1
0
mirror of git://sourceware.org/git/lvm2.git synced 2024-12-29 15:22:30 +03:00
Commit Graph

355 Commits

Author SHA1 Message Date
Alasdair G Kergon
366493a1d1 device: Suppress repeated reads of the same data.
If the data being requested is present in last_[extra_]devbuf,
return that directly instead of reading it from disk again.

Typical LVM2 access patterns request data within two adjacent 4k blocks
so we eliminate some read() system calls by always reading at least 8k.
2018-01-10 15:52:03 +00:00
Alasdair G Kergon
dcb2a5a611 device: Remove some data copying between buffers.
Callers that read larger amounts of data now get a pointer to read-only
data directly without copying it through an intermediate buffer.  This
data is owned by the device layer so the callers no longer free it.
2018-01-10 15:48:03 +00:00
Alasdair G Kergon
4d568b709c device: Free cached device bufs when metadata invalid or dev closed. 2018-01-10 15:48:03 +00:00
Alasdair G Kergon
bd0967a4b1 device: Keep the last data buffer read off each device.
If there's a second metadata area on device, we record that separately.

Note that the memory requirements aren't restricted yet.
2018-01-10 15:48:03 +00:00
Alasdair G Kergon
ea96381534 libdm: Introduce dm_malloc_aligned 2018-01-10 15:48:03 +00:00
Alasdair G Kergon
f4675af4cf format_text: Use vgsummary callbacks 2018-01-09 03:14:30 +00:00
Alasdair G Kergon
5e7d3ad749 device: Introduce dev_read_callback
If it obtains the data, it passes it into the supplied callback function
and returns 1.  Otherwise the callback receives failed = 1.

Updated config_file_read_fd to use this and similarly return the data
via a callback fn of its own.
2018-01-06 02:40:12 +00:00
Alasdair G Kergon
946f07af3e metadata: Use a consistent format for callback fn parameters 2018-01-05 14:24:56 +00:00
Alasdair G Kergon
22b6c482ec config: Split config buffer processing into new fn.
Wrap its parameters into struct process_config_file_params allocated
from a mempool now passed into the config_file_read* fns.
2018-01-02 21:10:46 +00:00
Alasdair G Kergon
17649d4ac8 device: Move dev_read memory allocation into device layer.
Rename dev_read() to dev_read_buf() - the function that reads data
into a supplied buffer.

Introduce a new dev_read() that allocates the buffer it returns and
switch the important users over to this.  No caller may change the
returned data.  (For now, callers are responsible for freeing it after
use, but later the device layer will take full ownership.)

dev_read_buf() should only be used for tiny buffers or unimportant code
(such as the old disk formats).
2017-12-19 01:31:50 +00:00
Alasdair G Kergon
5f45cb90a7 format_text: Transfer circular buf alloc to device layer.
Instead of the caller passing dev_read_circular() a buffer to fill with
data, the device layer itself now allocates it.
2017-12-15 22:34:26 +00:00
Alasdair G Kergon
beee9940a5 format_text: Separate out code paths for buffer wraparound
The creation of wrapped around metadata - where the start of metadata is
written up to the end of the buffer and the remainder follows back at
the start of the buffer - is now restricted to cases where writing the
metadata in one piece wouldn't fit.  This shouldn't happen in 'normal'
usage so let's begin treating the code for this as a special case that
can be ignored when optimising 'normal' cases.
2017-12-15 21:12:19 +00:00
Alasdair G Kergon
e932c5da50 device: Fix an unpaired device close.
dev_open_flags contains an unpaired dev_close_immediate so increment
open_count before calling it.
2017-12-12 17:56:58 +00:00
Alasdair G Kergon
c5ef76bf27 device: Internal error if writing 0 bytes to dev. 2017-12-12 12:57:25 +00:00
Alasdair G Kergon
d591d04103 device: Tag I/O for each mda on a device separately in log messages.
Mark the first metadata area on each text format PV as MDA_PRIMARY.
Pass this information down to the device layer so that when
there are two metadata areas on a block device, we can easily
distinguish two independent streams of I/O.
2017-12-07 03:48:11 +00:00
Alasdair G Kergon
7195df5aca device: Skip read-modify-write if replacing whole block. 2017-12-05 01:00:38 +00:00
Alasdair G Kergon
e4805e4883 device: categorise block i/o
Introduce enum dev_io_reason to categorise block device I/O
in debug messages so it's obvious what it is for.

