is there any Hard-disk shock-protection for 2.6.18 and above?
Subsystems (4)
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 3 lines · Jens Axboe
Quoted 2 lines · Pavel Machek
Hi!
[...]
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
After some googeling and digging in gamne i read that someone said that there are plans for some generic support for HD-parking in the kernel and thus making such patches obsolete.
[...]
Quoted 6 lines
Quoted 3 lines
I'm afraid we need your help with development here. Porting old patch to 2.6.19-rc6 should be easy, and then you can start 'how do I makethis generic' debate.
2.6.19 will finally have the generic block layer commands, so this can be implemented properly.
Eventually, I've ported the patch to 2.6.19-rc6 (attached). However, I'll need some gentle guidance by you developers for the next steps, I'm afraid. What exactly do you mean by "making this generic". Perhaps, you could give me a hint as to which generic block layer commands you have in mind that should be used in such a patch.
Here is a short description of what the patch does in its current shape:
1. Adds functions to ide-disk.c and scsi_lib.c that issue an idle immediate with head unload or a standby immediate command as appropriate and stop the queue on command completion.
2. Adds matching queue unfreeze functions that also reenable hd power management (see source code).
3. Adds an option to the ide-disk and libata module respectively which allows to indicate whether idle immediate is supported by the drive. Unfortunately, the implemented auto detection feature often gives false negatives.
4. Adds a sysfs attribute chich can be used to trigger the disk park command and the associated queue freeze.
5. Adds an unfreeze timer that calls the unfreeze function on expiry.
Regards,
Elias
--- block/ll_rw_blk.c | 154 ++++++++++++++++++++++++++++++++++ drivers/ata/libata-core.c | 4 drivers/ata/libata-scsi.c | 38 ++++++++ drivers/ata/libata.h | 1 drivers/ide/ide-disk.c | 155 ++++++++++++++++++++++++++++++++++ drivers/ide/ide-io.c | 14 +++ drivers/scsi/scsi_lib.c | 163 ++++++++++++++++++++++++++++++++++++ include/linux/ata.h | 1 include/linux/blkdev.h | 13 ++ include/linux/ide.h | 1 10 files changed, 544 insertions(+)
block/ll_rw_blk.c +154 -0
@@ -36,10 +36,14 @@ #include <scsi/scsi_cmnd.h>static void blk_unplug_work (void *data);static void blk_unplug_timeout (unsigned long data); +static void blk_unfreeze_work (void *data); +static void blk_unfreeze_timeout (unsigned long data);static void drive_stat_acct (struct request *rq,int nr_sectors,int new_io);static void init_request_from_bio (struct request *req,struct bio *bio);static int __make_request (request_queue_t *q,struct bio *bio);static struct io_context *current_io_context (gfp_t gfp_flags,int node); +static int blk_protect_register (request_queue_t *q); +static void blk_protect_unregister (request_queue_t *q);/* * For the allocated request tables
@@ -232,6 +236,13 @@ void blk_queue_make_request(request_queu q->unplug_timer.function = blk_unplug_timeout; q->unplug_timer.data = (unsigned long )q; + q->max_unfreeze =30 ; + +INIT_WORK (&q->unfreeze_work, blk_unfreeze_work, q); + + q->unfreeze_timer.function = blk_unfreeze_timeout; + q->unfreeze_timer.data = (unsigned long )q; +/* * by default assume old behaviour and bounce for any highmem page */
@@ -324,6 +335,18 @@ void blk_queue_issue_flush_fn(request_quEXPORT_SYMBOL (blk_queue_issue_flush_fn); +void blk_queue_issue_protect_fn (request_queue_t *q,issue_protect_fn *ipf) +{ + q->issue_protect_fn = ipf; +} +EXPORT_SYMBOL (blk_queue_issue_protect_fn); + +void blk_queue_issue_unprotect_fn (request_queue_t *q,issue_unprotect_fn *iuf) +{ + q->issue_unprotect_fn = iuf; +} +EXPORT_SYMBOL (blk_queue_issue_unprotect_fn); +/* * Cache flushing for ordered writes handling */
@@ -1842,6 +1865,7 @@ request_queue_t *blk_alloc_queue_node(gfmemset (q,0 ,sizeof (*q));init_timer (&q->unplug_timer); +init_timer (&q->unfreeze_timer);snprintf (q->kobj.name,KOBJ_NAME_LEN ,"%s" ,"queue" ); q->kobj.ktype = &queue_ktype;
@@ -3917,6 +3941,7 @@ int blk_register_queue(struct gendisk *dreturn ret; } +blk_protect_register (q);return 0 ; }
@@ -3925,6 +3950,7 @@ void blk_unregister_queue(struct gendiskrequest_queue_t *q = disk->queue;if (q && q->request_fn) { +blk_protect_unregister (q);elv_unregister_queue (q);kobject_uevent (&q->kobj,KOBJ_REMOVE );
