* [PATCH 0/2] Introduce new raid0 state 'broken' @ 2019-07-29 20:31 Guilherme G. Piccoli 2019-07-29 20:31 ` [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 Guilherme G. Piccoli 2019-07-29 20:31 ` [PATCH 2/2] mdadm: " Guilherme G. Piccoli 0 siblings, 2 replies; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-07-29 20:31 UTC (permalink / raw) To: linux-raid Cc: linux-block, dm-devel, gpiccoli, jay.vosburgh, neilb, songliubraving Currently the md/raid0 device behaves quite differently of other block devices when it comes to failure. While other md levels contain vast logic to deal with failures, and other non-md devices like scsi disks or nvme rely on a dying queue when they fail, md/raid0 for instance does not signal failures if an array member is removed while the array is mounted; in that case, udev signals the device removal but mdadm cannot succeed in the STOP_ARRAY ioctl, since it's mounted. This behavior was tentatively changed in the past to match the scsi/nvme devices (see [0]), but this attempt was quite complex, it had some corner cases and (the few) community reviews weren't generally positive. So, we are trying again with a simpler approach this time. This series introduces a new array state 'broken' (for raid0 only), which mimics the state 'clean'. The main goal for this new state is a way to signal the user that something is wrong with the array. We also included a warn_once-style message in kernel log to alert the user when the array has one failed member. The series encompass changes in the kernel and in mdadm tool. To get the 'broken' state completely functional one requires both changes, but mdadm and kernel can live without their counterpart changes (in case some users gets an updated mdadm for example, but keeps using an old kernel). This series does not affect at all the way md/raid0 will react to I/O failures. It was discussed in [0] that it should be better if raid0 could fail faster in case it gets a member removed; we just proposed a change in that realm too (see [1]), but it seems better to have them reviewed/treated separately. This series was tested with raid0 arrays holding both an ext4 and xfs filesystems. Thanks in advance for the reviews/feedbacks. Cheers, Guilherme [0] lore.kernel.org/linux-block/20190418220448.7219-1-gpiccoli@canonical.com [1] lore.kernel.org/linux-block/20190729193359.11040-1-gpiccoli@canonical.com Guilherme G. Piccoli (1): md/raid0: Introduce new array state 'broken' for raid0 [kernel part] drivers/md/md.c | 23 +++++++++++++++++++---- drivers/md/md.h | 2 ++ drivers/md/raid0.c | 26 ++++++++++++++++++++++++++ 3 files changed, 47 insertions(+), 4 deletions(-) [mdadm] Detail.c | 16 ++++++++++++++-- Monitor.c | 9 +++++++-- maps.c | 1 + mdadm.h | 1 + mdmon.h | 2 +- monitor.c | 4 ++-- 6 files changed, 26 insertions(+), 7 deletions(-) -- 2.22.0 ^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-29 20:31 [PATCH 0/2] Introduce new raid0 state 'broken' Guilherme G. Piccoli @ 2019-07-29 20:31 ` Guilherme G. Piccoli 2019-07-30 0:11 ` NeilBrown ` (2 more replies) 2019-07-29 20:31 ` [PATCH 2/2] mdadm: " Guilherme G. Piccoli 1 sibling, 3 replies; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-07-29 20:31 UTC (permalink / raw) To: linux-raid Cc: linux-block, dm-devel, gpiccoli, jay.vosburgh, neilb, songliubraving Currently if a md/raid0 array gets one or more members removed while being mounted, kernel keeps showing state 'clean' in the 'array_state' sysfs attribute. Despite udev signaling the member device is gone, 'mdadm' cannot issue the STOP_ARRAY ioctl successfully, given the array is mounted. Nothing else hints that something is wrong (except that the removed devices don't show properly in the output of 'mdadm detail' command). There is no other property to be checked, and if user is not performing reads/writes to the array, even kernel log is quiet and doesn't give a clue about the missing member. This patch changes this behavior; when 'array_state' is read we introduce a non-expensive check (only for raid0) that relies in the comparison of the total number of disks when array was assembled with gendisk flags of those devices to validate if all members are available and functional. A new array state 'broken' was added: it mimics the state 'clean' in every aspect, being useful only to distinguish if such array has some member missing. Also, we show a rate-limited warning in kernel log in such case. This patch has no proper functional change other than adding a 'clean'-like state; it was tested with ext4 and xfs filesystems. It requires a 'mdadm' counterpart to handle the 'broken' state. Cc: NeilBrown <neilb@suse.com> Cc: Song Liu <songliubraving@fb.com> Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com> --- drivers/md/md.c | 23 +++++++++++++++++++---- drivers/md/md.h | 2 ++ drivers/md/raid0.c | 26 ++++++++++++++++++++++++++ 3 files changed, 