOCK (); -refrigerator (); +run_todo_list (); }else {DECLARE_WAITQUEUE (wq, current);
fs/lockd/clntlock.c +1 -0
@@ -237,6 +237,7 @@ restart: fl->fl_u.nfs_fl.flags &= ~NFS_LCK_RECLAIM ;nlmclnt_reclaim (host, fl); +try_todo_list ();if (signalled ())break ; goto restart;
fs/lockd/clntproc.c +1 -1
@@ -311,7 +311,7 @@ static int nlm_wait_on_grace(wait_queue_prepare_to_wait (queue, &wait,TASK_INTERRUPTIBLE );if (!signalled ()) {schedule_timeout (NLMCLNT_GRACE_WAIT ); -try_to_freeze (); +try_todo_list ();if (!signalled ()) status =0 ; }
fs/lockd/svc.c +2 -0
@@ -138,6 +138,8 @@ lockd(struct svc_rqst *rqstp)while ((nlmsvc_users || !signalled ()) && nlmsvc_pid == current->pid) {long timeout =MAX_SCHEDULE_TIMEOUT ; +try_todo_list (); +if (signalled ()) {flush_signals (current);if (nlmsvc_ops) {
fs/xfs/linux-2.6/xfs_buf.c +2 -2
@@ -1677,9 +1677,9 @@ xfsbufd(INIT_LIST_HEAD (&tmp); do { - if (unlikely(freezing(current))) { + if (unlikely(todo_list_active(current))) { set_bit(XBT_FORCE_SLEEP, &target->bt_flags); - refrigerator(); + run_todo_list(); } else { clear_bit(XBT_FORCE_SLEEP, &target->bt_flags); }
fs/xfs/linux-2.6/xfs_super.c +1 -1
@@ -578,7 +578,7 @@ xfssyncd(for (;;) { timeleft =schedule_timeout_interruptible (timeleft);/* swsusp */ -try_to_freeze (); +try_todo_list ();if (kthread_should_stop () &&list_empty (&vfsp->vfs_sync_list))break ;
include/linux/sched.h +21 -59
113 lines not on this page · raw message
kernel/audit.c +1 -1
@@ -289,7 +289,7 @@ static int kauditd_thread(void *dummy) }else {DECLARE_WAITQUEUE (wait, current); -try_to_freeze (); +try_todo_list ();set_current_state (TASK_INTERRUPTIBLE );add_wait_queue (&kauditd_wait, &wait);
kernel/sched.c +1 -1
@@ -4589,7 +4589,7 @@ static int migration_thread(void *data)struct list_head *head;migration_req_t *req; -try_to_freeze (); +try_todo_list ();spin_lock_irq (&rq->lock);
kernel/signal.c +2 -2
@@ -214,7 +214,7 @@ static inline int has_pending_signals(si fastcall void recalc_sigpending_tsk(struct task_struct *t) {if (t->signal->group_stop_count >0 || - (freezing (t)) || + (t->todo) ||PENDING (&t->pending, &t->blocked) ||PENDING (&t->signal->shared_pending, &t->blocked))set_tsk_thread_flag (t,TIF_SIGPENDING );
@@ -2307,7 +2307,7 @@ sys_rt_sigtimedwait(const sigset_t __use timeout =schedule_timeout_interruptible (timeout); -try_to_freeze (); +try_todo_list ();spin_lock_irq (¤t->sighand->siglock); sig =dequeue_signal (current, &these, &info); current->blocked = current->real_blocked;
kernel/workqueue.c +1 -1
@@ -211,7 +211,7 @@ static int worker_thread(void *__cwq)schedule (); else__set_current_state (TASK_RUNNING ); -try_to_freeze (); +try_todo_list ();remove_wait_queue (&cwq->more_work, &wait);if (!list_empty (&cwq->worklist))
mm/page_alloc.c +1 -0
@@ -1030,6 +1030,7 @@ rebalance: do_retry =1 ; }if (do_retry) { +try_todo_list ();blk_congestion_wait (WRITE ,HZ /50 ); goto rebalance; }
mm/pdflush.c +1 -1
@@ -106,7 +106,7 @@ static int __pdflush(struct pdflush_workspin_unlock_irq (&pdflush_lock);schedule (); -if (try_to_freeze ()) { +if (try_todo_list ()) {spin_lock_irq (&pdflush_lock);continue ; }
mm/vmscan.c +1 -1
@@ -1463,7 +1463,7 @@ static int kswapd(void *p)for ( ; ; ) {unsigned long new_order; -try_to_freeze (); +try_todo_list ();prepare_to_wait (&pgdat->kswapd_wait, &wait,TASK_INTERRUPTIBLE ); new_order = pgdat->kswapd_max_order;
net/rxrpc/krxiod.c +1 -1
@@ -138,7 +138,7 @@ static int rxrpc_krxiod(void *arg)_debug ("### End Work" ); -try_to_freeze (); +try_todo_list ();/* discard pending signals */ rxrpc_discard_my_signals ();
net/rxrpc/krxsecd.c +1 -1
@@ -107,7 +107,7 @@ static int rxrpc_krxsecd(void *arg)_debug ("### End Inbound Calls" ); -try_to_freeze (); +try_todo_list ();/* discard pending signals */ rxrpc_discard_my_signals ();
net/rxrpc/krxtimod.c +1 -1
@@ -90,7 +90,7 @@ static int krxtimod(void *arg)complete_and_exit (&krxtimod_dead,0 ); } -try_to_freeze (); +try_todo_list ();/* discard pending signals */ rxrpc_discard_my_signals ();
net/sunrpc/sched.c +4 -0
@@ -671,6 +671,9 @@ static int __rpc_execute(struct rpc_task/* sync task: sleep here */ dprintk ("RPC: %4d sync task going to sleep\n" , task->tk_pid); + +try_todo_list (); +/* Note: Caller should be using rpc_clnt_sigmask() */ status = out_of_line_wait_on_bit(&task->tk_runstate, RPC_TASK_QUEUED, rpc_wait_bit_interruptible,
@@ -711,6 +714,7 @@ static int __rpc_execute(struct rpc_taskint rpc_execute (struct rpc_task *task) { +try_todo_list ();rpc_set_active (task);rpc_set_running (task);return __rpc_execute (task);
net/sunrpc/svcsock.c +2 -2
@@ -1177,7 +1177,7 @@ svc_recv(struct svc_serv *serv, struct s arg->len = (pages-1 )*PAGE_SIZE ; arg->tail[0 ].iov_len =0 ; -try_to_freeze (); +try_todo_list ();cond_resched ();if (signalled ())return -EINTR ;
@@ -1219,7 +1219,7 @@ svc_recv(struct svc_serv *serv, struct sschedule_timeout (timeout); -try_to_freeze (); +try_todo_list ();spin_lock_bh (&serv->sv_lock);remove_wait_queue (&rqstp->rq_wait, &wait);
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 5/23 Make the freezer use todo lists.
In reply to Nigel Cunningham(This is Christoph Lameter's original patch, slightly modified. I don't remove PF_FROZEN because it will still be used by later modifications. Christoph's original comment follows).
Make the suspend code use the todo list in the task_struct.
This patch makes the suspend code SMP clean by removing the current usage of PF_FREEZE and PF_FROZEN. Instead it relies on the new notification handler in the task_struct, a completion handler and an atomic counter for the number of processes frozen.
All the logic is local to the suspend code and no longer requires changes to other kernel components.
Documentation/power/kernel_threads.txt | 10 +- Documentation/power/swsusp.txt | 5 + include/linux/sched.h | 1 kernel/power/process.c | 134 ++++++++++++++++---------------- 4 files changed, 77 insertions(+), 73 deletions(-)
Documentation/power/kernel_threads.txt +5 -5
@@ -4,15 +4,15 @@ KERNEL THREADS Freezer Upon entering a suspended state the system will freeze all -tasks. This is done by delivering pseudosignals. This affects -kernel threads, too. To successfully freeze a kernel thread -the thread has to check for the pseudosignal and enter the -refrigerator. Code to do this looks like this: +tasks. This is done by making all processes execute a notifier. +This affects kernel threads, too. To successfully freeze a kernel thread +the thread has to check for the notifications and call the notifier +chain for the process. Code to do this looks like this: do { hub_events(); wait_event_interruptible(khubd_wait, !list_empty(&hub_event_list)); - try_to_freeze(); + try_todo_list(); } while (!signal_pending(current)); from drivers/usb/core/hub.c::hub_thread()
Documentation/power/swsusp.txt +3 -2
@@ -135,7 +135,8 @@ should be sent to the mailing list avail website, and not to the Linux Kernel Mailing List. We are working toward merging suspend2 into the mainline kernel. -Q: A kernel thread must voluntarily freeze itself (call 'refrigerator'). +Q: A kernel thread must work on the todo list (call 'run_todo_list') +to enter the refrigerator. I found some kernel threads that don't do it, and they don't freeze so the system can't sleep. Is this a known behavior?
@@ -144,7 +145,7 @@ place where the thread is safe to be fro should be held at that point and it must be safe to sleep there), and add: - try_to_freeze(); + try_todo_list(); If the thread is needed for writing the image to storage, you should instead set the PF_NOFREEZE process flag when creating the thread (and
include/linux/sched.h +0 -1
@@ -924,7 +924,6 @@ static inline void put_task_struct(struc#define PF_MEMALLOC 0x00000800/* Allocating memory */ #define PF_FLUSHER 0x00001000/* responsible for disk writeback */ #define PF_USED_MATH 0x00002000/* if unset the fpu must be initialized before use */ -#define PF_FREEZE 0x00004000/* this task is being frozen for suspend now */ #define PF_NOFREEZE 0x00008000/* this thread should not be frozen */ #define PF_FROZEN 0x00010000/* frozen for system suspend */ #define PF_FSTRANS 0x00020000/* inside a filesystem transaction */
kernel/power/process.c +69 -65
177 lines not on this page · raw message
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 6/23 Disable usermode helper invocations when the freezer is on.