@@ -3932,3 +3958,131 @@ void blk_unregister_queue(struct gendiskkobject_put (&disk->kobj); } } + +/* +* Issue lower level unprotect function if no timers are pending. +*/ +static void blk_unfreeze_work (void *data) +{ +request_queue_t *q = (request_queue_t *) data; +int pending; +unsigned long flags; + +spin_lock_irqsave (q->queue_lock, flags); + pending =timer_pending (&q->unfreeze_timer); +spin_unlock_irqrestore (q->queue_lock, flags); +if (!pending) + q->issue_unprotect_fn (q); +} + +/* +* Called when the queue freeze timeout expires... +*/ +static void blk_unfreeze_timeout (unsigned long data) +{ +request_queue_t *q = (request_queue_t *) data; + +kblockd_schedule_work (&q->unfreeze_work); +} + +/* +* The lower level driver parks and freezes the queue, and this block layer +* function sets up the freeze timeout timer on return. If the queue is +* already frozen then this is called to extend the timer... +*/ +void blk_freeze_queue (request_queue_t *q,int seconds) +{ +/* Don't accept arbitrarily long freezes */ +if (seconds >= q->max_unfreeze) + seconds = q->max_unfreeze; +/* set/reset the timer */ +mod_timer (&q->unfreeze_timer,msecs_to_jiffies (seconds*1000 ) + jiffies); +} + +/* +* When reading the 'protect' attribute, we return seconds remaining +* before unfreeze timeout expires +*/ +static ssize_t queue_protect_show (struct request_queue *q,char *page) +{ +unsigned int seconds =0 ; + +spin_lock_irq (q->queue_lock); +if (blk_queue_stopped (q) &&timer_pending (&q->unfreeze_timer)) +/* +* Adding 1 in order to guarantee nonzero value until timer +* has actually expired. +*/ + seconds =jiffies_to_msecs (q->unfreeze_timer.expires + - jiffies) /1000 +1 ; +spin_unlock_irq (q->queue_lock); +return queue_var_show (seconds, (page)); +} + +/* +* When writing the 'protect' attribute, input is the number of seconds +* to freeze the queue for. We call a lower level helper function to +* park the heads and freeze/block the queue, then we make a block layer +* call to setup the thaw timeout. If input is 0, then we thaw the queue. +*/ +static ssize_t queue_protect_store (struct request_queue *q,const char *page,size_t count) +{ +unsigned long freeze =0 ; + +queue_var_store (&freeze, page, count); + +if (freeze>0 ) { +/* Park and freeze */ +if (!blk_queue_stopped (q)) + q->issue_protect_fn (q); +/* set / reset the thaw timer */ +spin_lock_irq (q->queue_lock); +blk_freeze_queue (q, freeze); +spin_unlock_irq (q->queue_lock); + }else { +spin_lock_irq (q->queue_lock); + freeze =del_timer (&q->unfreeze_timer); +spin_unlock_irq (q->queue_lock); +if (freeze) + q->issue_unprotect_fn (q); + } + +return count; +} + +static struct queue_sysfs_entry queue_protect_entry = { + .attr = {.name ="protect" , .mode =S_IRUGO |S_IWUSR }, + .show = queue_protect_show, + .store = queue_protect_store, +}; + +static int blk_protect_register (request_queue_t *q) +{ +int error =0 ; + +/* check that the lower level driver has a protect handler */ +if (!q->issue_protect_fn) +return 1 ; + +/* create the attribute */ + error =sysfs_create_file (&q->kobj, &queue_protect_entry.attr); +if (error){ +printk (KERN_ERR +"blk_protect_register(): failed to create protect queue attribute!\n" ); +return error; + } + +kobject_get (&q->kobj); +return 0 ; +} + +static void blk_protect_unregister (request_queue_t *q) +{ +/* check that the lower level driver has a protect handler */ +if (!q->issue_protect_fn) +return ; + +/* remove the attribute */ +sysfs_remove_file (&q->kobj,&queue_protect_entry.attr); +kobject_put (&q->kobj); +}
drivers/ata/libata-core.c +4 -0
@@ -74,6 +74,10 @@ static struct workqueue_struct *ata_wq;struct workqueue_struct *ata_aux_wq; +int libata_protect_method =0 ; +module_param_named (protect_method, libata_protect_method, int,0444 ); +MODULE_PARM_DESC (protect_method,"hdaps disk protection method (0=autodetect, 1=unload, 2=standby)" ); +int atapi_enabled =1 ;module_param (atapi_enabled, int,0444 );MODULE_PARM_DESC (atapi_enabled,"Enable discovery of ATAPI devices (0=off, 1=on)" );
drivers/ata/libata-scsi.c +38 -0
@@ -841,6 +841,42 @@ static void ata_scsi_dev_config(struct s } } +extern int scsi_protect_queue (request_queue_t *q,int unload); +extern int scsi_unprotect_queue (request_queue_t *q); + +static int ata_scsi_issue_protect_fn (request_queue_t *q) +{ +struct scsi_device *sdev = q->queuedata; +struct ata_port *ap = (struct ata_port *) &sdev->host->hostdata[0 ]; +struct ata_device *dev = &ap->device[sdev->id]; +int unload; + +if (libata_protect_method ==1 ) { + unload =1 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload method requested, overriding drive capability check..\n" ); + } +else if (libata_protect_method ==2 ) { + unload =0 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): standby method requested, overriding drive capability check..\n" ); + } +else if (ata_id_has_unload (dev->id)) { + unload =1 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload support reported by drive..\n" ); + } +else { + unload =0 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload support NOT reported by drive!..\n" ); + } + +/* call scsi_protect_queue, requesting either unload or standby */ +return scsi_protect_queue (q, unload); +} + +static int ata_scsi_issue_unprotect_fn (request_queue_t *q) +{ +return scsi_unprotect_queue (q); +} +/** * ata_scsi_slave_config - Set SCSI device attributes * @sdev: SCSI device to examine