47 insertions(+), 4 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index fba49918d591..b80f36084ec1 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -4160,12 +4160,18 @@ __ATTR_PREALLOC(resync_start, S_IRUGO|S_IWUSR, * active-idle * like active, but no writes have been seen for a while (100msec). * + * broken + * RAID0-only: same as clean, but array is missing a member. + * It's useful because RAID0 mounted-arrays aren't stopped + * when a member is gone, so this state will at least alert + * the user that something is wrong. + * */ enum array_state { clear, inactive, suspended, readonly, read_auto, clean, active, - write_pending, active_idle, bad_word}; + write_pending, active_idle, broken, bad_word}; static char *array_states[] = { "clear", "inactive", "suspended", "readonly", "read-auto", "clean", "active", - "write-pending", "active-idle", NULL }; + "write-pending", "active-idle", "broken", NULL }; static int match_word(const char *word, char **list) { @@ -4181,7 +4187,7 @@ array_state_show(struct mddev *mddev, char *page) { enum array_state st = inactive; - if (mddev->pers) + if (mddev->pers) { switch(mddev->ro) { case 1: st = readonly; @@ -4201,7 +4207,15 @@ array_state_show(struct mddev *mddev, char *page) st = active; spin_unlock(&mddev->lock); } - else { + + if ((mddev->pers->level == 0) && + ((st == clean) || (st == broken))) { + if (mddev->pers->is_missing_dev(mddev)) + st = broken; + else + st = clean; + } + } else { if (list_empty(&mddev->disks) && mddev->raid_disks == 0 && mddev->dev_sectors == 0) @@ -4315,6 +4329,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len) break; case write_pending: case active_idle: + case broken: /* these cannot be set */ break; } diff --git a/drivers/md/md.h b/drivers/md/md.h index 41552e615c4c..e7b42b75701a 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -590,6 +590,8 @@ struct md_personality int (*congested)(struct mddev *mddev, int bits); /* Changes the consistency policy of an active array. */ int (*change_consistency_policy)(struct mddev *mddev, const char *buf); + /* Check if there is any missing/failed members - RAID0 only for now. */ + bool (*is_missing_dev)(struct mddev *mddev); }; struct md_sysfs_entry { diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index 58a9cc5193bf..79618a6ae31a 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -455,6 +455,31 @@ static inline int is_io_in_chunk_boundary(struct mddev *mddev, } } +bool raid0_is_missing_dev(struct mddev *mddev) +{ + struct md_rdev *rdev; + static int already_missing; + int def_disks, work_disks = 0; + struct r0conf *conf = mddev->private; + + def_disks = conf->strip_zone[0].nb_dev; + rdev_for_each(rdev, mddev) + if (rdev->bdev->bd_disk->flags & GENHD_FL_UP) + work_disks++; + + if (unlikely(def_disks - work_disks)) { + if (!already_missing) { + already_missing = 1; + pr_warn("md: %s: raid0 array has %d missing/failed members\n", + mdname(mddev), (def_disks - work_disks)); + } + return true; + } + + already_missing = 0; + return false; +} + static void raid0_handle_discard(struct mddev *mddev, struct bio *bio) { struct r0conf *conf = mddev->private; @@ -789,6 +814,7 @@ static struct md_personality raid0_personality= .takeover = raid0_takeover, .quiesce = raid0_quiesce, .congested = raid0_congested, + .is_missing_dev = raid0_is_missing_dev, }; static int __init raid0_init (void) -- 2.22.0 ^ permalink raw reply related [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-29 20:31 ` [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 Guilherme G. Piccoli @ 2019-07-30 0:11 ` NeilBrown 2019-07-30 11:43 ` Guilherme G. Piccoli 2019-07-30 6:20 ` Bob Liu 2019-07-31 19:43 ` Song Liu 2 siblings, 1 reply; 13+ messages in thread From: NeilBrown @ 2019-07-30 0:11 UTC (permalink / raw) To: Guilherme G. Piccoli, linux-raid Cc: jay.vosburgh, songliubraving, dm-devel, Neil F Brown, linux-block [-- Attachment #1: Type: text/plain, Size: 5759 bytes --] On Mon, Jul 29 2019, Guilherme G. Piccoli wrote: > Currently if a md/raid0 array gets one or more members removed while > being mounted, kernel keeps showing state 'clean' in the 'array_state' > sysfs attribute. Despite udev signaling the member device is gone, 'mdadm' > cannot issue the STOP_ARRAY ioctl successfully, given the array is mounted. > > Nothing else hints that something is wrong (except that the removed devices > don't show properly in the output of 'mdadm detail' command). There is no > other property to be checked, and if user is not performing reads/writes > to the array, even kernel log is quiet and doesn't give a clue about the > missing member. > > This patch changes this behavior; when 'array_state' is read we introduce > a non-expensive check (only for raid0) that relies in the comparison of > the total number of disks when array was assembled with gendisk flags of > those devices to validate if all members are