Lots of people are having trouble with the default timeouts, which are too
low. These new values are derived from an informal survey taken on
ocfs2-users, as well as data from bug reports. This should reduce the amount
of cluster disconnects and subsequent fencing seen during normal workloads.
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Explicitely convert loff_t to long long in printf. Just for sure...
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
We should use generic_file_llseek() and not default_llseek() so that
s_maxbytes gets properly checked when seeking.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
In ocfs2_read_inline_data() we should store file size in loff_t. Although
the file size should fit in 32 bits we cannot be sure in case filesystem is
corrupted.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Create separate lockdep lock classes for system file's i_mutexes. They are
used to guard allocations and similar things and thus rank differently
than i_mutex of a regular file or directory.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Hook up ocfs2_flock(), using the new flock lock type in dlmglue.c. A new
mount option, "localflocks" is added so that users can revert to old
functionality as need be.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
This adds a new dlmglue lock type which is intended to back flock()
requests.
Since these locks are driven from userspace, usage rules are much more
liberal than the typical Ocfs2 internal cluster lock. As a result, we can't
make use of most dlmglue features - lock caching and lock level
optimizations in particular. Additionally, userspace is free to deadlock
itself, so we have to deal with that in the same way as the rest of the
kernel - by allowing a signal to abort a lock request.
In order to keep ocfs2_cluster_lock() complexity down, ocfs2_file_lock()
does it's own dlm coordination. We still use the same helper functions
though, so duplicated code is kept to a minimum.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Local alloc is a performance optimization in ocfs2 in which a node
takes a window of bits from the global bitmap and then uses that for
all small local allocations. This window size is fixed to 8MB currently.
This patch allows users to specify the window size in MB including
disabling it by passing in 0. If the number specified is too large,
the fs will use the default value of 8MB.
mount -o localalloc=X /dev/sdX /mntpoint
Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Mostly taken from ext3. This allows the user to set the jbd commit interval,
in seconds. The default of 5 seconds stays the same, but now users can
easily increase the commit interval. Typically, this would be increased in
order to benefit performance at the expense of data-safety.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Check that an online resize is being driven by a user with permission to
change system resource limits.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
This patch adds the ability for a userspace program to request that a
properly formatted cluster group be added to the main allocation bitmap for
an Ocfs2 file system. The request is made via an ioctl, OCFS2_IOC_GROUP_ADD.
On a high level, this is similar to ext3, but we use a different ioctl as
the structure which has to be passed through is different.
During an online resize, tunefs.ocfs2 will format any new cluster groups
which must be added to complete the resize, and call OCFS2_IOC_GROUP_ADD on
each one. Kernel verifies that the core cluster group information is valid
and then does the work of linking it into the global allocation bitmap.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
This patch adds the ability for a userspace program to request an extend of
last cluster group on an Ocfs2 file system. The request is made via ioctl,
OCFS2_IOC_GROUP_EXTEND. This is derived from EXT3_IOC_GROUP_EXTEND, but is
obviously Ocfs2 specific.
tunefs.ocfs2 would call this for an online-resize operation if the last
cluster group isn't full.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
This patch sets the last_sector_bug flag to 1 for all USB disks. This is
needed to makes the cardreader on various HP multifunction printers work.
Since the performance impact is negible we set this flag for all USB disks to
avoid an unusual_devs.h nightmare.
Signed-off-by: Hans de Goede <j.w.r.degoede@hhs.nl>
Acked-by: Matthew Dharm <mdharm-usb@one-eyed-alien.net>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
We need to reserve at least two ioctls for online-resize. Reserve a small
range of ioctls for Ocfs2 use in Documentation/ioctl-number.txt. This should
give us enough room for future growth.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
This value is initialized from global_bitmap->id2.i_chain.cl_cpg. If there
is only 1 group, it will be equal to the total clusters in the volume. So
as for online resize, it should change for all the nodes in the cluster.
It isn't easy and there is no corresponding lock for it.
bitmap_cpg is only used in 2 areas:
1. Check whether the suballoc is too large for us to allocate from the global
bitmap, so it is little used. And now the suballoc size is 2048, it rarely
meet this situation and the check is almost useless.
2. Calculate which group a cluster belongs to. We use it during truncate to
figure out which cluster group an extent belongs too. But we should be OK
if we increase it though as the cluster group calculated shouldn't change
and we only ever have a small bitmap_cpg on file systems with a single
cluster group.
Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Currently it is possible to select HW_RANDOM as a module and have
hifn795x built-in. This causes a build problem because hifn795x
will then call hwrng_register which isn't built-in.
