IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
With clusters larger than 3 nodes, the 32-byte debug buffer in
cpg_join_callback() is too small to contain all the node IDs, because
32-bit identifiers are generally rendered in 10 decimal digits. No fixed
size is good in all cases, but this is conditionally logged debug info,
so we can simply truncate it. Double the size, nevertheless.
Cmirrord has endian bugs, which cause failure to lvcreate a mirrored lv
on s390.
- data_size is uint32, should not use xlate64 to convert, which will
cause data_size 0 after xlate.
- request_type and data_size still used by local(v5_data_switch),
should convert later. If request_type xlate too early, it will
cause request_type judge error; if data_size xlate too early, it
will cause coredump in case DM_ULOG_CLEAR_REGION.
- when receiving package in clog_request_from_network. vp[0] will always
be little endian. We could use xlate64(vp[0]) == vp[0] to decide if
the local node is little endian or not.
Signed-off-by: Lidong Zhong<lzhong@suse.com> & Liuhua Wang <lwang@suse.com>
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
There are two types of CPG communications in a corosync cluster:
messages and state transitions. Cmirrord processes the state
transitions first.
When a cluster mirror issues a POSTSUSPEND, it signals the end of
cluster communication with the rest of the nodes in the cluster.
The POSTSUSPEND marks the last communication of the 'message'
type that will go around the cluster. The node then calls
cpg_leave which causes a final 'state transition' communication to
all of the nodes. Once the out-going node receives its own state
transition notice from the cluster, it finalizes the leave. At this
point, the state of the log is 'INVALID'; but it is possible that
there remains some cluster trafic that was queued up behind the
state transition that still wants to be processed. It is harmless
to attempt to dispatch any remaining messages - they won't be
delivered because the node is no longer in the cluster. However,
there was a warning message that was being printed in this case
that is now removed by this patch. The failure of the dispatch
created a false positive condition that triggered the message.
cmirrord polls for messages on the kernel and cluster interfaces.
Sometimes it is possible for messages to be received on the cluster
interface and be waiting for processing while the node is in the
process of leaving the cluster group. When this happens, the
messages received on the cluster interface are attempted to be
dispatched, but an error is returned because the connection is no
longer valid. It is a harmless situation. So, if we get the
specific error (CS_ERR_BAD_HANDLE) and we know that we have left
the group, then simply don't print the message.
The PIE and RELRO compiler/linker options can be used to produce a code
some techniques applied that makes the code more immune to some attacks:
- PIE (Position Independent Executable). It can make use of the ASLR
(Address Space Layout Randomization) provided by kernel to avoid
static locations for .text regions of executables (this is the 'pie'
compiler and linker option)
- RELRO (Relocation Read-Only). This prevents overwrite attacks of
the GOT (Global Offset Table) and PLT (Procedure Lookup Table)
used for relocations by making it read-only after all relocations
are resolved (these are the 'relro' and 'now' linker options) -
hence all symbols are resolved at the very start so there's no
need for those tables to be writeable later.
These compiler/linker options are now used by default for daemons
if the compiler/linker supports it.
The bug addressed by this patch manifested itself during testing
by showing a mirror that never became 'in-sync' after creation.
The bug is isolated to distributions that do not have support
for openAIS checkpointing (i.e. > RHEL6, > F16).
When a node joins a group that is managing a mirror log, the other
machines in the group send it a checkpoint representing the current
state of the bitmap. More than one machine can send a checkpoint,
but only the initial one should be imported. Once the bitmap state
has been imported from the initial checkpoint, operations (such
as resync, mark, and clear operations) can begin. When subsequent
checkpoints are allowed to be imported, it has the effect of erasing
all the log operations between the initial checkpoint and the ones
that follow.
When cmirrord was updated to handle the absence of openAIS
checkpointing (commit 62e38da133),
the new import_checkpoint() function failed to honor the 'no_read'
parameter. This parameter was designed to avoid reading all but
the initial checkpoint. Honoring this parameter has solved the
issue of corrupting bitmap data with secondary checkpoints.
Looking at the code in cmirrord/local.c, we can see the various different
request types handled in different ways. Some information that is non-changing
does not need to go around the cluster and can be short-circuited. For
example, once the cluster mirror is in-sync, it is pointless to continue
sending that query around the cluster. We can save network bandwidth and reply
directly back to the kernel. When it comes to status information, there are
two types 'TABLE' and 'INFO'. The 'TABLE' information never changes and
belongs to the group of requests that can be safely short-circuited. The
'STATUS' information can change - and will change if a device fails. Thus it
cannot be short-circuited, but this is exactly what was found. The 'STATUS'
information request was being short-circuited and therefore never reporting the
failure condition to anyone other than the "server" that experienced it
directly.
F17 is getting rid of OpenAIS libraries (and checkpointing). While the
CPG stuff is staying, some if its constants are being removed. So, we
must adjust and use the remaining constants which the CPG constants were based on.
[~]# egrep 'CPG_DISPATCH_ALL|CPG_OK' /usr/include/*/*
corosync/corotypes.h:#define CPG_DISPATCH_ALL CS_DISPATCH_ALL
corosync/corotypes.h:#define CPG_OK CS_OK
The OpenAIS checkpoint library is going away; therefore, cmirrord must
operate without it. The algorithms the handle the timing of when to send
a checkpoint, the determination of what to send, and which ongoing cluster
requests are relevent with respect to the checkpoints are unaffected. We
need only replace the functions that actually perform the storing/transmitting
and retrieving/receiving of the checkpoint data. Rather than store the
checkpoint data in an OpenAIS checkpoint file, we simply transmit it along
with the message that notifies the incoming node that the checkpoint is
ready.