@@ -864,6 +900,8 @@ int ata_scsi_slave_config(struct scsi_deif (dev)ata_scsi_dev_config (sdev, dev); +blk_queue_issue_protect_fn (sdev->request_queue, ata_scsi_issue_protect_fn); +blk_queue_issue_unprotect_fn (sdev->request_queue, ata_scsi_issue_unprotect_fn);return 0 ;/* scsi layer doesn't check return value, sigh */ }
drivers/ata/libata.h +1 -0
@@ -40,6 +40,7 @@ struct ata_scsi_args {/* libata-core.c */ extern struct workqueue_struct *ata_aux_wq; +extern int libata_protect_method;extern int atapi_enabled;extern int atapi_dmadir;extern int libata_fua;
drivers/ide/ide-disk.c +155 -0
173 lines not on this page · raw message
drivers/ide/ide-io.c +14 -0
@@ -1261,6 +1261,17 @@ #endif }/* +* Don't accept a request when the queue is stopped (unless we +* are resuming from suspend). Prevents existing queue entries +* being processed after queue is stopped by the hard disk +* protection mechanism... +*/ +if (test_bit (QUEUE_FLAG_STOPPED , &drive->queue->queue_flags) && !blk_pm_resume_request (rq)) { + hwgroup->busy =0 ; +break ; + } + +/* * Sanity: don't accept a request that isn't a PM request * if we are currently power managed. This is very important as * blk_stop_queue() doesn't prevent the elv_next_request()
@@ -1744,6 +1755,9 @@ int ide_do_drive_cmd (ide_drive_t *drive where =ELEVATOR_INSERT_FRONT ; rq->cmd_flags |=REQ_PREEMPT ; } +if (action == ide_next) + where =ELEVATOR_INSERT_FRONT ; +__elv_add_request (drive->queue, rq, where,0 );ide_do_request (hwgroup,IDE_NO_IRQ );spin_unlock_irqrestore (&ide_lock, flags);
drivers/scsi/scsi_lib.c +163 -0
166 lines not on this page · raw message
include/linux/ata.h +1 -0
@@ -282,6 +282,7 @@ #define ata_id_is_sata(id) ((id)[93] ==#define ata_id_rahead_enabled (id) ((id)[85] & (1 << 6))#define ata_id_wcache_enabled (id) ((id)[85] & (1 << 5))#define ata_id_hpa_enabled (id) ((id)[85] & (1 << 10)) +#define ata_id_has_unload (id) ((id)[84] & (1 << 13))#define ata_id_has_fua (id) ((id)[84] & (1 << 6))#define ata_id_has_flush (id) ((id)[83] & (1 << 12))#define ata_id_has_flush_ext (id) ((id)[83] & (1 << 13))
include/linux/blkdev.h +13 -0
@@ -346,6 +346,8 @@ typedef void (activity_fn) (void *data,typedef int (issue_flush_fn ) (request_queue_t *,struct gendisk *,sector_t *);typedef void (prepare_flush_fn ) (request_queue_t *,struct request *);typedef void (softirq_done_fn )(struct request *); +typedef int (issue_protect_fn ) (request_queue_t *); +typedef int (issue_unprotect_fn ) (request_queue_t *);enum blk_queue_state { Queue_down,
@@ -388,6 +390,8 @@ struct request_queueissue_flush_fn *issue_flush_fn;prepare_flush_fn *prepare_flush_fn;softirq_done_fn *softirq_done_fn; +issue_protect_fn *issue_protect_fn; +issue_unprotect_fn *issue_unprotect_fn;/* * Dispatch queue sorting
@@ -403,6 +407,13 @@ struct request_queueunsigned long unplug_delay;/* After this many jiffies */ struct work_struct unplug_work; +/* +* Auto-unfreeze state +*/ +struct timer_list unfreeze_timer; +int max_unfreeze;/* At most this many seconds */ +struct work_struct unfreeze_work; +struct backing_dev_info backing_dev_info;/*
@@ -760,6 +771,8 @@ extern int blk_do_ordered(request_queue_extern unsigned blk_ordered_cur_seq (request_queue_t *);extern unsigned blk_ordered_req_seq (struct request *);extern void blk_ordered_complete_seq (request_queue_t *,unsigned ,int ); +extern void blk_queue_issue_protect_fn (request_queue_t *,issue_protect_fn *); +extern void blk_queue_issue_unprotect_fn (request_queue_t *,issue_unprotect_fn *);extern int blk_rq_map_sg (request_queue_t *,struct request *,struct scatterlist *);extern void blk_dump_rq_flags (struct request *,char *);
include/linux/ide.h +1 -0
@@ -1090,6 +1090,7 @@ extern u64 ide_get_error_location(ide_dr*/ typedef enum { ide_wait, /* insert rq at end of list, and wait for it */ + ide_next, /* insert rq immediately after current request */ ide_preempt, /* insert rq in front of current request */ ide_head_wait, /* insert rq in front of current request and wait for it */ ide_end /* insert rq at end of list, but don't wait for it */
Hi!
Quoted 18 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
After some googeling and digging in gamne i read that someone said that there are plans for some generic support for HD-parking in the kernel and thus making such patches obsolete.
[...]
Quoted 6 lines
Quoted 3 lines
I'm afraid we need your help with development here. Porting old patch to 2.6.19-rc6 should be easy, and then you can start 'how do I makethis generic' debate.
2.6.19 will finally have the generic block layer commands, so this can be implemented properly.
