Currently we may have work scheduled in default kernel workqueue when
the device is going down. The device could get freed before this
workqueue gets serviced. I am actually seeing this causing system
hangs.
The following patch fixes this by using ipoib_workqueue which gets
flushed when the device is going down.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Fix deadlock condition resulting from trying to destroy a cm_id
from the context of a CM thread. The synchronization around the
ucm context structure is simplified as a result, and some simple
code cleanup is included.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Always make sure that the full membership bit is set in the P_Keys
that IPoIB uses. This makes sure that all hosts join the correct
multicast groups so that hosts that are partial partition members
can talk to the rest of the network.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
When creating a table in context memory where the table is smaller
than our chunk size, we don't want to allocate and map a full chunk.
Instead, allocate just enough memory to cover the table.
This can be pretty simple because all tables are a power-of-2 size, so
either the table is a multiple of the chunk size, or it's smaller than
one chunk.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Move the definitions of the WQE structures from mthca_qp.c into
mthca_wqe.h, so that we'll be able to share them when we add the
SRQ code in mthca_srq.c.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Mem-free HCAs never generate error CQEs that complete multiple WQEs,
so just skip the call to mthca_free_err_wqe() for them rather than
having logic to handle the mem-free case in mthca_free_err_wqe().
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Clean up the allocation of memory for queues by factoring out the
common code into mthca_buf_alloc() and mthca_buf_free(). Now CQs and
QPs share the same queue allocation code, which we'll also use for SRQs.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add SRQ support to userspace verbs module. This adds several commands
and associated structures, but it's OK to do this without bumping the
ABI version because the commands are added at the end of the list so
they don't change the existing numbering. There are two cases to
worry about:
1. New kernel, old userspace. This is OK because old userspace simply
won't try to use the new SRQ commands. None of the old commands are
changed.
2. Old kernel, new userspace. This works perfectly as long as
userspace doesn't try to use SRQ commands. If userspace tries to
use SRQ commands, it will get EINVAL, which is perfectly
reasonable: the kernel doesn't support SRQs, so we couldn't do any
better.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Set the max_msg_sz port property correctly in mthca's port_query
function. Also zero out the attr struct so that we don't leave
any other members uninitialized.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
When we call the INIT_IB firmware command to bring up a port, use
the actual port width capability returned by the QUERY_DEV_LIM
command instead of always trying to enable both 1X and 4X. This
fixes breakage seen when the firmware is build to allow 4X only.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
changing CONFIG_LOCALVERSION rebuilds too much, for no appearent reason.
Remove unneeded includes of <linux/version.h>.
Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Change ib_mad_thread_completion_handler to conform to ib_comp_handler
declaration.
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Use the generic key_to_hw_index() function instead of the Arbel-specific
version in mthca_free_region().
Signed-off-by: Guy German <guyg@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Make sure that all FMRs are unmapped before we deallocate them so that
we don't leak references to our protection domain when destroying an
FMR pool. (Bug reported by Guy German <guyg@voltaire.com>)
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add support for reporting HCA board ID returned from QUERY_ADAPTER
firmware command through sysfs.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Make some lawyers happy and add copyright notices for people who
forgot to include them when they actually touched the code.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Update FW versions in mthca according to July 05 Mellanox release
Signed-off-by: Tziporet Koren <tziporet@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The problem arises if an entity in sysfs is created and removed without
ever having been made completely visible. In SCSI this is triggered by
removing a device while it's initialising.
The problem appears to be that because it was never made visible in sysfs,
the sysfs dentry has a null d_inode which oopses when a reference is made
to it. The solution is simply to check d_inode and assume the object was
never made visible (and thus doesn't need deleting) if it's NULL.
(akpm: possibly a stopgap for 2.6.13 scsi problems. May not be the
long-term fix)
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
It's possible for this to still have flags in it and a previous instance
has been stopped, and that confused the new array using the same mddev.
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I just discovered this is needed for module auto-loading.
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Fix a use-after-free bug in userspace verbs cleanup: we can't touch
mr->device after we free mr by calling ib_dereg_mr().