In reply to Nigel CunninghamDisable usermode helper invocations when the freezer is on. This avoids deadlocks due to hotplug events occuring while processes are frozen.
kernel/kmod.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)
kernel/kmod.c +4 -0
@@ -36,6 +36,7 @@#include <linux/mount.h> #include <linux/kernel.h> #include <linux/init.h> +#include <linux/freezer.h> #include <asm/uaccess.h> extern int max_threads;
@@ -249,6 +250,9 @@ int call_usermodehelper_keys(char *path,if (!khelper_wq)return -EBUSY ; +if (freezer_is_on ()) +return 0 ; +if (path[0 ] =='\0' )return 0 ;
--
Nigel Cunningham nigel at suspend2 dot net
Re: [ 06/23] [Suspend2] Disable usermode helper invocations when the freezer is on.
Hi,
Quoted 22 lines · Nigel Cunningham
Disable usermode helper invocations when the freezer is on. This avoids deadlocks due to hotplug events occuring while processes are frozen.
Signed-off-by Nigel Cunningham <[email protected]>kernel/kmod.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)kernel/kmod.c +1
-0@@ -36,6 +36,7 @@#include <linux/mount.h> #include <linux/kernel.h> #include <linux/init.h> +#include <linux/freezer.h> #include <asm/uaccess.h> extern int max_threads;
kernel/kmod.c int call_usermodehelper_keys(char *path, · line 258
Quoted 9 lines
kernel/kmod.c +3
-0@@ -249,6 +250,9 @@ int call_usermodehelper_keys(char *path,if (!khelper_wq)return -EBUSY ; +if (freezer_is_on ()) +return 0 ; +if (path[0 ] =='\0' )return 0 ;
Disabling the usermode helper while freeze_processes() is executed seems to be a good idea to me, but I think it should be done with a mutex or something like that.
Greetings,
Rafael
Re: [ 06/23] [Suspend2] Disable usermode helper invocations when the freezer is on.
Hi.
Quoted 27 lines · Rafael J. Wysocki
Hi,
Quoted 24 lines · Nigel Cunningham
Disable usermode helper invocations when the freezer is on. This avoids deadlocks due to hotplug events occuring while processes are frozen.
Signed-off-by Nigel Cunningham <[email protected]>kernel/kmod.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)kernel/kmod.c +1
-0@@ -36,6 +36,7 @@#include <linux/mount.h> #include <linux/kernel.h> #include <linux/init.h> +#include <linux/freezer.h> #include <asm/uaccess.h> extern int max_threads;@@ -249,6 +250,9 @@ int call_usermodehelper_keys(char *path,if (!khelper_wq)return -EBUSY ;
Quoted 9 lines
kernel/kmod.c int call_usermodehelper_keys(char *path, · line 257
Quoted 5 lines
kernel/kmod.c +3
-0+if (freezer_is_on ()) +return 0 ; +if (path[0 ] =='\0' )return 0 ;Disabling the usermode helper while
freeze_processes()is executed seems to be a good idea to me, but I think it should be done with a mutex or something like that.
With the refrigerator code you guys are using at the moment, ouldn't that result in deadlocks when we later try to freeze the process in preparation for the atomic restore? (Or perhaps you don't freeze processes at that point?)
Regards,
Nigel
Quoted 7 lines
Greetings,
Rafael
-
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/
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
Re: [ 06/23] [Suspend2] Disable usermode helper invocations when the freezer is on.
Hi,
Quoted 36 lines · Nigel Cunningham
Quoted 25 lines · Rafael J. Wysocki
Quoted 24 lines · Nigel Cunningham
Disable usermode helper invocations when the freezer is on. This avoids deadlocks due to hotplug events occuring while processes are frozen.
Signed-off-by Nigel Cunningham <[email protected]>kernel/kmod.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)kernel/kmod.c +1
-0@@ -36,6 +36,7 @@#include <linux/mount.h> #include <linux/kernel.h> #include <linux/init.h> +#include <linux/freezer.h> #include <asm/uaccess.h> extern int max_threads;@@ -249,6 +250,9 @@ int call_usermodehelper_keys(char *path,if (!khelper_wq)return -EBUSY ;Quoted 9 lines
kernel/kmod.c int call_usermodehelper_keys(char *path, · line 257
Quoted 5 lines
kernel/kmod.c +3
-0+if (freezer_is_on ()) +return 0 ; +if (path[0 ] =='\0' )return 0 ;Disabling the usermode helper while
freeze_processes()is executed seems to be a good idea to me, but I think it should be done with a mutex or something like that.
Quoted 4 lines
With the refrigerator code you guys are using at the moment, ouldn't that result in deadlocks when we later try to freeze the process in preparation for the atomic restore? (Or perhaps you don't freeze processes at that point?)
I'm not sure what you mean. I said "mutex" because you seem to have a race here (the freezer may be started right after the freezer_is_on() check). IMO the freezer should disable the invocations of new usermode helpers and wait util all of the already running helpers are finished. For this purpose two variables would be needed and a lock.
Greetings,
Rafael
Re: [ 06/23] [Suspend2] Disable usermode helper invocations when the freezer is on.
Hi.
Quoted 17 lines · Rafael J. Wysocki
Hi,
Quoted 11 lines
Quoted 10 lines
kernel/kmod.c int call_usermodehelper_keys(char *path, · line 257
Quoted 6 lines
kernel/kmod.c +3
-0+if (freezer_is_on ()) +return 0 ; +if (path[0 ] =='\0' )return 0 ;Disabling the usermode helper while
freeze_processes()is executed seems to be a good idea to me, but I think it should be done with a mutex or something like that.Quoted 4 lines
With the refrigerator code you guys are using at the moment, ouldn't that result in deadlocks when we later try to freeze the process in preparation for the atomic restore? (Or perhaps you don't freeze processes at that point?)
Quoted 5 lines
I'm not sure what you mean. I said "mutex" because you seem to have a race here (the freezer may be started right after the
freezer_is_on()check). IMO the freezer should disable the invocations of new usermode helpers and wait util all of the already running helpers are finished. For this purpose two variables would be needed and a lock.
Sorry. Being a bit thick.
I wasn't worried about already-running usermode helpers (or about-to-run helpers either) because the spawned processes would either complete or be caught be the usual freezing code. My concern was more that new invocations of this path don't leave us with unfrozen processes hanging around. That said, I think you have a valid point about letting existing helpers run to completion. It does make me concerned though about the possibility of a long-lived helper (not that I know that there really are such things at the moment). Do you think that needs to be taken as a genuine concern? If so, I guess we then need to make freezing a four stage process:
1. Stop new usermodehelpers from starting & let existing ones run to completion.
2. Freeze userspace.
3. Freezer bdevs.
4. Freezer kernel space.
Regards,
Nigel
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
Re: [ 06/23] [Suspend2] Disable usermode helper invocations when the freezer is on.
Hi,
Quoted 36 lines · Nigel Cunningham
Quoted 17 lines · Rafael J. Wysocki
Hi,
Quoted 11 lines
Quoted 10 lines
kernel/kmod.c int call_usermodehelper_keys(char *path, · line 257
Quoted 6 lines
kernel/kmod.c +3
-0+if (freezer_is_on ()) +return 0 ; +if (path[0 ] =='\0' )return 0 ;Disabling the usermode helper while
freeze_processes()is executed seems to be a good idea to me, but I think it should be done with a mutex or something like that.Quoted 4 lines
With the refrigerator code you guys are using at the moment, ouldn't that result in deadlocks when we later try to freeze the process in preparation for the atomic restore? (Or perhaps you don't freeze processes at that point?)
Quoted 5 lines
I'm not sure what you mean. I said "mutex" because you seem to have a race here (the freezer may be started right after the
freezer_is_on()check). IMO the freezer should disable the invocations of new usermode helpers and wait util all of the already running helpers are finished. For this purpose two variables would be needed and a lock.Sorry. Being a bit thick.
I wasn't worried about already-running usermode helpers (or about-to-run helpers either) because the spawned processes would either complete or be caught be the usual freezing code. My concern was more that new invocations of this path don't leave us with unfrozen processes hanging around. That said, I think you have a valid point about letting existing helpers run to completion. It does make me concerned though about the possibility of a long-lived helper (not that I know that there really are such things at the moment). Do you think that needs to be taken as a genuine concern? If so, I guess we then need to make freezing a four stage process:
Quoted 5 lines
1. Stop new usermodehelpers from starting & let existing ones run to completion.
2. Freeze userspace.
3. Freezer bdevs.
4. Freezer kernel space.
Well, I know a little about bdevs, but generally I think that's reasonable.
Greetings,
Rafael
PATCH 7/23 Add which-to-thaw parameter to thaw_processes() calls.