Eventually, I've ported the patch to 2.6.19-rc6 (attached). However, I'll need some gentle guidance by you developers for the next steps, I'm afraid. What exactly do you mean by "making this generic". Perhaps, you could give me a hint as to which generic block layer commands you have in mind that should be used in such a patch.
Quoted 6 lines
Here is a short description of what the patch does in its current shape:
1. Adds functions to
ide-disk.candscsi_lib.cthat issue an idle immediate with head unload or a standby immediate command as appropriate and stop the queue on command completion.
Can we get short Documentation/ patch?
block/ll_rw_blk.c void blk_unregister_queue(struct gendisk · line 3975
Quoted 2 lines
block/ll_rw_blk.c +2
-0+if (!pending) + q->issue_unprotect_fn (q);
Minor tab vs spaces problem here.
block/ll_rw_blk.c void blk_unregister_queue(struct gendisk · line 4019
Quoted 2 lines
block/ll_rw_blk.c +2
-0+spin_unlock_irq (q->queue_lock); +return queue_var_show (seconds, (page));
And here.
Review
Quoted 2 lines
(quoted diff) +2
-0+static ssize_t queue_protect_store(struct request_queue *q, const char *page, size_t count) +{
80 colums would be nice.
block/ll_rw_blk.c void blk_unregister_queue(struct gendisk · line 4034
Quoted 1 line
block/ll_rw_blk.c +1
-0+if (freeze>0 ) {
...and space between if and (
block/ll_rw_blk.c void blk_unregister_queue(struct gendisk · line 4054
Quoted 2 lines
block/ll_rw_blk.c +2
-0+static struct queue_sysfs_entry queue_protect_entry = { + .attr = {.name ="protect" , .mode =S_IRUGO |S_IWUSR },
And space between { and . .
block/ll_rw_blk.c void blk_unregister_queue(struct gendisk · line 4069
Quoted 3 lines
block/ll_rw_blk.c +3
-0+/* create the attribute */ + error =sysfs_create_file (&q->kobj, &queue_protect_entry.attr); +if (error){
if (error) {
drivers/ata/libata-core.c static struct workqueue_struct *ata_wq; · line 79
Quoted 2 lines
drivers/ata/libata-core.c +2
-0+module_param_named (protect_method, libata_protect_method, int,0444 ); +MODULE_PARM_DESC (protect_method,"hdaps disk protection method (0=autodetect, 1=unload, 2=standby)" );
Should this be configurable by module parameter? Why not tell each unload what to do?
Is /sys interface right thing to do?
drivers/ata/libata-scsi.c static void ata_scsi_dev_config(struct s · line 857
Quoted 4 lines
drivers/ata/libata-scsi.c +4
-0+if (libata_protect_method ==1 ) { + unload =1 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload method requested, overriding drive capability check..\n" ); + }
} and else on same line...
drivers/ata/libata-scsi.c static void ata_scsi_dev_config(struct s · line 869
Quoted 12 lines
drivers/ata/libata-scsi.c +12
-0+else if (libata_protect_method ==2 ) { + unload =0 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): standby method requested, overriding drive capability check..\n" ); + } +else if (ata_id_has_unload (dev->id)) { + unload =1 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload support reported by drive..\n" ); + } +else { + unload =0 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload support NOT reported by drive!..\n" ); + }
Can we consolidate the strings somehow?
drivers/ide/ide-disk.c #include <asm/uaccess.h> · line 78
Quoted 10 lines
drivers/ide/ide-disk.c +4
-0@@ -72,6 +72,10 @@ #include <asm/uaccess.h>#include <asm/io.h> #include <asm/div64.h> +int idedisk_protect_method =0 ; +module_param_named (protect_method, idedisk_protect_method, int,0444 ); +MODULE_PARM_DESC (protect_method,"hdaps disk protection method (0=autodetect, 1=unload, 2=standby)" ); +
Oh and do not mention hdaps, there are more different accelerometer types.
drivers/ide/ide-disk.c #include <asm/uaccess.h>
Quoted 9 lines
drivers/ide/ide-disk.c +7
-0+/* +* Auto-unfreeze state +*/ +struct timer_list unfreeze_timer; +int max_unfreeze;/* At most this many seconds */ +struct work_struct unfreeze_work; +struct backing_dev_info backing_dev_info;
Should we have kernel doing auto-unfreeze? Perhaps we can just mlock() the daemon?
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Quoted 2 lines · Pavel Machek
Should we have kernel doing auto-unfreeze? Perhaps we can just
mlock()the daemon?
You could be in the middle of suspend with by-now-frozen userspace; or maybe the daemon had a SEGV or was accidentally killed. Can't trust that.
Shem
Hi Pavel,
thanks a lot for your first review. See comments below.
Quoted 2 lines · Pavel Machek
Hi!
[...]
Quoted 8 lines
Quoted 6 lines
Here is a short description of what the patch does in its current shape:
1. Adds functions to
ide-disk.candscsi_lib.cthat issue an idle immediate with head unload or a standby immediate command as appropriate and stop the queue on command completion.Can we get short
Documentation/patch?
Sure. Would Documentation/block/disk-protection.txt be an appropriate location?
[...]
Quoted 5 lines
drivers/ata/libata-core.c static struct workqueue_struct *ata_wq; · line 79
Quoted 2 lines
drivers/ata/libata-core.c +2
-0+module_param_named (protect_method, libata_protect_method, int,0444 ); +MODULE_PARM_DESC (protect_method,"hdaps disk protection method (0=autodetect, 1=unload, 2=standby)" );Should this be configurable by module parameter? Why not tell each unload what to do?