available and functional. > A new array state 'broken' was added: it mimics the state 'clean' in every > aspect, being useful only to distinguish if such array has some member > missing. Also, we show a rate-limited warning in kernel log in such case. > > This patch has no proper functional change other than adding a 'clean'-like > state; it was tested with ext4 and xfs filesystems. It requires a 'mdadm' > counterpart to handle the 'broken' state. > > Cc: NeilBrown <neilb@suse.com> > Cc: Song Liu <songliubraving@fb.com> > Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com> > --- > drivers/md/md.c | 23 +++++++++++++++++++---- > drivers/md/md.h | 2 ++ > drivers/md/raid0.c | 26 ++++++++++++++++++++++++++ > 3 files changed, 47 insertions(+), 4 deletions(-) > > diff --git a/drivers/md/md.c b/drivers/md/md.c > index fba49918d591..b80f36084ec1 100644 > --- a/drivers/md/md.c > +++ b/drivers/md/md.c > @@ -4160,12 +4160,18 @@ __ATTR_PREALLOC(resync_start, S_IRUGO|S_IWUSR, > * active-idle > * like active, but no writes have been seen for a while (100msec). > * > + * broken > + * RAID0-only: same as clean, but array is missing a member. > + * It's useful because RAID0 mounted-arrays aren't stopped > + * when a member is gone, so this state will at least alert > + * the user that something is wrong. > + * > */ > enum array_state { clear, inactive, suspended, readonly, read_auto, clean, active, > - write_pending, active_idle, bad_word}; > + write_pending, active_idle, broken, bad_word}; > static char *array_states[] = { > "clear", "inactive", "suspended", "readonly", "read-auto", "clean", "active", > - "write-pending", "active-idle", NULL }; > + "write-pending", "active-idle", "broken", NULL }; > > static int match_word(const char *word, char **list) > { > @@ -4181,7 +4187,7 @@ array_state_show(struct mddev *mddev, char *page) > { > enum array_state st = inactive; > > - if (mddev->pers) > + if (mddev->pers) { > switch(mddev->ro) { > case 1: > st = readonly; > @@ -4201,7 +4207,15 @@ array_state_show(struct mddev *mddev, char *page) > st = active; > spin_unlock(&mddev->lock); > } > - else { > + > + if ((mddev->pers->level == 0) && Don't test if ->level is 0. Instead, test if ->is_missing_dev is not NULL. NeilBrown > + ((st == clean) || (st == broken))) { > + if (mddev->pers->is_missing_dev(mddev)) > + st = broken; > + else > + st = clean; > + } > + } else { > if (list_empty(&mddev->disks) && > mddev->raid_disks == 0 && > mddev->dev_sectors == 0) > @@ -4315,6 +4329,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len) > break; > case write_pending: > case active_idle: > + case broken: > /* these cannot be set */ > break; > } > diff --git a/drivers/md/md.h b/drivers/md/md.h > index 41552e615c4c..e7b42b75701a 100644 > --- a/drivers/md/md.h > +++ b/drivers/md/md.h > @@ -590,6 +590,8 @@ struct md_personality > int (*congested)(struct mddev *mddev, int bits); > /* Changes the consistency policy of an active array. */ > int (*change_consistency_policy)(struct mddev *mddev, const char *buf); > + /* Check if there is any missing/failed members - RAID0 only for now. */ > + bool (*is_missing_dev)(struct mddev *mddev); > }; > > struct md_sysfs_entry { > diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c > index 58a9cc5193bf..79618a6ae31a 100644 > --- a/drivers/md/raid0.c > +++ b/drivers/md/raid0.c > @@ -455,6 +455,31 @@ static inline int is_io_in_chunk_boundary(struct mddev *mddev, > } > } > > +bool raid0_is_missing_dev(struct mddev *mddev) > +{ > + struct md_rdev *rdev; > + static int already_missing; > + int def_disks, work_disks = 0; > + struct r0conf *conf = mddev->private; > + > + def_disks = conf->strip_zone[0].nb_dev; > + rdev_for_each(rdev, mddev) > + if (rdev->bdev->bd_disk->flags & GENHD_FL_UP) > + work_disks++; > + > + if (unlikely(def_disks - work_disks)) { > + if (!already_missing) { > + already_missing = 1; > + pr_warn("md: %s: raid0 array has %d missing/failed members\n", > + mdname(mddev), (def_disks - work_disks)); > + } > + return true; > + } > + > + already_missing = 0; > + return false; > +} > + > static void raid0_handle_discard(struct mddev *mddev, struct bio *bio) > { > struct r0conf *conf = mddev->private; > @@ -789,6 +814,7 @@ static struct md_personality raid0_personality= > .takeover = raid0_takeover, > .quiesce = raid0_quiesce, > .congested = raid0_congested, > + .is_missing_dev = raid0_is_missing_dev, > }; > > static int __init raid0_init (void) > -- > 2.22.0 [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 832 bytes --] ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-30 0:11 ` NeilBrown @ 2019-07-30 11:43 ` Guilherme G. Piccoli 0 siblings, 0 replies; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-07-30 11:43 UTC (permalink / raw) To: NeilBrown, linux-raid Cc: jay.vosburgh, songliubraving, dm-devel, Neil F Brown, linux-block On 29/07/2019 21:11, NeilBrown wrote: > [...] >> - else { >> + >> + if ((mddev->pers->level == 0) && > > Don't test if ->level is 0. Instead, test if ->is_missing_dev is not > NULL. > > NeilBrown Hi Neil, thanks for the feedback. I'll change that in a potential V2, (if the patches are likely to be accepted), good idea. Cheers, Guilherme > > >> + ((st == clean) || (st == broken))) { >> + if (mddev->pers->is_missing_dev(mddev)) >> + st = broken; >> + else >> + st = clean; >> + } >> + } else { >> if (list_empty(&mddev->disks) && >> mddev->raid_disks == 0 && >> mddev->dev_sectors == 0) >> @@ -4315,6 +4329,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len) >> break; >> case write_pending: >> case active_idle: >> + case broken: >> /* these cannot be set */ >> break; >> } >> diff --git a/drivers/md/md.h b/drivers/md/md.h >> index 41552e615c4c..e7b42b75701a 100644 >> --- a/drivers/md/md.h >> +++ b/drivers/md/md.h >> @@ -590,6 +590,8 @@ struct md_personality >> int (*congested)(struct mddev *mddev, int bits); >> /* Changes the consistency policy of an active array. */ >> int (*change_consistency_policy)(struct mddev *mddev, const char *buf); >> + /* Check if there is any missing/failed members - RAID0 only for now. */ >> + bool (*is_missing_dev)(struct mddev *mddev); >> }; >> >> struct md_sysfs_entry { >> diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c >> index 58a9cc5193bf..79618a6ae31a 100644 >> --- a/drivers/md/raid0.c >> +++ b/drivers/md/raid0.c >> @@ -455,6 +455,31 @@ static inline int is_io_in_chunk_boundary(struct mddev *mddev, >> } >> } >> >> +bool raid0_is_missing_dev(struct mddev *mddev) >> +{ >> + struct md_rdev *rdev; >> + static int already_missing; >> + int def_disks, work_disks = 0; >> + struct r0conf *conf = mddev->private; >> + >> + def_disks = conf->strip_zone[0].nb_dev; >> + rdev_for_each(rdev, mddev) >> + if (rdev->bdev->bd_disk->flags & GENHD_FL_UP) >> + work_disks++; >> + >> + if (unlikely(def_disks - work_disks)) { >> + if (!already_missing) { >> + already_missing = 1; >> + pr_warn("md: %s: raid0 array has %d missing/failed members\n", >> + mdname(mddev), (def_disks - work_disks)); >> + } >> + return true; >> + } >> + >> + already_missing = 0; >> + return false; >> +} >> + >> static void raid0_handle_discard(struct mddev *mddev, struct bio *bio) >> { >> struct r0conf *conf = mddev->private; >> @@ -789,6 +814,7 @@ static struct md_personality raid0_personality= >> .takeover = raid0_takeover, >> .quiesce = raid0_quiesce, >> .congested = raid0_congested, >> + .is_missing_dev = raid0_is_missing_dev, >> }; >> >> static int __init raid0_init (void) >> -- >> 2.22.0 ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-29 20:31 ` [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 Guilherme G. Piccoli 2019-07-30 0:11 ` NeilBrown @ 2019-07-30 6:20 ` Bob Liu 2019-07-30 12:18 ` Guilherme G. Piccoli 2019-07-31 0:28 ` NeilBrown 2019-07-31 19:43 ` Song Liu 2 siblings, 2 replies; 13+ messages in thread From: Bob Liu @ 2019-07-30 6:20 UTC (permalink / raw) To: Guilherme G. Piccoli, linux-raid Cc: linux-block, dm-devel, jay.vosburgh, neilb, songliubraving On 7/30/19 4:31 AM, Guilherme G. Piccoli wrote: > Currently if a md/raid0 array gets one or more members removed while > being mounted, kernel keeps showing state 'clean' in the 'array_state' > sysfs attribute. Despite udev signaling the member device is gone, 'mdadm' > cannot issue the STOP_ARRAY ioctl successfully, given the array is mounted. > > Nothing else hints that something is wrong (except that the removed devices > don't show properly in the output of 'mdadm detail' command). There is no > other property to be checked, and if user is not performing reads/writes > to the array, even kernel log is quiet and doesn't give a clue about the > missing member. > > This patch changes this behavior; when 'array_state' is read we introduce > a non-expensive check (only for raid0) that relies in the comparison of > the total number of disks when array was assembled with gendisk flags of> those devices to validate if all members are available and functional. > A new array state 'broken' was added: it mimics the state 'clean' in every > aspect, being useful only to distinguish if such array has some member > missing. Also, we show a rate-limited warning in kernel log in such case. > > This patch has no proper functional change other than adding a 'clean'-like > state; it was tested with ext4 and xfs filesystems. It requires a 'mdadm' > counterpart to handle the 'broken' state. > > Cc: NeilBrown <neilb@suse.com> > Cc: Song Liu <songliubraving@fb.com> > Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com> > --- > drivers/md/md.c | 23 +++++++++++++++++++---- > drivers/md/md.h | 2 ++ > drivers/md/raid0.c | 26 ++++++++++++++++++++++++++ > 3 files changed, 47 insertions(+), 4 deletions(-) > > diff --git a/drivers/md/md.c b/drivers/md/md.c > index fba49918d591..b80f36084ec1 100644 > --- a/drivers/md/md.c > +++ b/drivers/md/md.c > @@ -4160,12 +4160,18 @@ __ATTR_PREALLOC(resync_start, S_IRUGO|S_IWUSR, > * active-idle > * like active, but no writes have been seen for a while (100msec). > * > + * broken > + * RAID0-only: same as clean, but array is missing a member. > + * It's useful because RAID0 mounted-arrays aren't stopped > + * when a member is gone, so this state will at least alert > + * the user that something is wrong. Curious why only raid0 has this issue? Thanks, -Bob > + * > */ > enum array_state { clear, inactive, suspended, readonly, read_auto, clean, active, > - write_pending, active_idle, bad_word}; > + write_pending, active_idle, broken, bad_word}; > static char *array_states[] = { > "clear", "inactive", "suspended", "readonly", "read-auto", "clean", "active", > - "write-pending", "active-idle", NULL }; > + "write-pending", "active-idle", "broken", NULL }; > > static int match_word(const char *word, char **list) > { > @@ -4181,7 +4187,7 @@ array_state_show(struct mddev *mddev, char *page) > { > enum array_state st = inactive; > > - if (mddev->pers) > + if (mddev->pers) { > switch(mddev->ro) { > case 1: > st = readonly; > @@ -4201,7 +4207,15 @@ array_state_show(struct mddev *mddev, char *page) > st = active; > spin_unlock(&mddev->lock); > } > - else { > + > + if ((mddev->pers->level == 0) && > + ((st == clean) || (st == broken))) { > + if (mddev->pers->is_missing_dev(mddev)) > + st = broken; > + else > + st = clean; > + } > + } else { > if (list_empty(&mddev->disks) && > mddev->raid_disks == 0 && > mddev->dev_sectors == 0) > @@ -4315,6 +4329,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len) > break; > case write_pending: > case active_idle: > + case broken: > /* these cannot be set */ > break; > } > diff --git a/drivers/md/md.h b/drivers/md/md.h > index 41552e615c4c..e7b42b75701a 100644 > --- a/drivers/md/md.h > +++ b/drivers/md/md.h > @@ -590,6 +590,8 @@ struct md_personality > int (*congested)(struct mddev *mddev, int bits); > /* Changes the consistency policy of an active array. */ > int (*change_consistency_policy)(struct mddev *mddev, const char *buf); > + /* Check if there is any missing/failed members - RAID0 only for now. */ > + bool (*is_missing_dev)(struct mddev *mddev); > }; > > struct md_sysfs_entry { > diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c > index 58a9cc5193bf..79618a6ae31a 100644 > --- a/drivers/md/raid0.c > +++ b/drivers/md/raid0.c > @@ -455,6 +455,31 @@ static inline int is_io_in_chunk_boundary(struct mddev *mddev, > } > } > > +bool raid0_is_missing_dev(struct mddev *mddev) > +{ > + struct md_rdev *rdev; > + static int already_missing; > + int def_disks, work_disks = 0; > + struct r0conf *conf = mddev->private; > + > + def_disks = conf->strip_zone[0].nb_dev; > + rdev_for_each(rdev, mddev) > + if (rdev->bdev->bd_disk->flags & GENHD_FL_UP) > + work_disks++; > + > + if (unlikely(def_disks - work_disks)) { > + if (!already_missing) { > + already_missing = 1; > + pr_warn("md: %s: raid0 array has %d missing/failed members\n", > + mdname(mddev), (def_disks - work_disks)); > + } > + return true; > + } > + > + already_missing = 0; > + return false; > +} > + > static void raid0_handle_discard(struct mddev *mddev, struct bio *bio) > { > struct r0conf *conf = mddev->private; > @@ -789,6 +814,7 @@ static struct md_personality raid0_personality= > .takeover = raid0_takeover, > .quiesce = raid0_quiesce, > .congested = raid0_congested, > + .is_missing_dev = raid0_is_missing_dev, > }; > > static int __init raid0_init (void) > ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-30 6:20 ` Bob Liu @ 2019-07-30 12:18 ` Guilherme G. Piccoli 2019-07-31 0:28 ` NeilBrown 1 sibling, 0 replies; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-07-30 12:18 UTC (permalink / raw) To: Bob Liu, linux-raid Cc: linux-block, dm-devel, jay.vosburgh, neilb, songliubraving On 30/07/2019 03:20, Bob Liu wrote: > [...] >> + * broken >> + * RAID0-only: same as clean, but array is missing a member. >> + * It's useful because RAID0 mounted-arrays aren't stopped >> + * when a member is gone, so this state will at least alert >> + * the user that something is wrong. > > > Curious why only raid0 has this issue? > > Thanks, -Bob Hi Bob, I understand that all other levels have fault-tolerance logic, while raid0 is just a "bypass" driver that selects the correct underlying device to send the BIO and blindly sends it. It's known to be a performance-only /lightweight solution whereas the other levels aim to be reliable. I've quickly tested raid5 and rai10, and see messages like this on kernel log when removing a device (in raid5): [35.764975] md/raid:md0: Disk failure on nvme1n1, disabling device. md/raid:md0: Operation continuing on 1 devices. The message seen in raid10 is basically the same. As a (cheap) comparison of the complexity among levels, look that: <...