In reply to Nigel CunninghamModify the existing invocations of thaw_processes so that they use the new parameter specifying which processes (kernel only or all) to thaw.
kernel/power/disk.c | 5 +++-- kernel/power/main.c | 7 ++++--- 2 files changed, 7 insertions(+), 5 deletions(-)
kernel/power/disk.c +3 -2
@@ -10,6 +10,7 @@*/ #include <linux/suspend.h> +#include <linux/freezer.h> #include <linux/syscalls.h> #include <linux/reboot.h> #include <linux/string.h>
@@ -106,7 +107,7 @@ static int prepare_processes(void)if (!(error =swsusp_shrink_memory ()))return 0 ; thaw: -thaw_processes (); +thaw_processes (FREEZER_ALL_THREADS );enable_nonboot_cpus ();pm_restore_console ();return error;
@@ -115,7 +116,7 @@ thaw:static void unprepare_processes (void ) {platform_finish (); -thaw_processes (); +thaw_processes (FREEZER_ALL_THREADS );enable_nonboot_cpus ();pm_restore_console (); }
kernel/power/main.c +4 -3
@@ -9,6 +9,7 @@*/ #include <linux/suspend.h> +#include <linux/freezer.h> #include <linux/kobject.h> #include <linux/string.h> #include <linux/delay.h>
@@ -95,7 +96,7 @@ static int suspend_prepare(suspend_stateif (pm_ops->finish) pm_ops->finish (state); Thaw: -thaw_processes (); +thaw_processes (FREEZER_ALL_THREADS ); Enable_cpu:enable_nonboot_cpus ();pm_restore_console ();
@@ -103,7 +104,7 @@ static int suspend_prepare(suspend_state } -static int suspend_enter (suspend_state_t state) +int suspend_enter (suspend_state_t state) {int error =0 ;unsigned long flags;
@@ -135,7 +136,7 @@ static void suspend_finish(suspend_statedevice_resume ();if (pm_ops && pm_ops->finish) pm_ops->finish (state); -thaw_processes (); +thaw_processes (FREEZER_ALL_THREADS );enable_nonboot_cpus ();pm_restore_console (); }
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 8/23 New freezer explanation for kernel/power/process.c
In reply to Nigel CunninghamModify kernel/power/process.c header to describe how the new freezer implementation works.
kernel/power/process.c | 26 +++++++++++++++++++++++--- 1 files changed, 23 insertions(+), 3 deletions(-)
kernel/power/process.c +23 -3
@@ -1,8 +1,28 @@/* -* drivers/power/process.c - Functions for starting/stopping processes on -* suspend transitions. +* kernel/power/process.c * -* Originally from swsusp. +* Copyright (C) 1998-2001 Gabor Kuti <[email protected]> +* Copyright (C) 1998,2001,2002 Pavel Machek <[email protected]> +* Copyright (C) 2002-2003 Florent Chabaud <[email protected]> +* Copyright (C) 2002-2006 Nigel Cunningham <[email protected]> +* +* This file is released under the GPLv2. +* +* It contains the routines used to freeze processes during a suspend +* cycle (or perhaps in future, a process migration). +* +* When quiescing the whole system, one process initiates this +* functionality by calling freeze_processes(). First, userspace +* is frozen. Then bdevs for mounted filesystems are frozen so +* as to ensure pending I/O is flushed and stopped. Finally, +* kernel threads which are no marked as unfreezeable (perhaps +* because they're needed for doing I/O) are frozen. Splitting +* kernel threads from userspace also improves reliability, +* because it ensures that a userspace thread doesn't deadlock, +* waiting for a kernel thread that has already been frozen to +* process some request, and because it stops the cause of most +* work for kernel threads (ie userspace), allowing those thread +* to complete their work more quickly. */
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 9/23 Quieten the freezer for normal operation.
In reply to Nigel CunninghamQuieten the freezer by replacing printks with a local #define that can be enabled if and when debugging is needed.
kernel/power/process.c | 19 +++++++++++++------ 1 files changed, 13 insertions(+), 6 deletions(-)
kernel/power/process.c +13 -6
@@ -33,6 +33,13 @@#include <linux/suspend.h> #include <linux/module.h> +#if 0 +//#ifdef CONFIG_PM_DEBUG +#define freezer_message(msg, a...) do { printk(msg, ##a); } while(0) +#else +#define freezer_message(msg, a...) do { } while(0) +#endif +/* * Timeout for stopping processes */
@@ -67,7 +74,7 @@ static int freeze_process(struct notifie current->flags |=PF_FROZEN ;notifier_chain_unregister (¤t->todo, nl);kfree (nl); -printk ("=" ); +freezer_message ("=" );spin_lock_irq (¤t->sighand->siglock);recalc_sigpending ();/* We sent fake signal, clean it up */
@@ -84,11 +91,11 @@ static int freeze_process(struct notifievoid thaw_processes (void ) { -printk ("Restarting tasks.." ); +freezer_message ("Restarting tasks.." );complete_all (&thaw);while (atomic_read (&nr_frozen) >0 )schedule (); -printk ("done\n" ); +freezer_message ("done\n" ); }static inline void freeze (struct task_struct *p)
@@ -127,7 +134,7 @@ int freeze_processes(void)atomic_set (&nr_frozen,0 );INIT_COMPLETION (thaw); -printk ("Stopping tasks: " ); +freezer_message ("Stopping tasks: " ); start_time = jiffies; do { todo = 0;
@@ -142,7 +149,7 @@ int freeze_processes(void)read_unlock (&tasklist_lock);yield ();/* Yield is okay here */ if (todo &&time_after (jiffies, start_time +TIMEOUT )) { -printk ("\n" ); +freezer_message ("\n" );printk (KERN_ERR " stopping tasks failed" "(%d tasks remaining)\n" , todo -atomic_read (&nr_frozen));
@@ -151,7 +158,7 @@ int freeze_processes(void) } }while (atomic_read (&nr_frozen) < todo); -printk ("|\n" ); +freezer_message ("|\n" );BUG_ON (in_atomic ());return 0 ; }
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 10/23 Add support for freezing filesystem bdevs.
In reply to Nigel CunninghamAdd support for freezing and thawing bdevs of filesystems. Filesystems are frozen in reverse order so that nested filesystems don't cause deadlocks.
kernel/power/process.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 50 insertions(+), 0 deletions(-)
kernel/power/process.c +50 -0
@@ -48,6 +48,56 @@DECLARE_COMPLETION (thaw);static atomic_t nr_frozen; +struct frozen_fs +{ +struct list_head fsb_list; +struct super_block *sb; +}; + +LIST_HEAD (frozen_fs_list); + +void freezer_make_fses_rw (void ) +{ +struct frozen_fs *fs, *next_fs; + +list_for_each_entry_safe (fs, next_fs, &frozen_fs_list, fsb_list) { +thaw_bdev (fs->sb->s_bdev, fs->sb); + +list_del (&fs->fsb_list); +kfree (fs); + } +} + +/* +* Done after userspace is frozen, so there should be no danger of +* fses being unmounted while we're in here. +*/ +int freezer_make_fses_ro (void ) +{ +struct frozen_fs *fs; +struct super_block *sb; + +/* Generate the list */ +list_for_each_entry (sb, &super_blocks, s_list) { +if (!sb->s_root || !sb->s_bdev || + (sb->s_frozen ==SB_FREEZE_TRANS ) || + (sb->s_flags &MS_RDONLY )) +continue ; + + fs =kmalloc (sizeof (struct frozen_fs ),GFP_ATOMIC ); + fs->sb = sb; +list_add_tail (&fs->fsb_list, &frozen_fs_list); + }; + +/* Do the freezing in reverse order so filesystems dependant +* upon others are frozen in the right order. (Eg loopback +* on ext3). */ +list_for_each_entry_reverse (fs, &frozen_fs_list, fsb_list) +freeze_bdev (fs->sb->s_bdev); + +return 0 ; +} +static inline int freezeable (struct task_struct * p) { if ((p == current) ||
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 11/23 Modify freezeable for freezing kernel threads separately.
In reply to Nigel CunninghamModify freezeable so that it takes an extra argument indicating whether we are interested in all threads, or just userspace ones, and ignores threads that are already in the refrigerator.
kernel/power/process.c | 14 ++++++++++++-- 1 files changed, 12 insertions(+), 2 deletions(-)
kernel/power/process.c +12 -2
@@ -98,14 +98,24 @@ int freezer_make_fses_ro(void)return 0 ; } -static inline int freezeable (struct task_struct * p) +/* +* freezeable +* +* Description: Determine whether a process should be frozen yet. +* Parameters: struct task_struct * The process to consider. +* int Boolean - 0 = userspace else all. +* Returns: int 0 if don't freeze yet, otherwise do. +*/ +static int freezeable (struct task_struct * p,int all_freezable) {if ((p == current) || + (p->flags &PF_FROZEN ) || (p->flags &PF_NOFREEZE ) || (p->exit_state ==EXIT_ZOMBIE ) || (p->exit_state ==EXIT_DEAD ) || (p->state ==TASK_STOPPED ) || - (p->state ==TASK_TRACED )) + (p->state ==TASK_TRACED ) || + (!p->mm && !all_freezable))return 0 ;return 1 ; }
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 12/23 Split freezing of threads according to whether user/kernelspace.