As I understand, ATA specs expect drives to indicate whether they support the head unload feature of the idle immediate command or not. Unfortunately, a whole lot of them doesn't, well, mine doesn't anyway. Since I know that my drive does actually support head unloading, I'd like to tell the module so in order to prevent it from falling back to standby immediate. Applications that issue disk parking requests should not be bothered with this issue, in my opinion.
Quoted 2 lines
Is /sys interface right thing to do?
Probably, you're right here. Since this feature is actually drive specific, it should not really be set globally as a libata or ide-disk parameter but specifically for each drive connected. Perhaps we should add another attribute to /sys/block/*/queue or enhance the scope of /sys/block/*/queue/protect?
[...]
Quoted 6 lines
Quoted 6 lines
drivers/ata/libata-scsi.c +6
-0+else if (libata_protect_method ==2 ) { + unload =0 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): standby method requested, overriding drive capability check..\n" ); + } +else if (ata_id_has_unload (dev->id)) { + unload =1 ;
Quoted 8 lines
drivers/ata/libata-scsi.c static void ata_scsi_dev_config(struct s · line 869
Quoted 6 lines
drivers/ata/libata-scsi.c +6
-0+printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload support reported by drive..\n" ); + } +else { + unload =0 ; +printk (KERN_DEBUG "ata_scsi_issue_protect_fn(): unload support NOT reported by drive!..\n" ); + }Can we consolidate the strings somehow?
Actually, I'd like to move this to the initialisation sequence of the drive. I still have to figure out how to do this properly.
[...]
Quoted 5 lines
Quoted 5 lines
include/linux/blkdev.h +5
-0+/* +* Auto-unfreeze state +*/ +struct timer_list unfreeze_timer; +int max_unfreeze;/* At most this many seconds */
Quoted 8 lines
include/linux/blkdev.h struct request_queue · line 417
Quoted 4 lines
include/linux/blkdev.h +2
-0+struct work_struct unfreeze_work; +struct backing_dev_info backing_dev_info;Should we have kernel doing auto-unfreeze? Perhaps we can just
mlock()the daemon?
Pavel
I'd strongly second Shem's comments on this. Besides, anybody with root privileges can issue diks park requests, not just hdapsd. Please note that this is not a hard timeout as userspace can always refresh the timer before it has actually expired. On the other hand I would not want to rely on user space to unfreeze my drives.
Regards
Elias
Hi!
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
1. Adds functions to
ide-disk.candscsi_lib.cthat issue an idle immediate with head unload or a standby immediate command as appropriate and stop the queue on command completion.Can we get short
Documentation/patch?Sure. Would
Documentation/block/disk-protection.txtbe an appropriate location?
Yes.
Quoted 6 lines
Quoted 5 lines
drivers/ata/libata-core.c static struct workqueue_struct *ata_wq; · line 79
Quoted 2 lines
drivers/ata/libata-core.c +2
-0+module_param_named (protect_method, libata_protect_method, int,0444 ); +MODULE_PARM_DESC (protect_method,"hdaps disk protection method (0=autodetect, 1=unload, 2=standby)" );Should this be configurable by module parameter? Why not tell each unload what to do?
Quoted 7 lines
As I understand, ATA specs expect drives to indicate whether they support the head unload feature of the idle immediate command or not. Unfortunately, a whole lot of them doesn't, well, mine doesn't anyway. Since I know that my drive does actually support head unloading, I'd like to tell the module so in order to prevent it from falling back to standby immediate. Applications that issue disk parking requests should not be bothered with this issue, in my opinion.
What if you have two disks and one supports head unload and second does not?
Quoted 7 lines
Quoted 1 line
Is /sys interface right thing to do?
Probably, you're right here. Since this feature is actually drive specific, it should not really be set globally as a libata or ide-disk parameter but specifically for each drive connected. Perhaps we should add another attribute to
/sys/block/*/queueor enhance the scope of/sys/block/*/queue/protect?
Certainly better than current solution. Or maybe ioctl similar to wat hdparm uses?
Pavel
Signature · 1 line
Thanks for all the (sleeping) penguins.
Hi Pavel,
Quoted 5 lines · Pavel Machek
Quoted 5 lines
Quoted 5 lines
Review
Quoted 2 lines
(quoted diff) +2
-0+module_param_named(protect_method, libata_protect_method, int, 0444); +MODULE_PARM_DESC(protect_method, "hdaps disk protection method (0=autodetect, 1=unload, 2=standby)");Should this be configurable by module parameter? Why not tell each unload what to do?
[...]
Quoted 11 lines
Quoted 7 lines
Quoted 1 line
Is /sys interface right thing to do?
Probably, you're right here. Since this feature is actually drive specific, it should not really be set globally as a libata or ide-disk parameter but specifically for each drive connected. Perhaps we should add another attribute to
/sys/block/*/queueor enhance the scope of/sys/block/*/queue/protect?Certainly better than current solution. Or maybe ioctl similar to wat hdparm uses?
Pavel
I'm not quite sure what you have in mind wrt ioctls. I'm still convinced that the administrator should take a conscious decision when forcing an idle immediate with unload feature on a drive which doesn't announce this capability according to the specs. This is because I have no idea as to how drives might react if they don't support it. Perhaps we should consult linux-ide on this topic.
Anyway, this is the reason why I favour the sysfs approach. The decision can be made, for instance, in a udev rule during device setup.