DEV_IO_SIGNATURES   /* Scanning device signatures */
DEV_IO_LABEL        /* LVM PV disk label */
DEV_IO_MDA_HEADER   /* Text format metadata area header */
DEV_IO_MDA_CONTENT  /* Text format metadata area content */
DEV_IO_FMT1         /* Original LVM1 metadata format */
DEV_IO_POOL         /* Pool metadata format */
DEV_IO_LV           /* Content written to an LV */
DEV_IO_LOG          /* Logging messages */
2017-12-04 23:45:26 +00:00
Alasdair G Kergon
115e66e9be device: log debug when I/O bounce buffer used 2017-11-16 19:16:10 +00:00
Alasdair G Kergon
02e9876665 log: Add io debug class 2017-11-15 01:02:15 +00:00
Alasdair G Kergon
6bf0f04ae2 log: Improve various device-related messages
- Use 'lvmcache' consistently instead of 'metadata cache'
- Always use 5 characters for source line number
- Remember to convert uuids into printable form
- Use <no name> rather than (null) when VG has no name.
2017-11-13 19:45:33 +00:00
Zdenek Kabelac
e9206fb93d devcache: track more udev errors
Add a bit more details for failing udev function.
2017-10-30 13:16:50 +01:00
Alasdair G Kergon
f1cc5b12fd tidy: Add missing underscores to statics. 2017-10-18 15:58:13 +01:00
Alasdair G Kergon
146745ad88 device: Separate errors for dev not found and filtered.
Replaced the confusing device error message "not found (or ignored by
filtering)" by either "not found" or "excluded by a filter".
(Later we should be able to say which filter.)

Left the the liblvm code paths alone.
2017-10-17 02:12:41 +01:00
Zdenek Kabelac
48ce8c7a49 tidy: drop unneeded cast
Avoid casting to the same type.
2017-07-20 11:20:44 +02:00
Zdenek Kabelac
0bf836aa14 tidy: prefer not using else after return
clang-tidy: avoid using  'else' after return - give more readable code,
and also saves indention level.
2017-07-20 11:18:29 +02:00
Zdenek Kabelac
0d0a3397c2 cleanup: add braces in macro 2017-07-20 11:18:29 +02:00
Zdenek Kabelac
767a5e1281 dev-cache: avoid hashing same data again
Before hashing device again with path, check if it's not already hashed.

TODO: maybe bigger chunk of executed code might be actually skipped.
2017-07-17 12:33:17 +02:00
Alasdair G Kergon
6a20b22151 devices: Recognise Veritas Dynamic Multipathing
VxDMP doesn't interact very well with udev so always set
  devices/obtain_device_list_from_udev = 0
in lvm.conf on these systems.
2017-01-10 22:23:23 +00:00
Zdenek Kabelac
1d58074d9f debug: more stacktrace corrections
Continue previous patch dropping some unneeded stack traces
after printed log_error/warn messages.
2016-11-25 14:58:28 +01:00
Peter Rajnoha
c8a14a29cd dev-type: check for DEVLINKS udev db variable existence if udev_device_get_is_initialized fn is not present
Older udev versions (udev < v165), don't have the official
udev_device_get_is_initialized function available to query for
device initialization state in udev database. Also, devices don't
have USEC_INITIALIZED udev db variable set - this is bound to the
udev_device_get_is_initialized fn functionality.

In this case, check for "DEVLINKS" variable instead - all block devices
have at least one symlink set for the node (the "/dev/block/<major:minor>".
This symlink is set by default basic udev rules provided by udev directly.
We'll use this as an alternative for the check that initial udev
processing for a device has already finished.
2016-09-06 13:21:29 +02:00
Peter Rajnoha
d7b282c601 dev-type: use more appropriate messages in udev_dev_is_mpath_component and use 10s timeout 2016-09-05 14:37:13 +02:00
Peter Rajnoha
5d323c37f3 refactor: move and rename _dev_is_mpath_component in lvmetad.c to udev_dev_is_mpath_component in dev-type.c 2016-09-05 12:55:25 +02:00
Zdenek Kabelac
d37a26b680 devices: handle partscan loop devices
Treat loop device created with 'losetup -P' as regular
partitioned device - so if it has partition table,
prevent its usage in commands like 'pvcreate'.

Before 'pvcreate /dev/loop0' could have erased and formated as PV,
after this patch, device is filtered out and cannot be used.
2016-06-01 17:37:47 +02:00
Alasdair G Kergon
b5314c2a6a device: Retry open without O_NOATIME if it fails. 2016-05-12 01:05:52 +01:00
Zdenek Kabelac
447daa9179 coverity: use wider type for whole expression
Coverity likes when the types are same through the whole expression.
And since dev_t is 64b - widen int type early.
2016-04-22 01:12:34 +02:00
Zdenek Kabelac
cbf99be43a coverity: fix memory access
Commit  52e0d0db44 introduced regression
as code may access   buf[0 - 1].