Version 2 of the userspace log protocol accepts return information during the
DM_ULOG_CTR exchange. The return information contains the name of the log
device that is being used (if there is one). The kernel can then register the
device via 'dm_get_device'. Amoung other things, this allows for userspace to
assemble a correct dependency tree of devices - critical for LVM handling of
suspend/resume calls.
Also, update dm-log-userspace.h to match the kernel header associated with
this protocol change. (Includes a version inc.)
(since data in statbuf are invalid).
Check whether sysconf managed to find _SC_PAGESIZE.
Report at least debug warning about failing unlink
(logging scheme here seems to be a different then in lvm).
Duplicate terminal FDs and use similar code as is made in clvmd
and cleanup warns about missing open/close tests.
FIXME: Looks like we already have 3 instancies of the same code in lvm repo.
Detect existence of new SELinux selabel interface during configure.
Use new dm_prepare_selinux_context instead of dm_set_selinux_context.
We should set the SELinux context before the actual file system object creation.
The new dm_prepare_selinux_context function sets this using the selabel_lookup
fn in conjuction with the setfscreatecon fn. If selinux/label.h interface
(that should be a part of the selinux library) is not found during configure,
we fallback to the original matchpathcon function instead.
Function pull_stateo() checks for NULL 'buf' - but return for this error
path was missing. cmirror code never calls this function with NULL 'buf',
so this fix has no effect on current code base, but makes clang happier.
can be opprobriously slow if created with '--nosync'.
One of the ways cluster mirrors coordinate I/O and recovery
amoung the different machines is by the use of the log
function 'is_remote_recovering()' which lets nodes know if
a region they wish to perform a write on is currently being
recovered on another node. If the region is being recovered,
the I/O is delayed.
The 'is_remote_recovering' routine has been optimized to
avoid the deluge of requests that would be issued to the
userspace log server by maintaining a marker of how far
the recovery has gotten. It can then immediately return
'not recovering' if the region being inquired about is
less than this mark. Additionally, if the region of
concern is greater than the mark, the function will
limit the number of transmissions to userspace by assuming
the region /is/ being recovered when skipping the
transmission. This limits the amount of processing
and updates the mark in 1/4 sec time steps.
This patch fixes a problem where 'the mark' is not being
updated because of faulty logic in the userspace log
daemon. When '--nosync' is used to create a cluster
mirror, the userspace log daemon never has a chance
to update the mark in the normal way. The fix is to set
the mark to "complete" if the mirror was created with
the --nosync flag.
The lvm repair issues I believe are the superficial symptoms of this
bug - there are worse issues that are not as clearly seen. From my
inline comments:
* If the mirror was successfully recovered, we want to always
* force every machine to write to all devices - otherwise,
* corruption will occur. Here's how:
* Node1 suffers a failure and marks a region out-of-sync
* Node2 attempts a write, gets by is_remote_recovering,
* and queries the sync status of the region - finding
* it out-of-sync.
* Node2 thinks the write should be a nosync write, but it
* hasn't suffered the drive failure that Node1 has yet.
* It then issues a generic_make_request directly to
* the primary image only - which is exactly the device
* that has suffered the failure.
* Node2 suffers a lost write - which completely bypasses the
* mirror layer because it had gone through generic_m_r.
* The file system will likely explode at this point due to
* I/O errors. If it wasn't the primary that failed, it is
* easily possible in this case to issue writes to just one
* of the remaining images - also leaving the mirror inconsistent.
*
* We let in_sync() return 1 in a cluster regardless of what is
* in the bitmap once recovery has successfully completed on a
* mirror. This ensures the mirroring code will continue to
* attempt to write to all mirror images. The worst that can
* happen for reads is that additional read attempts may be
* taken.
corruption bug in cmirror. 'dm_bit' is only ever used as a boolean operation
within LVM, but it can return a range of values. If the bit is set, a power of
2 is returned. If the bit is unset, 0 is returned.
'log_test_bit' (a function in the cluster mirror log daemon code) has switched
to using the dm bit operations in rhel6. There are two places in the daemon
code where 'log_test_bit' is not used merely as a boolean, but rather the
return value is used as the return value for the log functions 'is_clean' and
'in_sync' - having assumed that 'dm_bit' was returning 0 or 1 only.
One place the 'in_sync' function is utilized is in 'dm_rh_get_state' - a
function that informs the mirroring code how to treat I/O and which devices to
read/write from. 'dm_rh_get_state' was checking if the return value of
'in_sync' was 1 to determine if the region was DM_RH_CLEAN. Since 'dm_bit'
(and by extension 'log_test_bit' and 'in_sync') was returning powers of 2,
DM_RH_CLEAN was rarely being reported as it should have been. Thinking the
region was out-of-sync, the mirroring code would write only to the primary
device. When the primary device was failed, all of those writes were lost -
leaving the entire mirror corrupted.
Switch dmeventd to use dm_create_lockfile and drop duplicate code.
Allow clvmd pidfile to be configurable.
Switch cmirrord and clvmd to use dm_create_lockfile.
linux/kdev_t.h even though it wasn't needed. Strangely, it seems
to be causing problems on various architectures (i686) in the
function daemons/cmirrord/functions.c:disk_status_info()->sprintf.
I'm not sure why this is a problem since none of the macros in
kdev_t.h are used in that code, but it certainly doesn't hurt to
pull an unnecessary header and it seems to fix the problem.