So, here is a patch in which your remarks and suggestions have been incorporated. Additionally, I've added the requested kernel doc file and another sysfs attribute called protect_method. The usage of this attribute is described in Documentation/block/disk-protection.txt.
Patch applies to 2.6.19.
--- Documentation/block/disk-protection.txt | 79 +++++++ block/ll_rw_blk.c | 224 ++++++++++++++++++++++ drivers/ata/libata-scsi.c | 29 ++ drivers/ide/ide-disk.c | 142 +++++++++++++ drivers/ide/ide-io.c | 14 + drivers/scsi/scsi_lib.c | 163 ++++++++++++++++ include/linux/ata.h | 1 include/linux/blkdev.h | 14 + include/linux/ide.h | 1 9 files changed, 667 insertions(+)
Documentation/block/disk-protection.txt new file +79 -0
@@ -0,0 +1,79 @@ +Hard disk protection +==================== + + +Intro +----- +ATA/ATAPI-7 specifies the IDLE IMMEDIATE command with UNLOAD FEATURE. +Issuing this command should cause the drive to switch to idle mode and +unload disk heads. This feature is being used in modern laptops in +conjunction with accelerometers and appropriate software to implement +a shock protection facility. The idea is to stop all I/O operations on +the internal hard drive and park its heads on the ramp when critical +situations are anticipated. The desire to have such a feature +available on GNU/Linux systems has been the original motivation to +implement a generic disk parking interface in the Linux kernel. + + +The interface +------------- +The interface works as follows: Writing an integer value to +/sys/block/*/queue/protect will park the respective drive and freeze +the block layer queue for the specified number of seconds. When the +timeout expires and no further disk park request has been issued in +the meantime, the queue is unfrozen and accumulated I/O operations are +performed. + +IMPORTANT NOTE: +Not all ATA drives implement IDLE IMMEDIATE with UNLOAD FEATURE and +quite a few of those that do so, don't report this capability as +described in the specs. When a disk park has been requested through +sysfs as described above, the kernel will try to determine if the +drive supports the UNLOAD FEATURE by default. The kernel will only +rely on the IDLE IMMEDIATE with UNLOAD FEATURE command if it is +convinced that this command is actually supported by the disk drive; +otherwise, it will fall back to STANDBY IMMEDIATE. Resuming from the +latter will take much longer and it is generally more likely to have a +negative impact on the drive's lifetime due to the inclease of spin +down and up cycles. If you want to use this interface in a shock +protection framework and you know that your drive does indeed support +the IDLE IMMEDIATE with UNLOAD FEATURE command despite not saying so, +you can force the kernel to issue that command by doing the following +on the command line: +# echo -n unload > /sys/block/sda/queue/protect_method +(replace sda by the drive identifier as appropriate). + +/sys/block/*/queue/protect_method accepts auto, unload and standby +respectively. Reading from protect_method shows the available options +surrounding the active one with brackets. When auto is active, this +will change to whatever the kernel sees fit after the next disk park +command has been issued. + + +References +---------- + +There are several laptops from different brands featuring shock +protection capabilities. As manufacturers have refused to support open +source development of the required software components so far, Linux +support for shock protection varies considerably between different +hardware implementations. Ideally, this section should contain a list +of poiters at different projects aiming at an implementation of shock +protection on different systeems. Unfortunately, I only know of a +single project which, although still considered experimental, is fit +for use. Please feel free to add projects that have been the victims +of my ignorance. + +- http://www.thinkwiki.org/wiki/HDAPS + See this page for information about Linux support of the hard disk + active protection syystem as implemented in IBM/Lenovo Thinkpads. + + +CREDITS +------- + +The patch to implement the interface described in this file has +originally been published by Jon Escombe <[email protected]>. + + +05 Dec 2006, Elias Oltmanns <[email protected]>
block/ll_rw_blk.c +224 -0
267 lines not on this page · raw message
drivers/ata/libata-scsi.c +29 -0
@@ -841,6 +841,33 @@ static void ata_scsi_dev_config(struct s } } +extern int scsi_protect_queue (request_queue_t *q,int unload); +extern int scsi_unprotect_queue (request_queue_t *q); + +static int ata_scsi_issue_protect_fn (request_queue_t *q) +{ +struct scsi_device *sdev = q->queuedata; +struct ata_port *ap = (struct ata_port *) &sdev->host->hostdata[0 ]; +struct ata_device *dev = &ap->device[sdev->id]; +int unload = q->protect_method; +unsigned long flags; + +if (unload ==2 ) { + unload =ata_id_has_unload (dev->id) ?1 :0 ; +spin_lock_irqsave (q->queue_lock, flags); + q->protect_method = unload; +spin_unlock_irqrestore (q->queue_lock, flags); + } + +/* call scsi_protect_queue, requesting either unload or standby */ +return scsi_protect_queue (q, unload); +} + +static int ata_scsi_issue_unprotect_fn (request_queue_t *q) +{ +return scsi_unprotect_queue (q); +} +/** * ata_scsi_slave_config - Set SCSI device attributes * @sdev: SCSI device to examine