Reorder code to first remove '\n' and then check buffer size for
empty.
2016-04-22 01:11:57 +02:00
Zdenek Kabelac
556eba1835 cleanup: use kdev_t header in lvm tree
Reuse libdm header in lvm so we have single definition
of MAJOR/MINOR/MKDEV macros in use.
2016-04-22 00:23:28 +02:00
Zdenek Kabelac
0f7975cb35 cleanup: avoid declaring var in the middle of code
Easier to read code.
2016-04-12 11:47:51 +02:00
Zdenek Kabelac
fe65a86cbc devcache: do not insert devices without device node
When not obtaining device from udev, we are doing deep devdir scan,
and at the same time we try to insert everything what /sys/dev/block
knows about. However in case  lvm2 is configured to use nonstardard
devdir this way it will see (and scan) devices from a real system.

lvm2 test suite is using its own test devdir with its
own device nodes. To avoid touching real /dev  devices, validate
the device node exist in give dir and do not insert such device
into a cache.

With obtain list from udev this patch has no effect
(the normal user path).
2016-04-11 10:33:14 +02:00
Zdenek Kabelac
07a60b59f7 devcache: index devices also without udev
We have _insert_dirs() for  udev and non-udev compilation.
Compiling without udev missed to call dev_cache_index_devs().
Move the call after _insert_dirs() call so both compilation
gets it.
2016-04-11 10:32:19 +02:00
Peter Rajnoha
42f04a0f77 dev-cache: skip VGID/LVID indexing if /sys/dev/block is not present
/sys/dev/block is available since kernel version 2.2.26 (~ 2008):
https://www.kernel.org/doc/Documentation/ABI/testing/sysfs-dev

The VGID/LVID indexing code relies on this feature so skip indexing
if it's not available to avoid error messages about inability to open
/sys/dev/block directory.

We're not going to provide fallback code to read the /sys/block/
instead in this case as that's not that efficient - it needs extra
reads for getting major:minor and reading partitions would also
pose further reads and that's not worth it.
2016-04-01 17:09:15 +02:00
Peter Rajnoha
15d1824fac dev-cache: iterate devices in sysfs for VGID/LVID index if obtain_device_list_from_udev=0
If obtain_device_list_from_udev=0, LVM can make use of persistent .cache
file. This cache file contains only devices which underwent filters in
previous LVM command run. But we need to iterate over all block devices
to create the VGID/LVID index completely for the device mismatch check
to be complete as well.

This patch iterates over block devices found in sysfs to generate the
VGID/LVID index in dev cache if obtain_device_list_from_udev=0
(if obtain_device_list_from_udev=1, we always read complete list of
block devices from udev and we ignore .cache file so we don't need
to look in sysfs for the complete list).
2016-04-01 14:49:39 +02:00
Peter Rajnoha
7ed5a65ee5 dev-cache: also add dev name for device found in sysfs only
For the case when we print device name associated with struct device
that was not found in /dev, but in sysfs, for example when printing
devices where LV device mismatch is found.
2016-04-01 14:48:56 +02:00
Peter Rajnoha
91d32f9d1b refactor: dev-cache: move code adding sysfs-only device into separate fn 2016-04-01 11:47:06 +02:00
Peter Rajnoha
0e774d5ae7 refactor: dev-cache: use btree instead of hash table for sysfs-only devices
major:minor btree is more convenient and more suitable than dev name
hash table here.
2016-04-01 11:42:25 +02:00
Peter Rajnoha
9a086a6607 dev-cache: fix check for already indexed dev in _index_dev_by_vgid_and_lvid 2016-03-30 15:57:57 +02:00
Peter Rajnoha
8b258a005b dev-cache: dev_cache_index_devs fn is available unconditionally
The new dev_cache_index_devs fn was under ifdef UDEV_SYNC_SUPPORT by mistake,
move it out of this ifdef.
2016-03-30 13:06:20 +02:00
Peter Rajnoha
52e0d0db44 dev-cache: remove spurious error msg if no value found in /sys/dev/block/<major>:<minor>/dm/uuid during dev scan
It's correct to have a DM device that has no DM UUID assigned
so no need to issue error message in this case. Also, if the
device doesn't have DM UUID, it's also clear it's not an LVM LV
(...when looking for VGID/LVID while creating VGID/LVID indices
in dev cache).