In reply to Nigel CunninghamSplit the freezing of threads according to whether they're userspace or kernelspace threads. We use separate completion handlers, thereby allowing for the possibility of thawing kernel space without thawing userspace. This will be used to allow memory to be freed without it racing against userspace.
kernel/power/process.c | 71 +++++++++++++++++++++++++++++++++++------------- 1 files changed, 52 insertions(+), 19 deletions(-)
kernel/power/process.c +52 -19
@@ -45,8 +45,11 @@*/ #define TIMEOUT (6 * HZ) -DECLARE_COMPLETION (thaw); -static atomic_t nr_frozen; + +DECLARE_COMPLETION (kernelspace_thaw); +DECLARE_COMPLETION (userspace_thaw); +static atomic_t nr_userspace_frozen; +static atomic_t nr_kernelspace_frozen;struct frozen_fs {
@@ -120,32 +123,62 @@ static int freezeable(struct task_structreturn 1 ; } +static void __freeze_process (struct completion *completion_handler, +atomic_t *nr_frozen) +{ +long save; + +freezer_message ("%s (%d) frozen.\n" , + current->comm, current->pid); + save = current->state; + +atomic_inc (nr_frozen); +wait_for_completion (completion_handler); +atomic_dec (nr_frozen); + + current->state = save; +freezer_message ("%s (%d) leaving freezer.\n" , + current->comm, current->pid); +} +/* * Invoked by the task todo list notifier when the task to be * frozen is running. */ static int freeze_process (struct notifier_block *nl,unsigned long x,void *v) { -/* Hmm, should we be allowed to suspend when there are realtime -processes around? */ -long save; - save = current->state; -pr_debug ("%s frozen\n" , current->comm); +unsigned long flags; + +might_sleep (); + +/* Locking to handle race against waking the process in +* freeze threads. */ +spin_lock_irqsave (¤t->sighand->siglock, flags); current->flags |=PF_FROZEN ; -notifier_chain_unregister (¤t->todo, nl); -kfree (nl); -freezer_message ("=" ); - -spin_lock_irq (¤t->sighand->siglock); -recalc_sigpending ();/* We sent fake signal, clean it up */ -atomic_inc (&nr_frozen); -spin_unlock_irq (¤t->sighand->siglock); -wait_for_completion (&thaw); +if (nl) +notifier_chain_unregister (¤t->todo, nl); + +recalc_sigpending (); +spin_unlock_irqrestore (¤t->sighand->siglock, flags); + +if (nl) +kfree (nl); + +if (test_freezer_state (FREEZER_ON )) { +if (current->mm) +__freeze_process (&userspace_thaw, &nr_userspace_frozen); +else +__freeze_process (&kernelspace_thaw, + &nr_kernelspace_frozen); + } + +spin_lock_irqsave (¤t->sighand->siglock, flags); +recalc_sigpending (); +spin_unlock_irqrestore (¤t->sighand->siglock, flags); + current->flags &= ~PF_FROZEN ; -atomic_dec (&nr_frozen); -pr_debug ("%s thawed\n" , current->comm); - current->state = save; +return 0 ; }
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 13/23 Add support for thawing just kernel threads or all threads.
In reply to Nigel CunninghamModify the thaw_processes routine so that it takes and implements a parameter saying whether to thaw all processes, or just kernel space.
kernel/power/process.c | 25 +++++++++++++++++++------ 1 files changed, 19 insertions(+), 6 deletions(-)
kernel/power/process.c +19 -6
@@ -182,13 +182,26 @@ static int freeze_process(struct notifiereturn 0 ; } -void thaw_processes (void ) +void thaw_processes (int do_all_threads) { -freezer_message ("Restarting tasks.." ); -complete_all (&thaw); -while (atomic_read (&nr_frozen) >0 ) -schedule (); -freezer_message ("done\n" ); +if (do_all_threads) { +clear_freezer_state (FREEZER_ON ); +clear_freezer_state (ABORT_FREEZING ); + } + +complete_all (&kernelspace_thaw); +while (atomic_read (&nr_kernelspace_frozen) >0 ) +yield (); + +init_completion (&kernelspace_thaw); +freezer_make_fses_rw (); + +if (do_all_threads) { +complete_all (&userspace_thaw); +while (atomic_read (&nr_userspace_frozen) >0 ) +yield (); +init_completion (&userspace_thaw); + } }static inline void freeze (struct task_struct *p)
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 14/23 Helper for counting freezeable threads of a type.
In reply to Nigel CunninghamAdd a helper, which returns the number of threads of the given type (userspace only or all) that are still to be frozen.
kernel/power/process.c | 23 +++++++++++++++++++++++ 1 files changed, 23 insertions(+), 0 deletions(-)
kernel/power/process.c +23 -0
@@ -204,6 +204,29 @@ void thaw_processes(int do_all_threads) } } +/* +* num_freezeable +* +* Description: Determine how many processes of our type are still to be +* frozen. As a side effect, update the progress bar too. +* Parameters: int Which type we are trying to freeze. +* int Whether we are displaying our progress. +*/ +static int num_freezeable (int do_all_threads) +{ +struct task_struct *g, *p; +int todo_this_type =0 ; + +read_lock (&tasklist_lock); + do_each_thread(g, p) { +if (freezeable (p, do_all_threads)) + todo_this_type++; + }while_each_thread (g, p); +read_unlock (&tasklist_lock); + +return todo_this_type; +} +static inline void freeze (struct task_struct *p) {unsigned long flags;
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 15/23 Helper for counting uninterruptible threads of a type.
In reply to Nigel CunninghamAdd a helper which counts the number of patches of a type (all or userspace only) which are in TASK_UNINTERRUPTIBLE state. These tasks are signalled (just in case they leave that state at a later point), but we do not consider freezing to have failed if and when they do not enter the freezer.
Note that when they eventually leave TASK_UNINTERRUPTIBLE state, they will enter the refrigerator, but will immediately exit if we no longer want to freeze at that point.
kernel/power/process.c | 23 +++++++++++++++++++++++ 1 files changed, 23 insertions(+), 0 deletions(-)
kernel/power/process.c +23 -0
@@ -227,6 +227,29 @@ static int num_freezeable(int do_all_thrreturn todo_this_type; } +/* +* num_uninterruptible +* +* Description: Determine how many processes of our type are in state +* task uninterruptible. +* Parameters: int Which type we are trying to freeze. +*/ +static int num_uninterruptible (int do_all_threads) +{ +struct task_struct *g, *p; +int count =0 ; + +read_lock (&tasklist_lock); + do_each_thread(g, p) { +if (freezeable (p, do_all_threads) && + p->state ==TASK_UNINTERRUPTIBLE ) + count++; + }while_each_thread (g, p); +read_unlock (&tasklist_lock); + +return count; +} +static inline void freeze (struct task_struct *p) {unsigned long flags;
--
Nigel Cunningham nigel at suspend2 dot net
Re: [ 15/23] [Suspend2] Helper for counting uninterruptible threads of a type.
Hi,
Quoted 10 lines · Nigel Cunningham
Add a helper which counts the number of patches of a type (all or userspace only) which are in
TASK_UNINTERRUPTIBLEstate. These tasks are signalled (just in case they leave that state at a later point), but we do not consider freezing to have failed if and when they do not enter the freezer.Note that when they eventually leave
TASK_UNINTERRUPTIBLEstate, they will enter the refrigerator, but will immediately exit if we no longer want to freeze at that point.
I think we need to do something like this to prevent problems with freezing under load.
Greetings,
Rafael
Re: [ 15/23] [Suspend2] Helper for counting uninterruptible threads of a type.
Quoted 10 lines · Rafael J. Wysocki
Hi,
Quoted 7 lines · Nigel Cunningham
Add a helper which counts the number of patches of a type (all or userspace only) which are in
TASK_UNINTERRUPTIBLEstate. These tasks are signalled (just in case they leave that state at a later point), but we do not consider freezing to have failed if and when they do not enter the freezer.
Quoted 6 lines
Quoted 3 lines
Note that when they eventually leave
TASK_UNINTERRUPTIBLEstate, they will enter the refrigerator, but will immediately exit if we no longer want to freeze at that point.I think we need to do something like this to prevent problems with freezing under load.
That is dangerous... task in UNINTERRUPTIBLE may hold some lock, AFAICT.
No, there's some simple bug in refrigerator, and I/we need to fix that. Signals work under load, so refrigerator should, too.
Pavel
Signature · 1 line
Thanks, Sharp!
Re: [ 15/23] [Suspend2] Helper for counting uninterruptible threads of a type.
Hi,
Quoted 13 lines · Pavel Machek
Quoted 12 lines · Rafael J. Wysocki
Quoted 10 lines · Nigel Cunningham
Add a helper which counts the number of patches of a type (all or userspace only) which are in
TASK_UNINTERRUPTIBLEstate. These tasks are signalled (just in case they leave that state at a later point), but we do not consider freezing to have failed if and when they do not enter the freezer.Note that when they eventually leave
TASK_UNINTERRUPTIBLEstate, they will enter the refrigerator, but will immediately exit if we no longer want to freeze at that point.
Quoted 5 lines
Quoted 2 lines
I think we need to do something like this to prevent problems with freezing under load.
That is dangerous... task in UNINTERRUPTIBLE may hold some lock, AFAICT.
Yes, and we have discussed that already, but frankly I'm still unconvinced. ;-)
Quoted 2 lines
No, there's some simple bug in refrigerator, and I/we need to fix that. Signals work under load, so refrigerator should, too.
I don't think there's a bug as such. The refrigerator is just very simple and apparently does not cover all possible cases.
I think the problems with freezing tasks are generally related to uninterruptible processes waiting for events that never happen.
IMHO we can try to defer calling freeze() for kernel threads until all of the user space processes are frozen. If that doesn't help, we'll need to treat uninterruptible tasks in a special way, I'm afraid.
Greetings,
Rafael
[RFC/RFT] finally solve "swsusp fails with mysqld" problem
Place refrigerator hook at more clever place; avoids "system can't be suspended while mysqld running" problem.
I'd like you to test it. It looks correct to me, and it is actually a solution, not a workaround like my previous tries. It still does not solve suspend while running stress tests.
