* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx:
dmaengine: fix interrupt clearing for mv_xor
missing inline keyword for static function in linux/dmaengine.h
dma/shdma: move dereference below the NULL check
I have been seeing occasional pauses in transaction throughput up to
30s long under heavy parallel workloads. The only notable thing was
that the xfsaild was trying to be active during the pauses, but
making no progress. It was running exactly 20 times a second (on the
50ms no-progress backoff), and the number of pushbuf events was
constant across this time as well. IOWs, the xfsaild appeared to be
stuck on buffers that it could not push out.
Further investigation indicated that it was trying to push out inode
buffers that were pinned and/or locked. The xfsbufd was also getting
woken at the same frequency (by the xfsaild, no doubt) to push out
delayed write buffers. The xfsbufd was not making any progress
because all the buffers in the delwri queue were pinned. This scan-
and-make-no-progress dance went one in the trace for some seconds,
before the xfssyncd came along an issued a log force, and then
things started going again.
However, I noticed something strange about the log force - there
were way too many IO's issued. 516 log buffers were written, to be
exact. That added up to 129MB of log IO, which got me very
interested because it's almost exactly 25% of the size of the log.
He delayed logging code is suppose to aggregate the minimum of 25%
of the log or 8MB worth of changes before flushing. That's what
really puzzled me - why did a log force write 129MB instead of only
8MB?
Essentially what has happened is that no CIL pushes had occurred
since the previous tail push which cleared out 25% of the log space.
That caused all the new transactions to block because there wasn't
log space for them, but they kick the xfsaild to push the tail.
However, the xfsaild was not making progress because there were
buffers it could not lock and flush, and the xfsbufd could not flush
them because they were pinned. As a result, both the xfsaild and the
xfsbufd could not move the tail of the log forward without the CIL
first committing.
The cause of the problem was that the background CIL push, which
should happen when 8MB of aggregated changes have been committed, is
being held off by the concurrent transaction commit load. The
background push does a down_write_trylock() which will fail if there
is a concurrent transaction commit holding the push lock in read
mode. With 8 CPUs all doing transactions as fast as they can, there
was enough concurrent transaction commits to hold off the background
push until tail-pushing could no longer free log space, and the halt
would occur.
It should be noted that there is no reason why it would halt at 25%
of log space used by a single CIL checkpoint. This bug could
definitely violate the "no transaction should be larger than half
the log" requirement and hence result in corruption if the system
crashed under heavy load. This sort of bug is exactly the reason why
delayed logging was tagged as experimental....
The fix is to start blocking background pushes once the threshold
has been exceeded. Rework the threshold calculations to keep the
amount of log space a CIL checkpoint can use to below that of the
AIL push threshold to avoid the problem completely.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Alex Elder <aelder@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
In max8925_irq_sync_unlock(), irq control bit is set at the same time.
Zero means enabling irq, and one means disabling irq.
The original code is:
irq_chg[0] &= irq_data->enable;
It should be changed to:
irq_chg[0] &= ~irq_data->enable;
Otherwise, irq control bit is mess.
Signed-off-by: Kevin Liu <kliu5@marvell.com>
Signed-off-by: Haojian Zhuang <haojian.zhuang@marvell.com>
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
The driver was originally tested with an additional patch which
made this unneeded but that patch had issuges and got lost on the
way to mainline, causing problems when the errors are reported.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
Cc: stable@kernel.org
When caching is disabled on the MN10300 arch, the sys_cacheflush()
function is removed by conditional stuff in the makefiles, but is still
referred to by the syscall table.
Provide a null version that just returns 0 when caching is disabled (or
-EINVAL if the arguments are silly).
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Tssk. Apparently Al hadn't checked commit c52c2ddc1d ("alpha: switch
osf_sigprocmask() to use of sigprocmask()") at all. It doesn't compile.
Fixed as per suggestions from Michael Cree.
Reported-by: Michael Cree <mcree@orcon.net.nz>
Cc: Al Viro <viro@ftp.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
libata depends on scsi_host_template for module reference counting and
sht's should be owned by each low level driver. During libahci split,
the sht was left with libahci.ko leaving the actual low level drivers
not reference counted. This made ahci and ahci_platform always
unloadable even while they're being actively used.