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We are currently reserving one byte more than actually needed by the flash
device and overlapping into the next I/O expansion bus window. This a)
causes us to allocate an extra page of VM due to ARM ioremap() alignment
code and b) could cause problems if another driver tries to request the
next expansion bus window.
Signed-off-by: Deepak Saxena <dsaxena@plexity.net>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Some nodes can have large holes on x86-64.
This fixes problems with the VM allowing too many dirty pages because it
overestimates the number of available RAM in a node. In extreme cases you
can end up with all RAM filled with dirty pages which can lead to deadlocks
and other nasty behaviour.
This patch just tells the VM about the known holes from e820. Reserved
(like the kernel text or mem_map) is still not taken into account, but that
should be only a few percent error now.
Small detail is that the flat setup uses the NUMA free_area_init_node() now
too because it offers more flexibility.
(akpm: lotsa thanks to Martin for working this problem out)
Cc: Martin Bligh <mbligh@mbligh.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch fixes several instances of hwmon drivers kfree'ing the "wrong"
pointer; the existing code works somewhat by accident.
(akpm: plucked from Greg's queue based on lkml discussion. Finishes off the
patch from Jon Corbet)
Signed-off-by: Mark M. Hoffman <mhoffman@lightlink.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I recently had a BUG_ON() go off spuriously on a gcc 4.0 compiled kernel.
It turns out gcc-4.0 was removing a sign extension while earlier gcc
versions would not. Thinking this to be a compiler bug, I submitted a
report:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23422
It turns out we need to cast the input in order to tell gcc to sign extend
it.
Thanks to Andrew Pinski for his help on this bug.
Signed-off-by: Anton Blanchard <anton@samba.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
At the suggestion of Nick Piggin and Dinakar, totally disable
the facility to allow cpu_exclusive cpusets to define dynamic
sched domains in Linux 2.6.13, in order to avoid problems
first reported by John Hawkes (corrupt sched data structures
and kernel oops).
This has been built for ppc64, i386, ia64, x86_64, sparc, alpha.
It has been built, booted and tested for cpuset functionality
on an SN2 (ia64).
Dinakar or Nick - could you verify that it for sure does avoid
the problems Hawkes reported. Hawkes is out of town, and I don't
have the recipe to reproduce what he found.
Signed-off-by: Paul Jackson <pj@sgi.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The partial disabling of Dinakar's new facility to allow
cpu_exclusive cpusets to define dynamic sched domains
doesn't go far enough. At the suggestion of Nick Piggin
and Dinakar, let us instead totally disable this facility
for 2.6.13, in order to avoid problems first reported
by John Hawkes (corrupt sched data structures and kernel oops).
This patch removes the partial disabling code in 2.6.13-rc7,
in anticipation of the next patch, which will totally disable
it instead.
Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Coverity uncovered an off-by-one error in the fscpos driver, in function
set_temp_reset(). Writing to the temp3_reset sysfs file will lead to an
array overrun, in turn causing an I2C write to a random register of the
FSC Poseidon chip. Additionally, writing to temp1_reset and temp2_reset
will not work as expected. The fix is straightforward.
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Be more precise on deciding whether to call m8xx_ide_init() at
m8xx_setup.c:platform_init().
Compilation fails if CONFIG_BLK_DEV_IDE is defined but
CONFIG_BLK_DEV_MPC8xx_IDE isnt.
Signed-off-by: Marcelo Tosatti <marcelo.tosatti@cyclades.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
spinlock used in irq handler should be initialized before registering
irq, even if we know that our device has interrupts disabled; handler
is registered shared and taking spinlock is done unconditionally. As
it is, we can and do get oopsen on boot for some configuration, depending
on irq routing - I've got a reproducer.
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Ben Collins <bcollins@debian.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
In qdio_get_micros() volatile in return type is plain noise (even with old
gccisms it would make no sense - noreturn function returning __u64 is a
bit odd ;-)
Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The adm9240 driver, in adm9240_detect(), allocates a structure. The
error path attempts to kfree() ->client field of it (second one),
resulting in an oops (or slab corruption) if the hardware is not present.