---
commit 01d049aacf36961d361cfd382fcbf746afcbb61b tree 67e3d1fcb65cebfab058f9a699534b1689494dad parent 92dfb5b4ac96e200138006901837056825c758d3 author <[email protected]> Tue, 31 Jan 2006 10:18:22 +0100 committer <[email protected]> Tue, 31 Jan 2006 10:18:22 +0100
arch/i386/kernel/signal.c | 3 --- fs/jbd/journal.c | 1 + kernel/signal.c | 4 ++-- 3 files changed, 3 insertions(+), 5 deletions(-)
arch/i386/kernel/signal.c +0 -3
@@ -615,9 +615,6 @@ int fastcall do_signal(struct pt_regs *rif (!user_mode (regs))return 1 ; -if (try_to_freeze ()) - goto no_signal; -if (!oldset) oldset = ¤t->blocked;
fs/jbd/journal.c +1 -0
@@ -153,6 +153,7 @@ loop: }wake_up (&journal->j_wait_done_commit); +/* Race here? May someone already be waking at *next* commit? */ if (freezing (current)) {/* * The simpler the better. Flushing journal isn't a
kernel/signal.c +2 -1
@@ -1922,6 +1922,8 @@ int get_signal_to_deliver(siginfo_t *infsigset_t *mask = ¤t->blocked;int signr =0 ; +try_to_freeze (); + relock:spin_lock_irq (¤t->sighand->siglock);for (;;) {
@@ -2307,7 +2309,6 @@ sys_rt_sigtimedwait(const sigset_t __use timeout =schedule_timeout_interruptible (timeout); -try_to_freeze ();spin_lock_irq (¤t->sighand->siglock); sig =dequeue_signal (current, &these, &info); current->blocked = current->real_blocked;
arch/x86_64/kernel/signal.c +0 -3
@@ -443,9 +443,6 @@ int do_signal(struct pt_regs *regs, sigsif (!user_mode (regs))return 1 ; -if (try_to_freeze ()) - goto no_signal; -if (!oldset) oldset = ¤t->blocked;
Signature · 1 line
Thanks, Sharp!
Re: [RFC/RFT] finally solve "swsusp fails with mysqld" problem
Hi,
Quoted 7 lines · Pavel Machek
Place refrigerator hook at more clever place; avoids "system can't be suspended while mysqld running" problem.
I'd like you to test it. It looks correct to me, and it is actually a solution, not a workaround like my previous tries. It still does not solve suspend while running stress tests.
Which kernel is it against? It does not apply to the recent -mm ...
Rafael
Re: [RFC/RFT] finally solve "swsusp fails with mysqld" problem
Hi!
Quoted 8 lines
Quoted 6 lines
Place refrigerator hook at more clever place; avoids "system can't be suspended while mysqld running" problem.
I'd like you to test it. It looks correct to me, and it is actually a solution, not a workaround like my previous tries. It still does not solve suspend while running stress tests.
Which kernel is it against? It does not apply to the recent -mm ...
Against vanilla -- but it should be easy to make it work against -mm.
Pavel
Signature · 1 line
Thanks, Sharp!
Re: [RFC/RFT] finally solve "swsusp fails with mysqld" problem
Hi,
Quoted 10 lines · Pavel Machek
Quoted 8 lines
Quoted 6 lines
Place refrigerator hook at more clever place; avoids "system can't be suspended while mysqld running" problem.
I'd like you to test it. It looks correct to me, and it is actually a solution, not a workaround like my previous tries. It still does not solve suspend while running stress tests.
Which kernel is it against? It does not apply to the recent -mm ...
Against vanilla -- but it should be easy to make it work against -mm.
Yes, I've already done it.
BTW, the no_signal labels are no longer needed in signal.c for i386 and x86_64.
Greetings,
Rafael
Re: [ 15/23] [Suspend2] Helper for counting uninterruptible threads of a type.
In reply to Pavel MachekHi.
Quoted 14 lines · Pavel Machek
Quoted 13 lines · Rafael J. Wysocki
Hi,
Quoted 9 lines · Nigel Cunningham
Add a helper which counts the number of patches of a type (all or userspace only) which are in
TASK_UNINTERRUPTIBLEstate. These tasks are signalled (just in case they leave that state at a later point), but we do not consider freezing to have failed if and when they do not enter the freezer.Note that when they eventually leave
TASK_UNINTERRUPTIBLEstate, they will enter the refrigerator, but will immediately exit if we no longer want to freeze at that point.
Quoted 8 lines
Quoted 2 lines
I think we need to do something like this to prevent problems with freezing under load.
That is dangerous... task in UNINTERRUPTIBLE may hold some lock, AFAICT.
No, there's some simple bug in refrigerator, and I/we need to fix that. Signals work under load, so refrigerator should, too.
I know you understand English, Pavel! Re-read what I wrote previously! The problem is a race. You're telling processes that process I/O to freeze at the same time as you're telling processes that submit I/O to freeze. If kjournald (eg) enters the refrigerator while dd is still running, dd is going to have a chance of getting stuck, waiting for some I/O to complete.
This also means your sys_sync is totally useless if anything is submitting I/O, since there will still be I/O being submitted after the sync is done. This is why (in the numbers I submitted yesterday), the current implementation took so much longer than the new one. The sys_sync really only introduces a big delay. It needs to be done after the things that generate I/O have been stopped.
Regards,
Nigel
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
PATCH 16/23 Helper to signal all threads of a type.
In reply to Nigel CunninghamAdd a helper to do the actual signalling of processes, telling them to enter the refrigerator.
kernel/power/process.c | 25 +++++++++++++++++++++++++ 1 files changed, 25 insertions(+), 0 deletions(-)
kernel/power/process.c +25 -0
@@ -250,6 +250,31 @@ static int num_uninterruptible(int do_alreturn count; } +/* +* Tell threads of the type to enter the freezer. +*/ +static void signal_threads (int do_all_threads) +{ +struct task_struct *g, *p; +struct notifier_block *n; + +read_lock (&tasklist_lock); + do_each_thread(g, p) { +if (!freezeable (p, do_all_threads)) +continue ; + + n =kmalloc (sizeof (struct notifier_block ), +GFP_ATOMIC ); + +if (n) { + n->notifier_call = freeze_process; + n->priority =0 ; +notifier_chain_register (&p->todo, n); + } + }while_each_thread (g, p); +read_unlock (&tasklist_lock); +} +static inline void freeze (struct task_struct *p) {unsigned long flags;
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 17/23 Helper to prod processes that should have frozen but haven't.
In reply to Nigel CunninghamAdd a helper which prods processes of the type we are currently trying to freeze, seeking to get them to run their todo lists.
kernel/power/process.c | 23 +++++++++++++++++++++++ 1 files changed, 23 insertions(+), 0 deletions(-)
kernel/power/process.c +23 -0
@@ -275,6 +275,29 @@ static void signal_threads(int do_all_thread_unlock (&tasklist_lock); } +/* +* Prod processes that haven't entered the refrigerator yet. +*/ +static void prod_processes (int do_all_threads) +{ +struct task_struct *g, *p; +unsigned long flags; + +read_lock (&tasklist_lock); + do_each_thread(g, p) { +if (!freezeable (p, do_all_threads)) +continue ; + +spin_lock_irqsave (&p->sighand->siglock, flags); +if (!(p->flags &PF_FROZEN )) { +recalc_sigpending (); +signal_wake_up (p,0 ); + } +spin_unlock_irqrestore (&p->sighand->siglock, flags); + }while_each_thread (g, p); +read_unlock (&tasklist_lock); +} +static inline void freeze (struct task_struct *p) {unsigned long flags;
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 18/23 Helper: Did we fail to freeze all threads of a type?
In reply to Nigel CunninghamAdd a helper which checks whether we failed to freeze all the processes of a given type (ignoring uninterruptible threads). If such processes are found, we notify the user.
kernel/power/process.c | 40 ++++++++++++++++++++++++++++++++++++++++ 1 files changed, 40 insertions(+), 0 deletions(-)
kernel/power/process.c +40 -0
@@ -298,6 +298,46 @@ static void prod_processes(int do_all_thread_unlock (&tasklist_lock); } +/* +* Freezer failure. +* +* Check whether we failed to freeze all the processes that +* should be frozen. If we find a task that failed to freeze, +* we give useful information on what failed and how. +*/ +static int freezer_failure (int do_all_threads) +{ +int result =0 ; +struct task_struct *g, *p; + +read_lock (&tasklist_lock); + do_each_thread(g, p) { +if (!freezeable (p, do_all_threads) || + p->state ==TASK_UNINTERRUPTIBLE ) +continue ; + +if (!result) { +printk (KERN_ERR "Stopping tasks failed.\n" ); +printk (KERN_ERR "Tasks that refused to be " +"refrigerated and haven't since exited:\n" ); +set_freezer_state (ABORT_FREEZING ); + result =1 ; + } + +if ((freezing (p))) { +printk (" - %s (#%d) signalled but " +"didn't enter refrigerator.\n" , + p->comm, p->pid); + }else +printk (" - %s (#%d) signalled " +"and todo list empty.\n" , + p->comm, p->pid); + }while_each_thread (g, p); +read_unlock (&tasklist_lock); + +return result; +} +static inline void freeze (struct task_struct *p) {unsigned long flags;
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 19/23 Freeze all threads of a type.