>/linux-mainline/drivers/md# cat raid5* | wc -l 14191 <...>/linux-mainline/drivers/md# cat raid10* | wc -l 5135 <...>/linux-mainline/drivers/md# cat raid0* | wc -l 820 Cheers, Guilherme ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-30 6:20 ` Bob Liu 2019-07-30 12:18 ` Guilherme G. Piccoli @ 2019-07-31 0:28 ` NeilBrown 2019-07-31 13:04 ` Guilherme G. Piccoli 1 sibling, 1 reply; 13+ messages in thread From: NeilBrown @ 2019-07-31 0:28 UTC (permalink / raw) To: Bob Liu, Guilherme G. Piccoli, linux-raid Cc: jay.vosburgh, songliubraving, dm-devel, Neil F Brown, linux-block [-- Attachment #1: Type: text/plain, Size: 224 bytes --] On Tue, Jul 30 2019, Bob Liu wrote: > > > Curious why only raid0 has this issue? Actually, it isn't only raid0. 'linear' has the same issue. Probably the fix for raid0 should be applied to linear too. NeilBrown [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 832 bytes --] ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-31 0:28 ` NeilBrown @ 2019-07-31 13:04 ` Guilherme G. Piccoli 2019-07-31 19:47 ` Song Liu 0 siblings, 1 reply; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-07-31 13:04 UTC (permalink / raw) To: NeilBrown, Bob Liu, linux-raid Cc: jay.vosburgh, songliubraving, dm-devel, Neil F Brown, linux-block On 30/07/2019 21:28, NeilBrown wrote: > On Tue, Jul 30 2019, Bob Liu wrote: >> >> >> Curious why only raid0 has this issue? > > Actually, it isn't only raid0. 'linear' has the same issue. > Probably the fix for raid0 should be applied to linear too. > > NeilBrown > Thanks Neil, it makes sense! I didn't considered "linear" and indeed, after some testing, it reacts exactly as raid0/stripping. In case this patch gets good acceptance I can certainly include md/linear in that! Cheers, Guilherme ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-31 13:04 ` Guilherme G. Piccoli @ 2019-07-31 19:47 ` Song Liu 0 siblings, 0 replies; 13+ messages in thread From: Song Liu @ 2019-07-31 19:47 UTC (permalink / raw) To: Guilherme G. Piccoli Cc: NeilBrown, Bob Liu, linux-raid, Jay Vosburgh, Song Liu, dm-devel, Neil F Brown, linux-block On Wed, Jul 31, 2019 at 6:05 AM Guilherme G. Piccoli <gpiccoli@canonical.com> wrote: > > On 30/07/2019 21:28, NeilBrown wrote: > > On Tue, Jul 30 2019, Bob Liu wrote: > >> > >> > >> Curious why only raid0 has this issue? > > > > Actually, it isn't only raid0. 'linear' has the same issue. > > Probably the fix for raid0 should be applied to linear too. > > > > NeilBrown > > > > Thanks Neil, it makes sense! I didn't considered "linear" and indeed, > after some testing, it reacts exactly as raid0/stripping. > > In case this patch gets good acceptance I can certainly include > md/linear in that! > Cheers, This looks good. Please include Neil's feedback in v2. Btw, there is no 2/2 in this set, right? Song ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-29 20:31 ` [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 Guilherme G. Piccoli 2019-07-30 0:11 ` NeilBrown 2019-07-30 6:20 ` Bob Liu @ 2019-07-31 19:43 ` Song Liu 2019-08-01 12:07 ` Guilherme G. Piccoli 2 siblings, 1 reply; 13+ messages in thread From: Song Liu @ 2019-07-31 19:43 UTC (permalink / raw) To: Guilherme G. Piccoli Cc: linux-raid, linux-block, dm-devel, Jay Vosburgh, NeilBrown, Song Liu On Mon, Jul 29, 2019 at 1:33 PM Guilherme G. Piccoli <gpiccoli@canonical.com> wrote: > > Currently if a md/raid0 array gets one or more members removed while > being mounted, kernel keeps showing state 'clean' in the 'array_state' > sysfs attribute. Despite udev signaling the member device is gone, 'mdadm' > cannot issue the STOP_ARRAY ioctl successfully, given the array is mounted. > > Nothing else hints that something is wrong (except that the removed devices > don't show properly in the output of 'mdadm detail' command). There is no > other property to be checked, and if user is not performing reads/writes > to the array, even kernel log is quiet and doesn't give a clue about the > missing member. > > This patch changes this behavior; when 'array_state' is read we introduce > a non-expensive check (only for raid0) that relies in the comparison of > the total number of disks when array was assembled with gendisk flags of > those devices to validate if all members are available and functional. > A new array state 'broken' was added: it mimics the state 'clean' in every > aspect, being useful only to distinguish if such array has some member > missing. Also, we show a rate-limited warning in kernel log in such case. > > This patch has no proper functional change other than adding a 'clean'-like > state; it was tested with ext4 and xfs filesystems. It requires a 'mdadm' > counterpart to handle the 'broken' state. > > Cc: NeilBrown <neilb@suse.com> > Cc: Song Liu <songliubraving@fb.com> > Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com> > --- > [...] > @@ -4315,6 +4329,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len) > break; > case