->client field in adm1026, adm1031, smsc47b397 and smsc47m1 is the first in
${HWMON}_data structure, but fix them too.
Signed-off-by: Jonathan Corbet <corbet@lwn.net
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The recent change to locks_remove_flock code in fs/locks.c changes how
byte range locks are removed from closing files, which shows up a bug in
cifs.
The assumption in the cifs code was that the close call sent to the
server would remove any pending locks on the server on this file, but
that is no longer safe as the fs/locks.c code on the client wants unlock
of 0 to PATH_MAX to remove all locks (at least from this client, it is
not possible AFAIK to remove all locks from other clients made to the
server copy of the file).
Note that cifs locks are different from posix locks - and it is not
possible to map posix locks perfectly on the wire yet, due to
restrictions of the cifs network protocol, even to Samba without adding
a new request type to the network protocol (which we plan to do for
Samba 3.0.21 within a few months), but the local client will have the
correct, posix view, of the lock in most cases.
The correct fix for cifs for this would involve a bigger change than I
would like to do this late in the 2.6.13-rc cycle - and would involve
cifs keeping track of all unmerged (uncoalesced) byte range locks for
each remote inode and scanning that list to remove locks that intersect
or fall wholly within the range - locks that intersect may have to be
reaquired with the smaller, remaining range.
Signed-off-by: Steve French <sfrench@us.ibm.com>
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
While touching this code I noticed the error handling is bogus, so I
fixed it up.
I've removed the IS_ERR(proc_dentry) check, which will never trigger and
is clearly a typo: we must check proc_file instead.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Update hppfs for the symlink functions prototype change.
Yes, I know the code I leave there is still _bogus_, see next patch for
this.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There is an off by one problem with idr_get_new_above.
The comment and function name suggest that it will return an id >
starting_id, but it actually returned an id >= starting_id, and kernel
callers other than inotify treated it as such.
The patch below fixes the comment, and fixes inotifys usage. The
function name still doesn't match the behaviour, but it never did.
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Writing even a disabled value seems to mess up some matrox graphics
cards. It may be a card-related issue, but we may also be writing
reserved low bits in the result.
This was a fall-out of switching x86 over to the generic PCI resource
allocation code, and needs more debugging. In particular, the old x86
code defaulted to not doing any resource allocations at all for ROM
resources.
In the meantime, this has been reported to make X happier by Helge
Hafting <helgehaf@aitel.hist.no>.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
It may seem small, but most cards need much less, if any, and this not
only makes the code adhere to the comment, it seems to fix a boot-time
lockup on a ThinkPad 380XD laptop reported by Tero Roponen <teanropo@cc.jyu.fi>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The tg3_abort_hw() call in tg3_test_loopback() is causing lockups on
some devices. tg3_abort_hw() disables the memory arbiter, causing
tg3_reset_hw() to hang when it tries to write the pre-reset signature.
tg3_abort_hw() should only be called after the pre-reset signature has
been written. This is all done in tg3_reset_hw() so the tg3_abort_hw()
call is unnecessary and can be removed.
[ Also bump driver version and release date. -DaveM ]
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
One critical fix and two minor fixes for 2.6.13-rc7:
- Max depth must currently be 2 to allow barriers to function on SCSI
- Prefer sync request over async in choosing the next request
- Never allow async request to preempt or disturb the "anticipation" for
a single cfq process context. This is as-designed, the code right now
is buggy in that area.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
pcibios_bus_to_resource is exported on all architectures except ia64
and sparc. Add exports for the two missing architectures. Needed when
Yenta socket support is compiled as a module.
Signed-off-by: Keith Owens <kaos@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I mistakedly disabled fusion support in an earlier update. Fusion
is commonly used on many x86-64 systems, so this was a problem.
This patch fixes that.
Signed-off-by: And Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch puts back the export of machine_power_off() that was removed
by some janitor as it's used for emergency shutdown by the G5 thermal
control driver. Wether that driver should use kernel_power_off() instead
is debatable and a post-2.6.13 decision. In the meantime, please commit
that patch that fixes the driver for now.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Michael S. Tsirkin