In reply to Nigel CunninghamAdd a helper, which seeks to freeze all threads of a given type. If the type is kernel threads, it first freezes filesystems. Separate timeouts are used for periodically prodding processes that haven't yet entered the refrigerator, and for the timeout after which we consider failure to have occured.
kernel/power/process.c | 65 ++++++++++++++++++++++++++++++------------------ 1 files changed, 41 insertions(+), 24 deletions(-)
kernel/power/process.c +41 -24
@@ -40,11 +40,9 @@ #define freezer_message(msg, a...) do { } while(0)#endif -/* -* Timeout for stopping processes -*/ -#define TIMEOUT (6 * HZ) - +/* Timeouts when freezing */ +#define FREEZER_TOTAL_TIMEOUT (5 * HZ) +#define FREEZER_CHECK_TIMEOUT (HZ / 10)DECLARE_COMPLETION (kernelspace_thaw);DECLARE_COMPLETION (userspace_thaw);
@@ -338,30 +336,49 @@ static int freezer_failure(int do_all_threturn result; } -static inline void freeze (struct task_struct *p) +/* +* freeze_threads +* +* Freeze a set of threads having particular attributes. +* +* Types: +* 2: User threads. +* 3: Kernel threads. +*/ +static int freeze_threads (int do_all_threads) { -unsigned long flags; +int result =0 , still_to_do; +unsigned long start_time = jiffies; + +if (do_all_threads) +freezer_make_fses_ro (); + +signal_threads (do_all_threads); + +/* Watch them do it, wake them if they ignore us. */ +do { +prod_processes (do_all_threads); + +set_task_state (current,TASK_INTERRUPTIBLE ); +schedule_timeout (FREEZER_CHECK_TIMEOUT ); + + still_to_do =num_freezeable (do_all_threads) - +num_uninterruptible (do_all_threads); + + }while (still_to_do && (!test_freezer_state (ABORT_FREEZING )) && + !time_after (jiffies, start_time +FREEZER_TOTAL_TIMEOUT ));/* -* We only freeze if the todo list is empty to avoid -* putting multiple freeze handlers on the todo list. +* Did we time out? See if we failed to freeze processes as well. +* */ -if (!p->todo) { -struct notifier_block *n; +if ((time_after (jiffies, start_time +FREEZER_TOTAL_TIMEOUT )) + && (still_to_do)) + result =freezer_failure (do_all_threads); - n =kmalloc (sizeof (struct notifier_block ), -GFP_ATOMIC ); -if (n) { - n->notifier_call = freeze_process; - n->priority =0 ; -notifier_chain_register (&p->todo, n); - } - } -/* Make the process work on its todo list */ -spin_lock_irqsave (&p->sighand->siglock, flags); -recalc_sigpending (); -signal_wake_up (p,0 ); -spin_unlock_irqrestore (&p->sighand->siglock, flags); +BUG_ON (in_atomic ()); + +return 0 ; }/* 0 = success, else # of processes that we failed to stop */
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 20/23 Modify process.c includes and export freezer state.
In reply to Nigel CunninghamModify kernel/power/process #includes and add the exported freezer_state variable. (The symbol is exported so that modules can use the freezer_is_on() macro).
kernel/power/process.c | 10 ++++++++-- 1 files changed, 8 insertions(+), 2 deletions(-)
kernel/power/process.c +8 -2
@@ -28,10 +28,15 @@#undef DEBUG -#include <linux/smp_lock.h> -#include <linux/interrupt.h> #include <linux/suspend.h> +#include <linux/freezer.h> #include <linux/module.h> +#include <linux/mount.h> +#include <linux/namespace.h> +#include <linux/buffer_head.h> +#include <asm/tlbflush.h> + +unsigned long freezer_state =0 ;#if 0 //#ifdef CONFIG_PM_DEBUG
@@ -420,3 +425,4 @@ int freeze_processes(void)return 0 ; } +EXPORT_SYMBOL (freezer_state);
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 21/23 Remove unused DEBUG undef.
In reply to Nigel CunninghamRemove a now-unneeded #undef DEBUG.
kernel/power/process.c | 3 --- 1 files changed, 0 insertions(+), 3 deletions(-)
kernel/power/process.c +0 -3
@@ -25,9 +25,6 @@* to complete their work more quickly. */ - -#undef DEBUG -#include <linux/suspend.h> #include <linux/freezer.h> #include <linux/module.h>
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 22/23 Modify swsusp to thaw kernel threads while eating memory.
In reply to Nigel CunninghamModify swsusp so that while trying to eat memory, it allows kernel threads to run. This avoids a deadlock that could otherwise occur if access to a filesystem is needed while freeing the memory.
kernel/power/main.c | 4 ++++ kernel/power/swsusp.c | 3 +++ 2 files changed, 7 insertions(+), 0 deletions(-)
kernel/power/main.c +4 -0
@@ -72,6 +72,8 @@ static int suspend_prepare(suspend_state goto Thaw; } +thaw_processes (FREEZER_KERNEL_THREADS ); +if ((free_pages =nr_free_pages ()) <FREE_PAGE_NUMBER ) {pr_debug ("PM: free some memory\n" );shrink_all_memory (FREE_PAGE_NUMBER - free_pages);
@@ -82,6 +84,8 @@ static int suspend_prepare(suspend_state } } +freeze_processes (); +if (pm_ops->prepare) {if ((error = pm_ops->prepare (state))) goto Thaw;
kernel/power/swsusp.c +3 -0
@@ -575,6 +575,8 @@ int swsusp_shrink_memory(void)unsigned int i =0 ;char *p ="-\\|/" ; +thaw_processes (FREEZER_KERNEL_THREADS ); +printk ("Shrinking memory... " ); do { size = 2 * count_highmem_pages();
@@ -598,6 +600,7 @@ int swsusp_shrink_memory(void) }while (tmp >0 );printk ("\bdone (%lu pages freed)\n" , pages); +freeze_processes ();return 0 ; }
--
Nigel Cunningham nigel at suspend2 dot net
PATCH 23/23 Don't scan LRU while freezer is on.
In reply to Nigel CunninghamStop processes from initiating scanning of the LRU while the freezer is on. This is primarily for the benefit of Suspend2, which could conceivably trigger LRU scanning through this path while writing the image, and would thus generate an inconsistent image and through it all sorts of trouble.
mm/page_alloc.c | 5 +++-- 1 files changed, 3 insertions(+), 2 deletions(-)
mm/page_alloc.c +3 -2
@@ -25,6 +25,7 @@#include <linux/kernel.h> #include <linux/module.h> #include <linux/suspend.h> +#include <linux/freezer.h> #include <linux/pagevec.h> #include <linux/blkdev.h> #include <linux/slab.h>
@@ -957,8 +958,8 @@ restart:/* This allocation should allow future memory freeing. */ -if (((p->flags &PF_MEMALLOC ) ||unlikely (test_thread_flag (TIF_MEMDIE ))) - && !in_interrupt ()) { +if ((((p->flags &PF_MEMALLOC ) ||unlikely (test_thread_flag (TIF_MEMDIE ))) && + !in_interrupt ()) || (test_freezer_state (FREEZER_ON ))) {if (!(gfp_mask & __GFP_NOMEMALLOC)) { nofail_alloc:/* go through the zonelist yet again, ignoring mins */
--
Nigel Cunningham nigel at suspend2 dot net
Hi!
Quoted 4 lines
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
Quoted 8 lines
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).
Exactly. 6 seconds should be enough for signal to be delivered.
Quoted 5 lines
- The use of bdev freezing to ensure filesystems are properly frozen, thereby increasing the integrity of on-disk data in the case where a resume doesn't occur. This is also helpful in the case of Suspend2, where we don't atomically copy all memory, instead writing LRU pages separately.
We have talked about this before... we alredy have sync there, and that's enough. Your patch only makes code more complex for no gain.
Pavel
Signature · 1 line
Thanks, Sharp!
Hi.
Quoted 17 lines · Pavel Machek
Hi!
Quoted 4 lines
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
Quoted 8 lines
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).Exactly. 6 seconds should be enough for signal to be delivered.
Quoted 8 lines
Quoted 5 lines
- The use of bdev freezing to ensure filesystems are properly frozen, thereby increasing the integrity of on-disk data in the case where a resume doesn't occur. This is also helpful in the case of Suspend2, where we don't atomically copy all memory, instead writing LRU pages separately.
We have talked about this before... we alredy have sync there, and that's enough. Your patch only makes code more complex for no gain.
Yes, we have. Sorry for failing to back up my statement of the need for the changes with numbers. I'll do a couple of hundred attempts under load with and without the changes, and send a further message saying how many times the current implementation fails.
Regards,
Nigel
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
Hi,
Quoted 6 lines · Nigel Cunningham
Hi everyone.
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
Quoted 8 lines
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).
Could you please describe specific situation?
Quoted 5 lines
- The use of bdev freezing to ensure filesystems are properly frozen, thereby increasing the integrity of on-disk data in the case where a resume doesn't occur. This is also helpful in the case of Suspend2, where we don't atomically copy all memory, instead writing LRU pages separately.
Is this also needed when we do atomically copy all memory?
Greetings,
Rafael
Hi.
Quoted 18 lines · Rafael J. Wysocki
Hi,
Quoted 6 lines · Nigel Cunningham
Hi everyone.
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
Quoted 8 lines
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).
Quoted 1 line
Could you please describe specific situation?
The simplest example would be:
dd if=/dev/hda of=/dev/null
echo disk > /sys/power/state
Quoted 7 lines
Quoted 5 lines
- The use of bdev freezing to ensure filesystems are properly frozen, thereby increasing the integrity of on-disk data in the case where a resume doesn't occur. This is also helpful in the case of Suspend2, where we don't atomically copy all memory, instead writing LRU pages separately.
Is this also needed when we do atomically copy all memory?
I'm not thawing the bdevs until the end of resuming, so no.
Regards,
Nigel
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
Hi,
Quoted 17 lines · Nigel Cunningham
Quoted 16 lines · Rafael J. Wysocki
Quoted 14 lines · Nigel Cunningham
Hi everyone.
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).
Quoted 6 lines
Quoted 1 line
Could you please describe specific situation?
The simplest example would be:
dd if=/dev/hda of=/dev/null
echo disk >/sys/power/state
Well, I don't think it's a usual kind of workload. :-)
Anyway, could you please give some details? I mean how exactly your patch helps in this particular case?
Greetings,
Rafael
Hi.
Quoted 21 lines · Rafael J. Wysocki
Hi,
Quoted 18 lines · Nigel Cunningham
Quoted 17 lines · Rafael J. Wysocki
Quoted 15 lines · Nigel Cunningham
Hi everyone.
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Could you please describe specific situation?
The simplest example would be:
dd if=/dev/hda of=/dev/null
echo disk >/sys/power/stateWell, I don't think it's a usual kind of workload. :-)
No, but I/O alone shouldn't have such effect.