write_pending: > case active_idle: > + case broken: > /* these cannot be set */ > break; > } Maybe it is useful to set "broken" state? When user space found some issues with the drive? Thanks, Song > diff --git a/drivers/md/md.h b/drivers/md/md.h > index 41552e615c4c..e7b42b75701a 100644 > --- a/drivers/md/md.h > +++ b/drivers/md/md.h > @@ -590,6 +590,8 @@ struct md_personality > int (*congested)(struct mddev *mddev, int bits); > /* Changes the consistency policy of an active array. */ > int (*change_consistency_policy)(struct mddev *mddev, const char *buf); > + /* Check if there is any missing/failed members - RAID0 only for now. */ > + bool (*is_missing_dev)(struct mddev *mddev); > }; > > struct md_sysfs_entry { > diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c > index 58a9cc5193bf..79618a6ae31a 100644 > --- a/drivers/md/raid0.c > +++ b/drivers/md/raid0.c > @@ -455,6 +455,31 @@ static inline int is_io_in_chunk_boundary(struct mddev *mddev, > } > } > > +bool raid0_is_missing_dev(struct mddev *mddev) > +{ > + struct md_rdev *rdev; > + static int already_missing; > + int def_disks, work_disks = 0; > + struct r0conf *conf = mddev->private; > + > + def_disks = conf->strip_zone[0].nb_dev; > + rdev_for_each(rdev, mddev) > + if (rdev->bdev->bd_disk->flags & GENHD_FL_UP) > + work_disks++; > + > + if (unlikely(def_disks - work_disks)) { > + if (!already_missing) { > + already_missing = 1; > + pr_warn("md: %s: raid0 array has %d missing/failed members\n", > + mdname(mddev), (def_disks - work_disks)); > + } > + return true; > + } > + > + already_missing = 0; > + return false; > +} > + > static void raid0_handle_discard(struct mddev *mddev, struct bio *bio) > { > struct r0conf *conf = mddev->private; > @@ -789,6 +814,7 @@ static struct md_personality raid0_personality= > .takeover = raid0_takeover, > .quiesce = raid0_quiesce, > .congested = raid0_congested, > + .is_missing_dev = raid0_is_missing_dev, > }; > > static int __init raid0_init (void) > -- > 2.22.0 > ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-07-31 19:43 ` Song Liu @ 2019-08-01 12:07 ` Guilherme G. Piccoli 2019-08-16 13:48 ` Guilherme G. Piccoli 0 siblings, 1 reply; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-08-01 12:07 UTC (permalink / raw) To: Song Liu Cc: linux-raid, linux-block, dm-devel, Jay Vosburgh, NeilBrown, Song Liu On 31/07/2019 16:43, Song Liu wrote: >[...] >> @@ -4315,6 +4329,7 @@ array_state_store(struct mddev *mddev, const char *buf, size_t len) >> break; >> case write_pending: >> case active_idle: >> + case broken: >> /* these cannot be set */ >> break; >> } > > Maybe it is useful to set "broken" state? When user space found some issues > with the drive? > > Thanks, > Song Hi Song, thanks a lot for your feedback. I can easily add that in V2 along with Neil's suggestion, I agree with you. There is a 2/2 patch regarding the mdadm counterpart; you should have received the email, but for completeness, the link is: lore.kernel.org/linux-block/20190729203135.12934-3-gpiccoli@canonical.com Thanks, Guilherme ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 2019-08-01 12:07 ` Guilherme G. Piccoli @ 2019-08-16 13:48 ` Guilherme G. Piccoli 0 siblings, 0 replies; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-08-16 13:48 UTC (permalink / raw) To: Song Liu Cc: linux-raid, linux-block, dm-devel, Jay Vosburgh, NeilBrown, Song Liu V2 just sent: lore.kernel.org/linux-block/20190816134059.29751-1-gpiccoli@canonical.com Thanks, Guilherme ^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH 2/2] mdadm: Introduce new array state 'broken' for raid0 2019-07-29 20:31 [PATCH 0/2] Introduce new raid0 state 'broken' Guilherme G. Piccoli 2019-07-29 20:31 ` [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 Guilherme G. Piccoli @ 2019-07-29 20:31 ` Guilherme G. Piccoli 1 sibling, 0 replies; 13+ messages in thread From: Guilherme G. Piccoli @ 2019-07-29 20:31 UTC (permalink / raw) To: linux-raid Cc: linux-block, dm-devel, gpiccoli, jay.vosburgh, neilb, songliubraving Currently if a md/raid0 array gets one or more members removed while being mounted, kernel keeps showing state 'clean' in the 'array_state' sysfs attribute. Despite udev signaling the member device is gone, 'mdadm' cannot issue the STOP_ARRAY ioctl successfully, given the array is mounted. Nothing else hints that something is wrong (except that the removed devices don't show properly in the output of mdadm 'detail' command). There is no other property to be checked, and if user is not performing reads/writes to the array, even kernel log is quiet and doesn't give a clue about the missing member. This patch is the mdadm counterpart of kernel new array state 'broken'. The 'broken' state mimics the state 'clean' in every aspect, being useful only to distinguish if an array has some member missing. All necessary paths in mdadm were changed to deal with 'broken' state, and in case the tool runs in a kernel that is not updated, it'll work normally, i.e., it doesn't require the 'broken' state in order to work. Also, this patch changes the way the array state is showed in the 'detail' command (for raid0 only) - now it takes the 'array_state' sysfs attribute into account instead of only rely in the MD_SB_CLEAN flag. Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com> --- Detail.c | 16 ++++++++++++++-- Monitor.c | 9 +++++++-- maps.c | 1 + mdadm.h | 1 + mdmon.h | 2 +- monitor.c | 4 ++-- 6 files changed, 26 insertions(+), 7 deletions(-) diff --git a/Detail.c b/Detail.c index 20ea03a..4bf86b1 100644 --- a/Detail.c +++ b/Detail.c @@ -81,6 +81,7 @@ int Detail(char *dev, struct context *c) int external; int inactive; int is_container = 0; + char arrayst[12] = { 0 }; /* no state is >10 chars currently */ if (fd < 0) { pr_err("cannot open %s: %s\n", @@ -485,9 +486,20 @@ int Detail(char *dev, struct context *c) else st = ", degraded"; + if (array.state & (1 << MD_SB_CLEAN)) { + if (array.level == 0) + strncpy(arrayst, + map_num(sysfs_array_states, + sra->array_state), + sizeof(arrayst)-1); + else + strncpy(arrayst, "clean", + sizeof(arrayst)-1); + } else + strncpy(arrayst, "active", sizeof(arrayst)-1); + printf(" State : %s%s%s%s%s%s \n", - (array.state & (1 << MD_SB_CLEAN)) ? - "clean" : "active", st, + arrayst, st, (!e || (e->percent < 0 && e->percent != RESYNC_PENDING && e->percent != RESYNC_DELAYED)) ? diff --git a/Monitor.c b/Monitor.c index 036103f..9a6a250 100644 --- a/Monitor.c +++ b/Monitor.c @@ -1055,8 +1055,12 @@ int Wait(char *dev) } } +/* The state "broken" is used only for RAID0 - it's the same as "clean", but + * used in case the array has one or more members missing. + */ +#define CLEAN_STATES_LAST_POS 5 static char *clean_states[] = { - "clear", "inactive", "readonly", "read-auto", "clean", NULL }; + "clear", "inactive", "readonly", "read-auto", "clean", "broken", NULL }; int WaitClean(char *dev, int verbose) { @@ -1116,7 +1120,8 @@ int WaitClean(char *dev, int verbose) rv = read(state_fd, buf, sizeof(buf)); if (rv < 0) break; - if (sysfs_match_word(buf, clean_states) <= 4) + if (sysfs_match_word(buf, clean_states) + <= CLEAN_STATES_LAST_POS) break; rv = sysfs_wait(state_fd, &delay); if (rv < 0 && errno != EINTR) diff --git a/maps.c b/maps.c index 02a0474..98ddbbc 100644 --- a/maps.c +++ b/maps.c @@ -150,6 +150,7 @@ mapping_t sysfs_array_states[] = { { "read-auto", ARRAY_READ_AUTO }, { "clean", ARRAY_CLEAN }, { "write-pending", ARRAY_WRITE_PENDING }, + { "broken", ARRAY_BROKEN_RAID0 }, { NULL, ARRAY_UNKNOWN_STATE } }; diff --git a/mdadm.h b/mdadm.h index c36d7fd..72c2525 100644 --- a/mdadm.h +++ b/mdadm.h @@ -375,6 +375,7 @@ struct mdinfo { ARRAY_ACTIVE, ARRAY_WRITE_PENDING, ARRAY_ACTIVE_IDLE, + ARRAY_BROKEN_RAID0, ARRAY_UNKNOWN_STATE, } array_state; struct md_bb bb; diff --git a/mdmon.h b/mdmon.h index 818367c..b3d72ac 100644 --- a/mdmon.h +++ b/mdmon.h @@ -21,7 +21,7 @@ extern const char Name[]; enum array_state { clear, inactive, suspended, readonly, read_auto, - clean, active, write_pending, active_idle, bad_word}; + clean, active, write_pending, active_idle, broken, bad_word}; enum sync_action { idle, reshape, resync, recover, check, repair, bad_action }; diff --git a/monitor.c b/monitor.c index 81537ed..e0d3be6 100644 --- a/monitor.c +++ b/monitor.c @@ -26,7 +26,7 @@ static char *array_states[] = { "clear", "inactive", "suspended", "readonly", "read-auto", - "clean", "active", "write-pending", "active-idle", NULL }; + "clean", "active", "write-pending", "active-idle", "broken", NULL }; static char *sync_actions[] = { "idle", "reshape", "resync", "recover", "check", "repair", NULL }; @@ -476,7 +476,7 @@ static int read_and_act(struct active_array *a, fd_set *fds) a->next_state = clean; ret |= ARRAY_DIRTY; } - if (a->curr_state == clean) { + if ((a->curr_state == clean) || (a->curr_state == broken)) { a->container->ss->set_array_state(a, 1); } if (a->curr_state == active || -- 2.22.0 ^ permalink raw reply related [flat|nested] 13+ messages in thread
end of thread, other threads:[~2019-08-16 13:48 UTC | newest] Thread overview: 13+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2019-07-29 20:31 [PATCH 0/2] Introduce new raid0 state 'broken' Guilherme G. Piccoli 2019-07-29 20:31 ` [PATCH 1/2] md/raid0: Introduce new array state 'broken' for raid0 Guilherme G. Piccoli 2019-07-30 0:11 ` NeilBrown 2019-07-30 11:43 ` Guilherme G. Piccoli 2019-07-30 6:20 ` Bob Liu 2019-07-30 12:18 ` Guilherme G. Piccoli 2019-07-31 0:28 ` NeilBrown 2019-07-31 13:04 ` Guilherme G. Piccoli 2019-07-31 19:47 ` Song Liu 2019-07-31 19:43 ` Song Liu 2019-08-01 12:07 ` Guilherme G. Piccoli 2019-08-16 13:48 ` Guilherme G. Piccoli 2019-07-29 20:31 ` [PATCH 2/2] mdadm: " Guilherme G. Piccoli
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).