This patch introduces a new config option to control the hifn795x
RNG which lets us avoid this problem.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Remove 'readpages' from the list in ocfs2.txt. Instead of having two
identical lists, I just removed the list in the OCFS2 section of fs/Kconfig
and added a pointer to Documentation/filesystems/ocfs2.txt.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Add ->readpages support to Ocfs2. This is rather trivial - all it required
is a small update to ocfs2_get_block (for mapping full extents via b_size)
and an ocfs2_readpages() function which partially mirrors ocfs2_readpage().
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
This allows others to use the DLM constants without being tied to the
function API of fs/dlm.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Call this the "inode_lock" now, since it covers both data and meta data.
This patch makes no functional changes.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
The meta lock now covers both meta data and data, so this just removes the
now-redundant data lock.
Combining locks saves us a round of lock mastery per inode and one less lock
to ping between nodes during read/write.
We don't lose much - since meta locks were always held before a data lock
(and at the same level) ordered writeout mode (the default) ensured that
flushing for the meta data lock also pushed out data anyways.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
In order to extend inode lock coverage to inode data, we use the same data
downconvert worker with only a small modification to only do work for
regular files.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
The node maps that are set/unset by these votes are no longer relevant, thus
we can remove the mount and umount votes. Since those are the last two
remaining votes, we can also remove the entire vote infrastructure.
The vote thread has been renamed to the downconvert thread, and the small
amount of functionality related to managing it has been moved into
fs/ocfs2/dlmglue.c. All references to votes have been removed or updated.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (81 commits)
RDMA/cxgb3: Fix the T3A workaround checks
IB/ipath: Remove unnecessary cast
IPoIB: Constify seq_operations function pointer tables
RDMA/cxgb3: Mark QP as privileged based on user capabilities
RDMA/cxgb3: Fix page shift calculation in build_phys_page_list()
RDMA/cxgb3: Flush the receive queue when closing
IB/ipath: Trivial simplification of ipath_make_ud_req()
IB/mthca: Update latest "native Arbel" firmware revision
IPoIB: Remove redundant check of netif_queue_stopped() in xmit handler
IB/ipath: Add mappings from HW register to PortInfo port physical state
IB/ipath: Changes to support PIO bandwidth check on IBA7220
IB/ipath: Minor cleanup of unused fields and chip-specific errors
IB/ipath: New sysfs entries to control 7220 features
IB/ipath: Add new chip-specific functions to older chips, consistent init
IB/ipath: Remove unused MDIO interface code
IB/ehca: Prevent RDMA-related connection failures on some eHCA2 hardware
IB/ehca: Add "port connection autodetect mode"
IB/ehca: Define array to store SMI/GSI QPs
IB/ehca: Remove CQ-QP-link before destroying QP in error path of create_qp()
IB/iser: Add change_queue_depth method
...
Now that the dlm exposes domain information to us, we don't need generic
node up / node down callbacks. And since the DLM is only telling us when a
node goes down unexpectedly, we no longer need to optimize away node down
callbacks via the umount map.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
With this, a dlm client can take advantage of the group protocol in the dlm
to get full notification whenever a node within the dlm domain leaves
unexpectedly.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Correctly work around T3A issues by checking "hwtype != T3A" instead of
"hwtype == T3B". This will be needed for new hardware types.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The existing logic incorrectly maps this buffer list:
0: addr 0x10001000, size 0x1000
1: addr 0x10002000, size 0x1000
To this bogus page list:
0: 0x10000000
1: 0x10002000
The shift calculation must also take into account the address of the
first entry masked by the page_mask as well as the last address+size
rounded up to the next page size.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
- for kernel mode cqs, call event notification handler when flushing.
- flush QP when moving from RTS -> CLOSING.
- fix logic to identify a kernel mode qp.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Move the increment of s_hdrwords into the existing if block that tests
if we're doing a send with immediate, to save one test of the opcode.
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
qdisc_run() now tests for queue_stopped() before calling
__qdisc_run(), and the same check is done in every iteration of
__qdisc_run(), so another check is not required in the driver xmit.
This means that ipoib_start_xmit() no longer needs to test
netif_queue_stopped(); the test was added to fix earlier kernels,
where the networking stack did not guarantee that the xmit method of
an LLTX driver would not be called after the queue was stopped, but
current kernels do provide this guarantee.
To validate, I put a debug in the TX_BUSY path which never hit with 64
threads running overnight exercising this code a few 100 million
times.
Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add new mappings from port physical state (a HW register value) to the
IB SubnGet(PortInfo) port physical state.
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The IBA7220 uses a count-based triggering mechanism, and therefore
can't use the same bandwidth verification mechanism as older chips.
To support the 7220, allow enabling and disabling armlaunch errors on
application request. Minor robustness improvements as well.
Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Clean up some unused header fields, minor related cleanup.
Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
IBA7220 includes many more configurable IB settings. Getting/setting
these is now grouped into a pair of chip specific functions accessed via
function pointers. Provide sysfs access to these settings.