Quoted 2 lines
Anyway, could you please give some details? I mean how exactly your patch helps in this particular case?
I thought I did :). Freezing userspace first means the dd thread gets stopped first. Once the dd thread is stopped, the kernel threads processing the I/O requests have a finite amount of work to do (instead of having new work being submitted all the time), and can thus complete that and then be frozen in a far more deterministic fashion.
Regarding the stats I promised to Pavel, I'm heading home from LCA today, so I probably won't get them prepared until Monday now - unless I get lazy and only do 10 attempts instead of 100 :)
Regards,
Nigel
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
Quoted 44 lines · Nigel Cunningham
Hi.
Quoted 21 lines · Rafael J. Wysocki
Hi,
Quoted 18 lines · Nigel Cunningham
Quoted 17 lines · Rafael J. Wysocki
Quoted 15 lines · Nigel Cunningham
Hi everyone.
This set of patches represents the freezer upgrade patches from Suspend2.
The key features of this changeset are:
- Use of Christoph Lameter's todo list notifiers, which help with SMP cleanness.
- Splitting the freezing of kernel and userspace processes. Freezing currently suffers from a race because userspace processes can be submitting work for kernel threads, thereby stopping them from responding to freeze messages in a timely manner. The freezer can thus give up when it doesn't really need to. (This is not normally a problem only because load is not usually high).Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Could you please describe specific situation?
The simplest example would be:
dd if=/dev/hda of=/dev/null
echo disk >/sys/power/stateWell, I don't think it's a usual kind of workload. :-)
No, but I/O alone shouldn't have such effect.
Quoted 2 lines
Anyway, could you please give some details? I mean how exactly your patch helps in this particular case?
I thought I did :). Freezing userspace first means the dd thread gets stopped first. Once the dd thread is stopped, the kernel threads processing the I/O requests have a finite amount of work to do (instead of having new work being submitted all the time), and can thus complete that and then be frozen in a far more deterministic fashion.
Quoted 3 lines
Regarding the stats I promised to Pavel, I'm heading home from LCA today, so I probably won't get them prepared until Monday now - unless I get lazy and only do 10 attempts instead of 100 :)
You could swapoff -a to make your cycle a lot faster.... Freezing is still done in that case, IIRC.
Pavel
Signature · 1 line
Thanks, Sharp!
Hi.
Quoted 2 lines · Pavel Machek
You could swapoff -a to make your cycle a lot faster.... Freezing is still done in that case, IIRC.
That was right. I just did the test:
I used
stress -d 5 --hdd-bytes 100M -i 5 -c 5
to test.
In both cases, I:
1) booted to init S on an amd64 laptop with 1GB ram
2) for swsusp, swapoff'd /dev/hda1 and for suspend2 did echo 1 > /proc/suspend2/freezer_test
3) started stress
4) then started the attempts at suspending without restarting stress between each iteration.
I usually left a couple of seconds between each iteration to allow stress to fill buffers etc again, but was called away once during swsusp testing, leaving a much bigger period between 2 tests.
Kernel log output follows (sorry about the long lines):
Jan 30 15:33:08 localhost kernel: [ 372.681816] Stopping tasks: =================== Jan 30 15:33:08 localhost kernel: [ 378.714028] Restarting tasks...<6> Strange, kjournald not stopped Jan 30 15:33:08 localhost kernel: [ 378.734593] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 378.755084] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 378.775148] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 378.794802] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 378.813947] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 387.171737] done Jan 30 15:33:08 localhost kernel: [ 406.457958] Stopping tasks: ==========================| Jan 30 15:33:08 localhost kernel: [ 409.527817] Freeing memory... ^H-^H\^H|^H/^H-^H\^H|^H/^H-^H\^Hdone (130644 pages freed) Jan 30 15:33:08 localhost kernel: [ 410.276750] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 410.690283] ACPI: PCI Interrupt 0000:00:14.0[A] -> GSI 17 (level, low) -> IRQ 169 Jan 30 15:33:08 localhost kernel: [ 410.739835] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 410.758226] ACPI: PCI Interrupt 0000:00:1d.1[B] -> GSI 22 (level, low) -> IRQ 177 Jan 30 15:33:08 localhost kernel: [ 410.776764] pci_irq-0385 [03] pci_irq_derive : Unable to derive IRQ for device 0000:00:1f.0 Jan 30 15:33:08 localhost kernel: [ 410.795579] ACPI: PCI Interrupt 0000:00:1f.0[A]: no GSI Jan 30 15:33:08 localhost kernel: [ 411.472572] Restarting tasks... done Jan 30 15:33:08 localhost kernel: [ 489.731269] Stopping tasks: ======================= Jan 30 15:33:08 localhost kernel: [ 495.767731] Restarting tasks...<6> Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 495.786297] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 500.733572] done Jan 30 15:33:08 localhost kernel: [ 548.200154] Stopping tasks: ========================= Jan 30 15:33:08 localhost kernel: [ 554.229009] Restarting tasks...<6> Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 559.584968] done Jan 30 15:33:08 localhost kernel: [ 772.953163] Stopping tasks: ==========================| Jan 30 15:33:08 localhost kernel: [ 774.268679] Freeing memory... ^H-^H\^H|^H/^H-^H\^H|^H/^H-^H\^H|^H/^Hdone (157371 pages freed) Jan 30 15:33:08 localhost kernel: [ 774.948384] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 775.352905] ACPI: PCI Interrupt 0000:00:14.0[A] -> GSI 17 (level, low) -> IRQ 169 Jan 30 15:33:08 localhost kernel: [ 775.402457] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 775.421503] ACPI: PCI Interrupt 0000:00:1d.1[B] -> GSI 22 (level, low) -> IRQ 177 Jan 30 15:33:08 localhost kernel: [ 775.441055] pci_irq-0385 [03] pci_irq_derive : Unable to derive IRQ for device 0000:00:1f.0 Jan 30 15:33:08 localhost kernel: [ 775.461109] ACPI: PCI Interrupt 0000:00:1f.0[A]: no GSI Jan 30 15:33:08 localhost kernel: [ 776.140128] Restarting tasks... done Jan 30 15:33:08 localhost kernel: [ 793.729930] Stopping tasks: ================ Jan 30 15:33:08 localhost kernel: [ 799.765453] Restarting tasks...<6> Strange, kjournald not stopped Jan 30 15:33:08 localhost kernJan 30 15:33:08 localhost kernel: [ 825.247334] Stopping tasks: =============================| Jan 30 15:33:08 localhost kernel: [ 825.279168] Freeing memory... ^H-^H\^H|^Hdone (23358 pages freed) Jan 30 15:33:08 localhost kernel: [ 826.844573] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 827.409905] ACPI: PCI Interrupt 0000:00:14.0[A] -> GSI 17 (level, low) -> IRQ 169 Jan 30 15:33:08 localhost kernel: [ 827.459457] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 827.476078] ACPI: PCI Interrupt 0000:00:1d.1[B] -> GSI 22 (level, low) -> IRQ 177 Jan 30 15:33:08 localhost kernel: [ 827.492796] pci_irq-0385 [03] pci_irq_derive : Unable to derive IRQ for device 0000:00:1f.0 Jan 30 15:33:08 localhost kernel: [ 827.509962] ACPI: PCI Interrupt 0000:00:1f.0[A]: no GSI Jan 30 15:33:08 localhost kernel: [ 828.182831] Restarting tasks... done Jan 30 15:33:08 localhost kernel: [ 870.265642] Stopping tasks: ================ Jan 30 15:33:08 localhost kernel: [ 876.299090] Restarting tasks...<6> Strange, kjournald not stopped Jan 30 15:33:08 localhost kernel: [ 876.316040] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.332588] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.348638] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.364182] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.379489] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.394073] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.408172] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.422010] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.435853] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.449156] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.462449] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.475099] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 881.427744] done el: [ 799.785374] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 799.805440] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 799.825298] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 799.844855] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 799.863529] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 799.881792] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 799.899884] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 799.917305] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 799.934153] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 799.950444] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 805.247267] done Jan 30 15:33:08 localhost kernel: [ 825.247334] Stopping tasks: =============================| Jan 30 15:33:08 localhost kernel: [ 825.279168] Freeing memory... ^H-^H\^H|^Hdone (23358 pages freed) Jan 30 15:33:08 localhost kernel: [ 826.844573] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 827.409905] ACPI: PCI Interrupt 0000:00:14.0[A] -> GSI 17 (level, low) -> IRQ 169 Jan 30 15:33:08 localhost kernel: [ 827.459457] acpi_bus-0201 [02] bus_set_power : Device is not power manageable Jan 30 15:33:08 localhost kernel: [ 827.476078] ACPI: PCI Interrupt 0000:00:1d.1[B] -> GSI 22 (level, low) -> IRQ 177 Jan 30 15:33:08 localhost kernel: [ 827.492796] pci_irq-0385 [03] pci_irq_derive : Unable to derive IRQ for device 0000:00:1f.0 Jan 30 15:33:08 localhost kernel: [ 827.509962] ACPI: PCI Interrupt 0000:00:1f.0[A]: no GSI Jan 30 15:33:08 localhost kernel: [ 828.182831] Restarting tasks... done Jan 30 15:33:08 localhost kernel: [ 870.265642] Stopping tasks: ================ Jan 30 15:33:08 localhost kernel: [ 876.299090] Restarting tasks...<6> Strange, kjournald not stopped Jan 30 15:33:08 localhost kernel: [ 876.316040] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.332588] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.348638] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.364182] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.379489] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.394073] Strange, stress not stopped Jan 30 15:33:08 localhost kernel: [ 876.408172] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.422010] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.435853] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.449156] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.462449] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 876.475099] Strange, pdflush not stopped Jan 30 15:33:08 localhost kernel: [ 881.427744] done