@@ -864,6 +891,8 @@ int ata_scsi_slave_config(struct scsi_deif (dev)ata_scsi_dev_config (sdev, dev); +blk_queue_issue_protect_fn (sdev->request_queue, ata_scsi_issue_protect_fn); +blk_queue_issue_unprotect_fn (sdev->request_queue, ata_scsi_issue_unprotect_fn);return 0 ;/* scsi layer doesn't check return value, sigh */ }
drivers/ide/ide-disk.c +142 -0
154 lines not on this page · raw message
drivers/ide/ide-io.c +14 -0
@@ -1261,6 +1261,17 @@ #endif }/* +* Don't accept a request when the queue is stopped (unless we +* are resuming from suspend). Prevents existing queue entries +* being processed after queue is stopped by the hard disk +* protection mechanism... +*/ +if (test_bit (QUEUE_FLAG_STOPPED , &drive->queue->queue_flags) && !blk_pm_resume_request (rq)) { + hwgroup->busy =0 ; +break ; + } + +/* * Sanity: don't accept a request that isn't a PM request * if we are currently power managed. This is very important as * blk_stop_queue() doesn't prevent the elv_next_request()
@@ -1744,6 +1755,9 @@ int ide_do_drive_cmd (ide_drive_t *drive where =ELEVATOR_INSERT_FRONT ; rq->cmd_flags |=REQ_PREEMPT ; } +if (action == ide_next) + where =ELEVATOR_INSERT_FRONT ; +__elv_add_request (drive->queue, rq, where,0 );ide_do_request (hwgroup,IDE_NO_IRQ );spin_unlock_irqrestore (&ide_lock, flags);
drivers/scsi/scsi_lib.c +163 -0
166 lines not on this page · raw message
include/linux/ata.h +1 -0
@@ -282,6 +282,7 @@ #define ata_id_is_sata(id) ((id)[93] ==#define ata_id_rahead_enabled (id) ((id)[85] & (1 << 6))#define ata_id_wcache_enabled (id) ((id)[85] & (1 << 5))#define ata_id_hpa_enabled (id) ((id)[85] & (1 << 10)) +#define ata_id_has_unload (id) ((id)[84] & (1 << 13))#define ata_id_has_fua (id) ((id)[84] & (1 << 6))#define ata_id_has_flush (id) ((id)[83] & (1 << 12))#define ata_id_has_flush_ext (id) ((id)[83] & (1 << 13))
include/linux/blkdev.h +14 -0
@@ -346,6 +346,8 @@ typedef void (activity_fn) (void *data,typedef int (issue_flush_fn ) (request_queue_t *,struct gendisk *,sector_t *);typedef void (prepare_flush_fn ) (request_queue_t *,struct request *);typedef void (softirq_done_fn )(struct request *); +typedef int (issue_protect_fn ) (request_queue_t *); +typedef int (issue_unprotect_fn ) (request_queue_t *);enum blk_queue_state { Queue_down,
@@ -388,6 +390,8 @@ struct request_queueissue_flush_fn *issue_flush_fn;prepare_flush_fn *prepare_flush_fn;softirq_done_fn *softirq_done_fn; +issue_protect_fn *issue_protect_fn; +issue_unprotect_fn *issue_unprotect_fn;/* * Dispatch queue sorting
@@ -403,6 +407,14 @@ struct request_queueunsigned long unplug_delay;/* After this many jiffies */ struct work_struct unplug_work; +/* +* Auto-unfreeze state +*/ +struct timer_list unfreeze_timer; +int max_unfreeze;/* At most this many seconds */ +struct work_struct unfreeze_work; +int protect_method; +struct backing_dev_info backing_dev_info;/*
@@ -760,6 +772,8 @@ extern int blk_do_ordered(request_queue_extern unsigned blk_ordered_cur_seq (request_queue_t *);extern unsigned blk_ordered_req_seq (struct request *);extern void blk_ordered_complete_seq (request_queue_t *,unsigned ,int ); +extern void blk_queue_issue_protect_fn (request_queue_t *,issue_protect_fn *); +extern void blk_queue_issue_unprotect_fn (request_queue_t *,issue_unprotect_fn *);extern int blk_rq_map_sg (request_queue_t *,struct request *,struct scatterlist *);extern void blk_dump_rq_flags (struct request *,char *);
include/linux/ide.h +1 -0
@@ -1090,6 +1090,7 @@ extern u64 ide_get_error_location(ide_dr*/ typedef enum { ide_wait, /* insert rq at end of list, and wait for it */ + ide_next, /* insert rq immediately after current request */ ide_preempt, /* insert rq in front of current request */ ide_head_wait, /* insert rq in front of current request and wait for it */ ide_end /* insert rq at end of list, but don't wait for it */
Hi Jens,
Quoted 4 lines · Elias Oltmanns
So, here is a patch in which your remarks and suggestions have been incorporated. Additionally, I've added the requested kernel doc file and another sysfs attribute called
protect_method. The usage of this attribute is described inDocumentation/block/disk-protection.txt.
Just forgot to mention that your suggestions haven't been implemented yet. Thats because I'm only gradually beginning to understand the reasoning and how it might work out in the end. It will probably take me another weekend (or more) to come up with something fit for discussion. Bare with me, I'm rather busy at the moment and still new to the block layer (or kernel code, for that matter).
Regards,
Elias
Quoted 8 lines · Elias Oltmanns
Hi Jens,
Quoted 4 lines · Elias Oltmanns
So, here is a patch in which your remarks and suggestions have been incorporated. Additionally, I've added the requested kernel doc file and another sysfs attribute called
protect_method. The usage of this attribute is described inDocumentation/block/disk-protection.txt.