Signed-off-by: Michael Albaugh <michael.albaugh@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This adds the new (sometimes empty) chip-specific functions to the older
chips, and makes the initialization and related functions consistent across
all 3 chips.
Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This code has been unused for some time, but still had leftovers
from when it was used.
Signed-off-by: Dave Olson <dave.olson@qlogic.com
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Some HW revisions of eHCA2 may cause an RC connection to break if they
received RDMA Reads over that connection before. This can be
prevented by assuring that, after the first RDMA Read, the QP receives
a new RDMA Read every few million link packets.
Include code into the driver that inserts an empty (size 0) RDMA Read
into the message stream every now and then if the consumer doesn't
post them frequently enough.
Signed-off-by: Joachim Fenkes <fenkes@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This patch enhances ehca with a capability to "autodetect" the ports
being connected physically. In order to utilize that function the
module option nr_ports must be set to -1 (default is 2 - two
ports). This feature is experimental and will made the default later.
More detail:
If the user connects only one port to the switch, current code requires
1) port one to be connected and
2) module option nr_ports=1 to be given.
If autodetect is enabled, ehca will not wait at creation of the GSI QP
for the respective port to become active. Since firmware does not
accept modify_qp() while the port is down at initialization, we need
to cache all calls to modify_qp() for the SMI/GSI QP and just return a
good return code.
When a port is activated and we get a PORT_ACTIVE event, we replay the
cached modify-qp() parms and re-trigger any posted recv WRs. Only then
do we forward the PORT_ACTIVE event to registered clients.
The result of this autodetect patch is that all ports will be
accessible by the users. Depending on their respective cabling only
those ports that are connected properly will become operable. If a
user tries to modify a regular QP of a non-connected port, modify_qp()
will fail. Furthermore, ibv_devinfo should show the port state
accordingly.
Note that this patch primarily improves the loading behaviour of
ehca. If the cable is removed while the driver is operating and
plugged in again, firmware will handle that properly by sending an
appropriate async event.
Signed-off-by: Hoang-Nam Nguyen <hnguyen@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a .change_queue_depth handler to the scsi_host_template in the
iSER driver. iscsi_change_queue_depth was added to iscsi_tcp in order
to solve the problem of queue depth which was too high for some
targets. It is also applicable for iSER.
Signed-off-by: Erez Zilber <erezz@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Some RDMA CM events are not supported or not handled in iSER.
This patch adds some info (printk) for the user about them.
Signed-off-by: Erez Zilber <erezz@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
When a FMR is released via ib_fmr_pool_unmap(), the FMR usually ends
up on the free_list rather than the dirty_list (because we allow a
certain number of remappings before actually requiring a flush).
However, ib_fmr_batch_release() only looks at dirty_list when flushing
out old mappings. This means that when ib_fmr_pool_flush() is used to
force a flush of the FMR pool, some dirty FMRs that have not reached
their maximum remap count will not actually be flushed.
Fix this by flushing all FMRs that have been used at least once in
ib_fmr_batch_release().
Signed-off-by: Olaf Kirch <olaf.kirch@oracle.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Normally, the serial numbers for flush requests and flushes executed
for an FMR pool should be in sync.
However, if the FMR pool flushes dirty FMRs because the
dirty_watermark was reached, we wake up the cleanup thread and let it
do its stuff. As a side effect, the cleanup thread increments
pool->flush_ser, which leaves it one higher than pool->req_ser. The
next time the user calls ib_flush_fmr_pool(), the cleanup thread will
be woken up, but ib_flush_fmr_pool() won't wait for the flush to
complete because flush_ser is already past req_ser. This means the
FMRs that the user expects to be flushed may not have all been flushed
when the function returns.
Fix this by telling the cleanup thread to do work exclusively by
incrementing req_ser, and by moving the comparison of dirty_len and
dirty_watermark into ib_fmr_pool_unmap().
Signed-off-by: Olaf Kirch <olaf.kirch@oracle.com>
In addition to being overly complex, the locking in user_mad.c is
broken: there were multiple reports of deadlocks and lockdep warnings.
In particular it seems that a single thread may end up trying to take
the same rwsem for reading more than once, which is explicitly
forbidden in the comments in <linux/rwsem.h>.
To solve this, we change the locking to use plain mutexes instead of
rwsems. There is one mutex per open file, which protects the contents
of the struct ib_umad_file, including the array of agents and list of
queued packets; and there is one mutex per struct ib_umad_port, which
protects the contents, including the list of open files. We never
hold the file mutex across calls to functions like ib_unregister_mad_agent(),
which can call back into other ib_umad code to queue a packet, and we
always hold the port mutex as long as we need to make sure that a
device is not hot-unplugged from under us.
This even makes things nicer for users of the -rt patch, since we
remove calls to downgrade_write() (which is not implemented in -rt).
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Sunil Mushran