Then with the new version:
[ 86.975030] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 86.975035] suspend_userui: program not configured. suspend_userui disabled. [ 86.975223] Freezing processes [ 101.101759] Preparing Image. [ 101.960220] Suspend2 debugging info: [ 101.960223] - SUSPEND core : 2.2.0.1 [ 101.960224] - Kernel Version : 2.6.15 [ 101.960225] - Compiler vers. : 4.0 [ 101.960227] - Attempt number : 1 [ 101.960228] - Parameters : 0 128 0 0 0 0 [ 101.960229] - Overall expected compression percentage: 0. [ 101.960231] - Swapwriter active. [ 101.960232] Swap available for image: 295184 pages. [ 101.960234] - Filewriter inactive. [ 101.960235] - No I/O speed stats available. [ 113.041463] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 113.041470] suspend_userui: program not configured. suspend_userui disabled. [ 113.041557] Freezing processes [ 131.156259] Preparing Image. [ 132.038683] Suspend2 debugging info: [ 132.038685] - SUSPEND core : 2.2.0.1 [ 132.038687] - Kernel Version : 2.6.15 [ 132.038688] - Compiler vers. : 4.0 [ 132.038689] - Attempt number : 2 [ 132.038690] - Parameters : 0 128 0 0 0 0 [ 132.038692] - Overall expected compression percentage: 0. [ 132.038693] - Swapwriter active. [ 132.038694] Swap available for image: 295184 pages. [ 132.038696] - Filewriter inactive. [ 132.038697] - No I/O speed stats available. [ 145.228742] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 145.228747] suspend_userui: program not configured. suspend_userui disabled. [ 145.228974] Freezing processes [ 152.404097] Preparing Image. [ 153.366314] Suspend2 debugging info: [ 153.366316] - SUSPEND core : 2.2.0.1 [ 153.366317] - Kernel Version : 2.6.15 [ 153.366319] - Compiler vers. : 4.0 [ 153.366320] - Attempt number : 3 [ 153.366321] - Parameters : 0 128 0 0 0 0 [ 153.366323] - Overall expected compression percentage: 0. [ 153.366324] - Swapwriter active. [ 153.366325] Swap available for image: 295184 pages. [ 153.366327] - Filewriter inactive. [ 153.366328] - No I/O speed stats available. [ 163.088026] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 163.088032] suspend_userui: program not configured. suspend_userui disabled. [ 163.088217] Freezing processes [ 179.203517] Preparing Image. [ 180.083987] Suspend2 debugging info: [ 180.083989] - SUSPEND core : 2.2.0.1 [ 180.083990] - Kernel Version : 2.6.15 [ 180.083992] - Compiler vers. : 4.0 [ 180.083993] - Attempt number : 4 [ 180.083994] - Parameters : 0 128 0 0 0 0 [ 180.083996] - Overall expected compression percentage: 0. [ 180.083997] - Swapwriter active. [ 180.083998] Swap available for image: 295184 pages. [ 180.084000] - Filewriter inactive. [ 180.084001] - No I/O speed stats available. [ 190.528302] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 190.528308] suspend_userui: program not configured. suspend_userui disabled. [ 190.528461] Freezing processes [ 198.554222] Preparing Image. [ 199.473349] Suspend2 debugging info: [ 199.473351] - SUSPEND core : 2.2.0.1 [ 199.473353] - Kernel Version : 2.6.15 [ 199.473354] - Compiler vers. : 4.0 [ 199.473355] - Attempt number : 5 [ 199.473357] - Parameters : 0 128 0 0 0 0 [ 199.473358] - Overall expected compression percentage: 0. [ 199.473360] - Swapwriter active. [ 199.473361] Swap available for image: 295184 pages. [ 199.473362] - Filewriter inactive. [ 199.473364] - No I/O speed stats available. [ 208.377977] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 208.377984] suspend_userui: program not configured. suspend_userui disabled. [ 208.378079] Freezing processes [ 227.420522] Preparing Image. [ 228.395082] Suspend2 debugging info: [ 228.395084] - SUSPEND core : 2.2.0.1 [ 228.395085] - Kernel Version : 2.6.15 [ 228.395087] - Compiler vers. : 4.0 [ 228.395088] - Attempt number : 6 [ 228.395089] - Parameters : 0 128 0 0 0 0 [ 228.395091] - Overall expected compression percentage: 0. [ 228.395092] - Swapwriter active. [ 228.395093] Swap available for image: 295184 pages. [ 228.395095] - Filewriter inactive. [ 228.395097] - No I/O speed stats available. [ 238.455606] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 238.455611] suspend_userui: program not configured. suspend_userui disabled. [ 238.455802] Freezing processes [ 245.563055] Preparing Image. [ 246.461738] Suspend2 debugging info: [ 246.461741] - SUSPEND core : 2.2.0.1 [ 246.461742] - Kernel Version : 2.6.15 [ 246.461743] - Compiler vers. : 4.0 [ 246.461745] - Attempt number : 7 [ 246.461746] - Parameters : 0 128 0 0 0 0 [ 246.461747] - Overall expected compression percentage: 0. [ 246.461749] - Swapwriter active. [ 246.461750] Swap available for image: 295184 pages. [ 246.461752] - Filewriter inactive. [ 255.808530] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 255.808537] suspend_userui: program not configured. suspend_userui disabled. [ 255.808785] Freezing processes [ 275.877164] Preparing Image. [ 276.277097] Losing some ticks... checking if CPU frequency changed. [ 276.705838] Suspend2 debugging info: [ 276.705841] - SUSPEND core : 2.2.0.1 [ 276.705842] - Kernel Version : 2.6.15 [ 276.705844] - Compiler vers. : 4.0 [ 276.705845] - Attempt number : 8 [ 276.705846] - Parameters : 0 128 0 0 0 0 [ 276.705848] - Overall expected compression percentage: 0. [ 276.705849] - Swapwriter active. [ 276.705850] Swap available for image: 295184 pages. [ 276.705852] - Filewriter inactive. [ 276.705853] - No I/O speed stats available. [ 288.497553] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 288.497559] suspend_userui: program not configured. suspend_userui disabled. [ 288.497791] Freezing processes [ 290.355243] Preparing Image. [ 291.262225] Suspend2 debugging info: [ 291.262228] - SUSPEND core : 2.2.0.1 [ 291.262229] - Kernel Version : 2.6.15 [ 291.262230] - Compiler vers. : 4.0 [ 291.262232] - Attempt number : 9 [ 291.262233] - Parameters : 0 128 0 0 0 0 [ 291.262234] - Overall expected compression percentage: 0. [ 291.262236] - Swapwriter active. [ 291.262237] Swap available for image: 295184 pages. [ 291.262239] - Filewriter inactive. [ 291.262240] - No I/O speed stats available. [ 309.866461] Suspend2 2.2.0.1: Initiating a software suspend cycle. [ 309.866467] suspend_userui: program not configured. suspend_userui disabled. [ 309.866655] Freezing processes [ 322.885993] Preparing Image. [ 323.860115] Suspend2 debugging info: [ 323.860118] - SUSPEND core : 2.2.0.1 [ 323.860119] - Kernel Version : 2.6.15 [ 323.860121] - Compiler vers. : 4.0 [ 323.860122] - Attempt number : 10 [ 323.860123] - Parameters : 0 128 0 0 0 0 [ 323.860125] - Overall expected compression percentage: 0. [ 323.860126] - Swapwriter active. [ 323.860128] Swap available for image: 295184 pages. [ 323.860129] - Filewriter inactive. [ 323.860130] - No I/O speed stats available.
Signature · 2 lines
See our web page for Howtos, FAQs, the Wiki and mailing list info. http://www.suspend2.net IRC: #suspend2 on Freenode
Quoted 6 lines · Nigel Cunningham
Hi.
Quoted 2 lines · Pavel Machek
You could swapoff -a to make your cycle a lot faster.... Freezing is still done in that case, IIRC.
Quoted 7 lines
That was right. I just did the test:
I used
stress -d 5 --hdd-bytes 100M -i 5 -c 5
to test.
Quite easy to reproduce, indeed. I'll try to find out if simple solution exists. (Notice that we still have that famous mysqld bug. I still have testcase somewhere).
Pavel
Signature · 1 line
Thanks, Sharp!
Quoted 10 lines · Rafael J. Wysocki
Hi,
Quoted 4 lines · Nigel Cunningham
The simplest example would be:
dd if=/dev/hda of=/dev/null
echo disk >/sys/power/stateWell, I don't think it's a usual kind of workload. :-)
Compiling a kernel, having updatedb and mandb run in the background and then trying to suspend while the compile still runs might do as well.
And this can happen, think of "battery-critical => suspend" setup. This is exactly the case i do not like the suspend to fail, because there might not be enough juice left to do a second try.
Signature · 3 lines
Stefan Seyfried \ "I didn't want to write for pay. I QA / R&D Team Mobile Devices \ wanted to be paid for what I write." SUSE LINUX Products GmbH, Nürnberg \ -- Leonard Cohen
Hi,
Quoted 8 lines · Stefan Seyfried
Quoted 7 lines · Rafael J. Wysocki
Quoted 4 lines · Nigel Cunningham
The simplest example would be:
dd if=/dev/hda of=/dev/null
echo disk >/sys/power/state
Quoted 8 lines
Quoted 1 line
Well, I don't think it's a usual kind of workload. :-)
Compiling a kernel, having updatedb and mandb run in the background and then trying to suspend while the compile still runs might do as well.
And this can happen, think of "battery-critical => suspend" setup. This is exactly the case i do not like the suspend to fail, because there might not be enough juice left to do a second try.
Still IMO we don't need all of the patches in the series to prevent this from happening and it's quite difficult to single out the relevant change.
Greetings,
Rafael