For example:

$ dmsetup create test --table "0 1 linear /dev/sda 0"
And there's no PV in the system.

Before this patch (spurious error message issued):
$ pvs
  _get_sysfs_value: /sys/dev/block/253:2/dm/uuid: no value

With this patch applied (no spurious error message):
$ pvs
2016-03-30 11:30:09 +02:00
Peter Rajnoha
8c27c52749 dev-cache: also index VGIDs and LVIDs if using persistent .cache file
If we're using persistent .cache file, we're reading this file instead
of traversing the /dev content. Fix missing indexing by VGID and LVID
here - hook this into persistent_filter_load where we populate device
cache from persistent .cache file instead of scanning /dev.

For example, inducing situation in which we warn about different device
actually used than what LVM thinks should be used based on metadata:

$ lsblk -s /dev/vg/lvol0
NAME     MAJ:MIN RM  SIZE RO TYPE MOUNTPOINT
vg-lvol0 253:4    0  124M  0 lvm
`-loop1    7:1    0  128M  0 loop

$ lvmconfig --type diff

global {
	use_lvmetad=0
}
devices {
	obtain_device_list_from_udev=0
}

(obtain_device_list_from_udev=0 also means the persistent .cache file is used)

Before this patch - pvs is fine as it does the dev scan, but lvs relies
on persistent .cache file and it misses the VGID/LVID indices to check
and warn about incorrect devices used:

$ pvs
  Found duplicate PV B9gXTHkIdEIiMVwcOoT2LX3Ywh4YIHgR: using /dev/loop0 not /dev/loop1
  Using duplicate PV /dev/loop0 without holders, ignoring /dev/loop1
  WARNING: Device mismatch detected for vg/lvol0 which is accessing /dev/loop1 instead of /dev/loop0.
  PV          VG Fmt  Attr PSize   PFree
  /dev/loop0 vg lvm2 a--  124.00m    0

$ lvs
  Found duplicate PV B9gXTHkIdEIiMVwcOoT2LX3Ywh4YIHgR: using /dev/loop0 not /dev/loop1
  Using duplicate PV /dev/loop0 without holders, ignoring /dev/loop1
  LV    VG Attr       LSize
  lvol0 vg -wi-a----- 124.00m

With this patch applied - both pvs and lvs is fine - the indices are
always created correctly (lvs just an example here, other LVM commands
that rely on persistent .cache file are fixed with this patch too):

$ pvs
  Found duplicate PV B9gXTHkIdEIiMVwcOoT2LX3Ywh4YIHgR: using /dev/loop0 not /dev/loop1
  Using duplicate PV /dev/loop0 without holders, ignoring /dev/loop1
  WARNING: Device mismatch detected for vg/lvol0 which is accessing /dev/loop1 instead of /dev/loop0.
  PV          VG Fmt  Attr PSize   PFree
  /dev/loop0 vg lvm2 a--  124.00m    0

$ lvs
  Found duplicate PV B9gXTHkIdEIiMVwcOoT2LX3Ywh4YIHgR: using /dev/loop0 not /dev/loop1
  Using duplicate PV /dev/loop0 without holders, ignoring /dev/loop1
  WARNING: Device mismatch detected for vg/lvol0 which is accessing /dev/loop1 instead of /dev/loop0.
  LV    VG Attr       LSize
  lvol0 vg -wi-a----- 124.00m
2016-03-30 11:00:01 +02:00
Peter Rajnoha
91bb202ded dev-cache: handle situation where device is referenced in sysfs, but the node is not yet in dev dir
It's possible that while a device is already referenced in sysfs, the node
is not yet in /dev directory.

This may happen in some rare cases right after LVs get created - we sync
with udev (or alternatively we create /dev content ourselves) while VG
lock is held. However, dev scan is done without VG lock so devices may
already be in sysfs, but /dev may not be updated yet if we call LVM command
right after LV creation (so the fact that fs_unlock is done within VG
lock is not usable here much). This is not a problem with devtmpfs as
there's at least kernel name for device in /dev as soon as the sysfs
item exists, but we still support environments without devtmpfs or
where different directory for dev nodes is used (e.g. our test suite).

This patch covers these situations by tracking such devices in
_cache.sysfs_only_names helper hash for the vgid/lvid check to work still.

This also resolves commit 6129d2e64d
which was then reverted by commit 109b7e2095
due to performance issues it may have brought (...and it didn't resolve
the problem fully anyway).
2016-03-30 10:56:46 +02:00