Fix it by defining AHCI_SHT() macro in ahci.h and defining a sht for
each low level ahci driver.
stable: only applicable to 2.6.35.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Pedro Francisco <pedrogfrancisco@gmail.com>
Tested-by: Michael Tokarev <mjt@tls.msk.ru>
Cc: stable@kernel.org
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (47 commits)
tcp: Fix >4GB writes on 64-bit.
net/9p: Mount only matching virtio channels
de2104x: fix ethtool
tproxy: check for transparent flag in ip_route_newports
ipv6: add IPv6 to neighbour table overflow warning
tcp: fix TSO FACK loss marking in tcp_mark_head_lost
3c59x: fix regression from patch "Add ethtool WOL support"
ipv6: add a missing unregister_pernet_subsys call
s390: use free_netdev(netdev) instead of kfree()
sgiseeq: use free_netdev(netdev) instead of kfree()
rionet: use free_netdev(netdev) instead of kfree()
ibm_newemac: use free_netdev(netdev) instead of kfree()
smsc911x: Add MODULE_ALIAS()
net: reset skb queue mapping when rx'ing over tunnel
br2684: fix scheduling while atomic
de2104x: fix TP link detection
de2104x: fix power management
de2104x: disable autonegotiation on broken hardware
net: fix a lockdep splat
e1000e: 82579 do not gate auto config of PHY by hardware during nominal use
...
Commit e40cc4bdfd introduced
a build breakage if CONFIG_SMP is undefined. This commit
fixes the problem.
This fix is only a workaround. For a real fix, cpu_sibling_mask() should
be defined in UP include code, eg in linux/smp.h, and asm/smp.h should not be
included directly. This fix is currently not possible because asm/smp.h defines
cpu_sibling_mask() unconditionally and is included directly from many source
files.
Reported-by: Ingo Molnar <mingo@elte.hu>
Tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
* 'x86/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86: Avoid 'constant_test_bit()' misoptimization due to cast to non-volatile
Fixes kernel bugzilla #16603
tcp_sendmsg() truncates iov_len to an 'int' which a 4GB write to write
zero bytes, for example.
There is also the problem higher up of how verify_iovec() works. It
wants to prevent the total length from looking like an error return
value.
However it does this using 'int', but syscalls return 'long' (and
thus signed 64-bit on 64-bit machines). So it could trigger
false-positives on 64-bit as written. So fix it to use 'long'.
Reported-by: Olaf Bonorden <bono@onlinehome.de>
Reported-by: Daniel Büse <dbuese@gmx.de>
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The PKT_CTRL_CMD_STATUS device ioctl retrieves a pointer to a
pktcdvd_device from the global pkt_devs array. The index into this
array is provided directly by the user and is a signed integer, so the
comparison to ensure that it falls within the bounds of this array will
fail when provided with a negative index.
This can be used to read arbitrary kernel memory or cause a crash due to
an invalid pointer dereference. This can be exploited by users with
permission to open /dev/pktcdvd/control (on many distributions, this is
readable by group "cdrom").
Signed-off-by: Dan Rosenberg <dan.j.rosenberg@gmail.com>
[ Rather than add a cast, just make the function take the right type -Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The configuration choice for the port on which the GDB stub listens has
a default of GDBSTUB_TTYSM0, but this should be GDBSTUB_ON_TTYSM0 to
match the option.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
p9_virtio_create will only compare the the channel's tag characters
against the device name till the end of the channel's tag but not till
the end of the device name. This means that if a user defines channels
with the tags foo and foobar then he would mount foo when he requested
foonot and may mount foo when he requested foobar.
Thus it is necessary to check both string lengths against each other in
case of a successful partial string match.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the interface is up, using ethtool breaks it because:
a) link is put down but media_timer interval is not shortened to NO_LINK
b) rxtx is stopped but not restarted
Also manual 10baseT-HD (and probably FD too - untested) mode does not work -
the link is forced up, packets are transmitted but nothing is received.
Changing CSR14 value to match documentation (not disabling link check) fixes this.
Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
Acked-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
IPv4 and IPv6 have separate neighbour tables, so
the warning messages should be distinguishable.
[ Add a suitable message prefix on the ipv4 side as well -DaveM ]
Signed-off-by: Ulrich Weber <uweber@astaro.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When TCP uses FACK algorithm to mark lost packets in
tcp_mark_head_lost(), if the number of packets in the (TSO) skb is
greater than the number of packets that should be marked lost, TCP
incorrectly exits the loop and marks no packets lost in the skb. This
underestimates tp->lost_out and affects the recovery/retransmission.