Quoted 6 lines
Just forgot to mention that your suggestions haven't been implemented yet. Thats because I'm only gradually beginning to understand the reasoning and how it might work out in the end. It will probably take me another weekend (or more) to come up with something fit for discussion. Bare with me, I'm rather busy at the moment and still new to the block layer (or kernel code, for that matter).
No worries, take your time. The target for the code wont be 2.6.20 anyways, so you have at least a month to get things designed right and solid for a 2.6.21 target.
Signature · 1 line
Jens Axboe
Hi1
Quoted 17 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Review
Quoted 2 lines
(quoted diff) +2
-0+module_param_named(protect_method, libata_protect_method, int, 0444); +MODULE_PARM_DESC(protect_method, "hdaps disk protection method (0=autodetect, 1=unload, 2=standby)");Should this be configurable by module parameter? Why not tell each unload what to do?
[...]
Quoted 10 lines
Quoted 7 lines
Quoted 1 line
Is /sys interface right thing to do?
Probably, you're right here. Since this feature is actually drive specific, it should not really be set globally as a libata or ide-disk parameter but specifically for each drive connected. Perhaps we should add another attribute to
/sys/block/*/queueor enhance the scope of/sys/block/*/queue/protect?Certainly better than current solution. Or maybe ioctl similar to wat hdparm uses?
Quoted 6 lines
I'm not quite sure what you have in mind wrt ioctls. I'm still convinced that the administrator should take a conscious decision when forcing an idle immediate with unload feature on a drive which doesn't announce this capability according to the specs. This is because I have no idea as to how drives might react if they don't support it. Perhaps we should consult linux-ide on this topic.
Yep, I guess linux-ide would have some comments.
Quoted 2 lines
So, here is a patch in which your remarks and suggestions have been incorporated. Additionally, I've added the requested kernel doc file
Additional suggestion is to keep lines < 80 columns.... Sorry it took me so long to comment.
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Hi Pavel,
Quoted 19 lines · Pavel Machek
Hi1
Quoted 17 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Review
Quoted 2 lines
(quoted diff) +2
-0+module_param_named(protect_method, libata_protect_method, int, 0444); +MODULE_PARM_DESC(protect_method, "hdaps disk protection method (0=autodetect, 1=unload, 2=standby)");Should this be configurable by module parameter? Why not tell each unload what to do?
[...]
Quoted 10 lines
Quoted 7 lines
Quoted 1 line
Is /sys interface right thing to do?
Probably, you're right here. Since this feature is actually drive specific, it should not really be set globally as a libata or ide-disk parameter but specifically for each drive connected. Perhaps we should add another attribute to
/sys/block/*/queueor enhance the scope of/sys/block/*/queue/protect?Certainly better than current solution. Or maybe ioctl similar to wat hdparm uses?
Quoted 8 lines
Quoted 6 lines
I'm not quite sure what you have in mind wrt ioctls. I'm still convinced that the administrator should take a conscious decision when forcing an idle immediate with unload feature on a drive which doesn't announce this capability according to the specs. This is because I have no idea as to how drives might react if they don't support it. Perhaps we should consult linux-ide on this topic.
Yep, I guess linux-ide would have some comments.
At least, I can now see the benefits of an ioctl approach. Although the protect_method attribute allows to configure each drive individually, that is actually not quite what we want. Since this attribute is meant to deal with drives that don't comply strictly with the ATA specs, it is, at least, misleading to associate this with each queue. Instead, it would be desirable to offer this option to ATA drives only and to do this consistently, regardless whether the IDE or the ATA drivers are in use. So, I'm now quite convinced that ioctls are the easiest way to achieve just that - thanks for the hint.
Quoted 7 lines
Quoted 3 lines
So, here is a patch in which your remarks and suggestions have been incorporated. Additionally, I've added the requested kernel doc file
Additional suggestion is to keep lines < 80 columns.... Sorry it took me so long to comment.
Pavel
In the meantime, I've started to implement the generic block layer commands for queue freezing as Jens suggested in this thread. Unfortunately, I've been very busy lately and actually I still am but I'll try hard to get this finished rather sooner thn later.
Thanks for your support.
Elias
Quoted 13 lines · Elias Oltmanns
Quoted 3 lines · Jens Axboe
Quoted 2 lines · Pavel Machek
Hi!
[...]
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
After some googeling and digging in gamne i read that someone said that there are plans for some generic support for HD-parking in the kernel and thus making such patches obsolete.
[...]
Quoted 4 lines
Quoted 3 lines
I'm afraid we need your help with development here. Porting old patch to 2.6.19-rc6 should be easy, and then you can start 'how do I makethis generic' debate.
Quoted 8 lines
Quoted 2 lines
2.6.19 will finally have the generic block layer commands, so this can be implemented properly.
Eventually, I've ported the patch to 2.6.19-rc6 (attached). However, I'll need some gentle guidance by you developers for the next steps, I'm afraid. What exactly do you mean by "making this generic". Perhaps, you could give me a hint as to which generic block layer commands you have in mind that should be used in such a patch.
If you look in <linux/blkdev.h>, you can see the definition of REQ_TYPE_LINUX_BLOCK and the current sub commands (none are implemented right now). So you want to add REQ_LB_OPT_PARK (or whatever the name should be) to implement the freezing operation, and then add support for the drivers to understand this command and do what they need to do. You can add block layer helpers to perform the freezing of the actual queue, to be called when the PARK command completes. You can handle the queue draining for tagged devices like the barriers do for FUA barriers.
Signature · 1 line
Jens Axboe