Actions
Bug #14022
open
zpool online -e breaks access to pool
Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Start date:
Due date:
% Done:
0%
Estimated time:
Difficulty:
Medium
Tags:
Gerrit CR:
Description
Attempting to expand a ZFS pool in one operation using zpool online -e does not work, causing pool access (and the machine) to hang.
If the partition is expanded first using format, however, then it succeeds.
root@omnios:~# zpool status
pool: rpool
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
rpool ONLINE 0 0 0
c1t0d0 ONLINE 0 0 0
errors: No known data errors
root@omnios:~# zpool get expandsz rpool
NAME PROPERTY VALUE SOURCE
rpool expandsize 2G -
root@omnios:~# prtvtoc /dev/dsk/c1t0d0
* /dev/dsk/c1t0d0 EFI partition map
*
* Dimensions:
* 512 bytes/sector
* 20971520 sectors
* 16777149 accessible sectors
*
* Flags:
* 1: unmountable
* 10: read-only
*
* Unallocated space:
* First Sector Last
* Sector Count Sector
* 34 222 255
*
* First Sector Last
* Partition Tag Flags Sector Count Sector Mount Directory
0 12 00 256 524288 524543
1 4 00 524544 16236255 16760798
8 11 00 16760799 16384 16777182
root@omnios:~# zpool online -e rpool c1t0d0
[0]> ::stacks -c vdev_online
THREAD STATE SOBJ COUNT
fffffe064f725400 SLEEP CV 1
swtch+0x133
cv_wait+0x68
zfs`txg_wait_synced_impl+0xa5
zfs`txg_wait_synced+0xd
zfs`spa_vdev_state_exit+0xb9
zfs`vdev_online+0x23a
zfs`zfs_ioc_vdev_set_state+0x7b
zfs`zfsdev_ioctl+0x1fd
cdev_ioctl+0x2b
specfs`spec_ioctl+0x45
fop_ioctl+0x5b
ioctl+0x153
[0]> ::msgbuf
...
WARNING: Pool 'rpool' has encountered an uncorrectable I/O failure and has been suspended; `zpool clear` will be required before the pool can be written to.
[0]> ::zfs_dbgmsg disk vdev '/dev/dsk/c1t0d0s1': vdev_disk_open: failed to get size spa=rpool async request task=1 rpool: metaslab allocation failure: zio fffffe065a5b49f8, size 512, error 28 rpool: metaslab allocation failure: zio fffffe065a5d3dd0, size 4096, error 28 rpool: metaslab allocation failure: zio fffffe065a5c2228, size 4096, error 28 rpool: metaslab allocation failure: zio fffffe065a5c3508, size 4096, error 28 rpool: metaslab allocation failure: zio fffffe065a586cb0, size 49152, error 28 rpool: metaslab allocation failure: zio fffffe065a5d1200, size 4096, error 28 rpool: zil block allocation failure: size 36864, error 28 rpool: metaslab allocation failure: zio fffffe065a584930, size 4096, error 28
[0]> ::spa ADDR STATE NAME fffffe064a8eb000 ACTIVE rpool [0]> fffffe064a8eb000::spa_vdevs ADDR STATE AUX DESCRIPTION fffffe064b8ee000 CANT_OPEN NO_REPLICAS root fffffe064b8eb000 CANT_OPEN OPEN_FAILED /dev/dsk/c1t0d0s1
This is a bhyve VM, so:
carolina# mdb /dev/zvol/rdsk/rpool/test/root > ::load disk_label > ::gpt Signature: EFI PART (valid) Revision: 1.0 HeaderSize: 92 bytes HeaderCRC32: 0xfcd5dc82 (should be 0xfcd5dc82) Reserved1: 0 (should be 0x0) MyLBA: 1 (should be 1) AlternateLBA: 20971519 FirstUsableLBA: 34 LastUsableLBA: 20971486 DiskGUID: 5897e0f9-b01a-4cd7-d30f-95c37f715c2f PartitionEntryLBA: 2 NumberOfPartitionEntries: 9 SizeOfPartitionEntry: 0x80 bytes PartitionEntryArrayCRC32: 0x59d5ed6a (should be 0x59d5ed6a) PART TYPE STARTLBA ENDLBA ATTR NAME 0 EFI_SYSTEM 256 524543 0 loader 1 EFI_USR 524544 20955102 0 zfs 2 EFI_UNUSED 3 EFI_UNUSED 4 EFI_UNUSED 5 EFI_UNUSED 6 EFI_UNUSED 7 EFI_UNUSED 8 EFI_RESERVED 20955103 20971486 0
Updated by Toomas Soome 2 months ago
Andy Fiddaman wrote:
Attempting to expand a ZFS pool in one operation using
zpool online -edoes not work, causing pool access (and the machine) to hang.
If the partition is expanded first usingformat, however, then it succeeds.
EFI wholedisk or custom partitioning?
Updated by Andy Fiddaman 2 months ago
Toomas Soome wrote in #note-1:
EFI wholedisk or custom partitioning?
I've expanded the description - the reproducer I'm using is EFI whole disk, so that zpool online -e tries to expand it
Updated by Andy Fiddaman about 2 months ago
The pool ends up being suspended because ldi_get_size() returns -1 - called from vdev_disk_open() in the re-open case
dtrace -n 'fbt::ldi_get_size:entry{v = (struct ldi_handle *)arg0; print(*v); self->t++}' -n 'fbt::ldi*:entry/self->t/{}' -n 'fbt::ldi*:return/self->t/{trace(arg1)}' -F
CPU FUNCTION
2 -> ldi_get_size struct ldi_handle {
struct ldi_handle *lh_next = 0
uint_t lh_ref = 0x1
uint_t lh_flags = 0
uint_t lh_type = 0x2
struct ldi_ident *lh_ident = 0xfffffe064a4abbf8
vnode_t *lh_vp = 0xfffffe064a4c3a40
kmutex_t [1] lh_lock = [
kmutex_t {
void *[1] _opaque = [ 0 ]
}
]
struct ldi_event *lh_events = 0
}
2 | ldi_get_size:entry
2 -> ldi_get_otyp
2 <- ldi_get_otyp 0
2 -> ldi_prop_exists
2 <- ldi_prop_exists 0
2 -> ldi_prop_exists
2 <- ldi_prop_exists 0
2 -> ldi_prop_exists
2 <- ldi_prop_exists 0
2 -> ldi_prop_exists
2 <- ldi_prop_exists 0
2 <- ldi_get_size 4294967295
Actions