This patch fargments the skb and marks the correct amount of packets
lost.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
* master.kernel.org:/home/rmk/linux-2.6-arm: (28 commits)
ARM: 6411/1: vexpress: set RAM latencies to 1 cycle for PL310 on ct-ca9x4 tile
ARM: 6409/1: davinci: map sram using MT_MEMORY_NONCACHED instead of MT_DEVICE
ARM: 6408/1: omap: Map only available sram memory
ARM: 6407/1: mmu: Setup MT_MEMORY and MT_MEMORY_NONCACHED L1 entries
ARM: pxa: remove pr_<level> uses of KERN_<level>
ARM: pxa168fb: clear enable bit when not active
ARM: pxa: fix cpu_is_pxa*() not expanding to zero when not configured
ARM: pxa168: fix corrected reset vector
ARM: pxa: Use PIO for PI2C communication on Palm27x
ARM: pxa: Fix Vpac270 gpio_power for MMC
ARM: 6401/1: plug a race in the alignment trap handler
ARM: 6406/1: at91sam9g45: fix i2c bus speed
leds: leds-ns2: fix locking
ARM: dove: fix __io() definition to use bus based offset
dmaengine: fix interrupt clearing for mv_xor
ARM: kirkwood: Unbreak PCIe I/O port
ARM: Fix build error when using KCONFIG_CONFIG
ARM: 6383/1: Implement phys_mem_access_prot() to avoid attributes aliasing
ARM: 6400/1: at91: fix arch_gettimeoffset fallout
ARM: 6398/1: add proc info for ARM11MPCore/Cortex-A9 from ARM
...
* git://git.infradead.org/iommu-2.6:
intel-iommu: Use symbolic values instead of magic numbers in Lenovo w/a
intel-iommu: Abort IOMMU setup for igfx if BIOS gave no shadow GTT space
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86/amd-iommu: Fix rounding-bug in __unmap_single
x86/amd-iommu: Work around S3 BIOS bug
x86/amd-iommu: Set iommu configuration flags in enable-loop
x86, setup: Fix earlyprintk=serial,0x3f8,115200
x86, setup: Fix earlyprintk=serial,ttyS0,115200
* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
perf, x86: Catch spurious interrupts after disabling counters
tracing/x86: Don't use mcount in kvmclock.c
tracing/x86: Don't use mcount in pvclock.c
It really has no business being there; short of a serious kernel bug
we should already have USER_DS at that point. It shouldn't have been
done on x86 either...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
set ->orig_d0 to -1, same as what sigreturn does
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
get rid of a useless wrapper, while we are at it
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch (commit 690a1f2002) added a
new call site for acpi_set_WOL() without checking that the function is
actually suitable to be called via
vortex_set_wol+0xcd/0xe0 [3c59x]
dev_ethtool+0xa5a/0xb70
dev_ioctl+0x2e0/0x4b0
T.961+0x49/0x50
sock_ioctl+0x47/0x290
do_vfs_ioctl+0x7f/0x340
sys_ioctl+0x80/0xa0
system_call_fastpath+0x16/0x1b
i.e. outside of code paths run when the device is not yet enabled or
already disabled. In particular, putting the device into D3hot is a
pretty bad idea when it was already brought up.
Furthermore, all prior callers of the function made sure they're
actually dealing with a PCI device, while the newly added one didn't.
In the same spirit, the .get_wol handler shouldn't indicate support
for WOL for non-PCI devices.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The HW by default has RX coalescing on. For iWARP connections, this
causes a 100ms delay in connection establishement due to the ingress
MPA Start message being stalled in HW. So explicitly turn RX
coalescing off when setting up iWARP connections.
This was causing very bad performance for NP64 gather operations using
Open MPI, due to the way it sets up connections on larger jobs.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Cc: <stable@kernel.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The PL310 on the ct-ca9x4 tile for the Versatile Express does not need
to add additional latency when accessing its cache RAMs. Unfortunately,
the boot monitor sets this up for an 8-cycle delay on reads and writes,
resulting in greatly reduced memory performance when the L2 cache is
enabled.
This patch sets the L2 RAM latencies to the correct value of 1 cycle
on the ct-ca9x4 tile before enabling the L2 cache.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Return -ENOMEM when erroring on kmalloc and fix memory leaks when returning on error.
Signed-off-by: Davidlohr Bueso <dave@gnu.org>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
f4347553b3 removed the edac polling
mechanism in favor of using a notifier chain for conveying MCE
information to edac. However, the module removal path didn't test
whether the driver had setup the polling function workqueue at all and
the rmmod process was hanging in the kernel at try_to_del_timer_sync()
in the cancel_delayed_work() path, trying to cancel an uninitialized
work struct.
Fix that by adding a balancing check to the workqueue removal path.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
While debugging bit_spin_lock() hang, it was tracked down to gcc-4.4
misoptimization of non-inlined constant_test_bit() due to non-volatile
addr when 'const volatile unsigned long *addr' cast to 'unsigned long *'
with subsequent unconditional jump to pause (and not to the test) leading
to hang.
Compiling with gcc-4.3 or disabling CONFIG_OPTIMIZE_INLINING yields inlined
constant_test_bit() and correct jump, thus working around the kernel bug.
Other arches than asm-x86 may implement this slightly differently;
2.6.29 mitigates the misoptimization by changing the function prototype
(commit c4295fbb60) but probably fixing the issue
itself is better.
Signed-off-by: Alexander Chumachenko <ledest@gmail.com>
Signed-off-by: Michael Shigorin <mike@osdn.org.ua>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
