RE: Single-drive RAID0
flat view
From: Wojcik, Krzysztof <hidden>
Date: 2011-02-22 11:50:11
-----Original Message----- From: NeilBrown [mailto:neilb@suse.de] Sent: Tuesday, February 22, 2011 1:50 AM To: Wojcik, Krzysztof Cc: linux-raid@vger.kernel.org Subject: Re: Single-drive RAID0 On Mon, 21 Feb 2011 14:11:51 +0000 "Wojcik, Krzysztof" [off-list ref] wrote:quoted
Hi, I tested your proposal. It resolve the problem.Thanks!quoted
What are your future plans? Do you need to consult the patch with someone involved in the blockdevices code?quoted
When can I expect final patch?Here it is. As you can see it is quite different - I figured out what is really going on..quoted
quoted
If it does, then we can explore more... Did you say that this works better on earlier kernels? If so, whatisquoted
quoted
the latest kernel that you have tried that you don't have this problemon?quoted
I've tested on kernel 2.6.32 from RHEL6.0 and, if I remembercorrectly, 2.6.34- they work properly. That fits - I discovered that I broke it in 2.6.35. Thanks, NeilBrown
Hi, I confirm. This patch resolve the problem. I did tests on your current top of for-next branch and issue is not reproducible. Thank for your cooperation! :-) Regards Krzysztof
quoted hunk ↗ jump to hunk
From b7fea734c7780ca80f58fb87a6c9f55aa83314d7 Mon Sep 17 00:00:00 2001 From: NeilBrown <redacted> Date: Tue, 22 Feb 2011 11:48:03 +1100 Subject: [PATCH] md: Fix - again - partition detection when array becomes active Revert b821eaa572fd737faaf6928ba046e571526c36c6 and f3b99be19ded511a1bf05a148276239d9f13eefa When I wrote the first of these I had a wrong idea about the lifetime of 'struct block_device'. It can disappear at any time that the block device is not open if it falls out of the inode cache. So relying on the 'size' recorded with it to detect when the device size has changed and so we need to revalidate, is wrong. Rather, we really do need the 'changed' attribute stored directly in the mddev and set/tested as appropriate. Without this patch, a sequence of: mknod / open / close / unlink (which can cause a block_device to be created and then destroyed) will result in a rescan of the partition table and consequence removal and addition of partitions. Several of these in a row can get udev racing to create and unlink and other code can get confused. With the patch, the rescan is only performed when needed and so there are no races. This is suitable for any stable kernel from 2.6.35. Reported-by: "Wojcik, Krzysztof" <redacted> Signed-off-by: NeilBrown <redacted> Cc: stable@kernel.org --- drivers/md/md.c | 22 +++++++++++++++++++++- drivers/md/md.h | 2 ++ 2 files changed, 23 insertions(+), 1 deletions(-)diff --git a/drivers/md/md.c b/drivers/md/md.c index 330addf..818313e 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c@@ -4627,6 +4627,7 @@ static int do_md_run(mddev_t *mddev) } set_capacity(mddev->gendisk, mddev->array_sectors); revalidate_disk(mddev->gendisk); + mddev->changed = 1; kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE); out: return err;@@ -4715,6 +4716,7 @@ static void md_clean(mddev_t *mddev) mddev->sync_speed_min = mddev->sync_speed_max = 0; mddev->recovery = 0; mddev->in_sync = 0; + mddev->changed = 0; mddev->degraded = 0; mddev->safemode = 0; mddev->bitmap_info.offset = 0;@@ -4830,6 +4832,7 @@ static int do_md_stop(mddev_t * mddev, int mode,int is_open) set_capacity(disk, 0); mutex_unlock(&mddev->open_mutex); + mddev->changed = 1; revalidate_disk(disk); if (mddev->ro)@@ -6014,7 +6017,7 @@ static int md_open(struct block_device *bdev,fmode_t mode) atomic_inc(&mddev->openers); mutex_unlock(&mddev->open_mutex); - check_disk_size_change(mddev->gendisk, bdev); + check_disk_change(bdev); out: return err; }@@ -6029,6 +6032,21 @@ static int md_release(struct gendisk *disk,fmode_t mode) return 0; } + +static int md_media_changed(struct gendisk *disk) +{ + mddev_t *mddev = disk->private_data; + + return mddev->changed; +} + +static int md_revalidate(struct gendisk *disk) +{ + mddev_t *mddev = disk->private_data; + + mddev->changed = 0; + return 0; +} static const struct block_device_operations md_fops = { .owner = THIS_MODULE,@@ -6039,6 +6057,8 @@ static const struct block_device_operationsmd_fops = .compat_ioctl = md_compat_ioctl, #endif .getgeo = md_getgeo, + .media_changed = md_media_changed, + .revalidate_disk= md_revalidate, }; static int md_thread(void * arg)diff --git a/drivers/md/md.h b/drivers/md/md.h index 7e90b85..12215d4 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h@@ -274,6 +274,8 @@ struct mddev_s atomic_t active; /* general refcount */ atomic_t openers; /* number of active opens */ + int changed; /* True if we might need to + * reread partition info */ int degraded; /* whether md should consider * adding a spare */ --1.7.1