01 Nov, 2011
8 commits
-
The basic idea behind cross memory attach is to allow MPI programs doing
intra-node communication to do a single copy of the message rather than a
double copy of the message via shared memory.The following patch attempts to achieve this by allowing a destination
process, given an address and size from a source process, to copy memory
directly from the source process into its own address space via a system
call. There is also a symmetrical ability to copy from the current
process's address space into a destination process's address space.- Use of /proc/pid/mem has been considered, but there are issues with
using it:
- Does not allow for specifying iovecs for both src and dest, assuming
preadv or pwritev was implemented either the area read from or
written to would need to be contiguous.
- Currently mem_read allows only processes who are currently
ptrace'ing the target and are still able to ptrace the target to read
from the target. This check could possibly be moved to the open call,
but its not clear exactly what race this restriction is stopping
(reason appears to have been lost)
- Having to send the fd of /proc/self/mem via SCM_RIGHTS on unix
domain socket is a bit ugly from a userspace point of view,
especially when you may have hundreds if not (eventually) thousands
of processes that all need to do this with each other
- Doesn't allow for some future use of the interface we would like to
consider adding in the future (see below)
- Interestingly reading from /proc/pid/mem currently actually
involves two copies! (But this could be fixed pretty easily)As mentioned previously use of vmsplice instead was considered, but has
problems. Since you need the reader and writer working co-operatively if
the pipe is not drained then you block. Which requires some wrapping to
do non blocking on the send side or polling on the receive. In all to all
communication it requires ordering otherwise you can deadlock. And in the
example of many MPI tasks writing to one MPI task vmsplice serialises the
copying.There are some cases of MPI collectives where even a single copy interface
does not get us the performance gain we could. For example in an
MPI_Reduce rather than copy the data from the source we would like to
instead use it directly in a mathops (say the reduce is doing a sum) as
this would save us doing a copy. We don't need to keep a copy of the data
from the source. I haven't implemented this, but I think this interface
could in the future do all this through the use of the flags - eg could
specify the math operation and type and the kernel rather than just
copying the data would apply the specified operation between the source
and destination and store it in the destination.Although we don't have a "second user" of the interface (though I've had
some nibbles from people who may be interested in using it for intra
process messaging which is not MPI). This interface is something which
hardware vendors are already doing for their custom drivers to implement
fast local communication. And so in addition to this being useful for
OpenMPI it would mean the driver maintainers don't have to fix things up
when the mm changes.There was some discussion about how much faster a true zero copy would
go. Here's a link back to the email with some testing I did on that:http://marc.info/?l=linux-mm&m=130105930902915&w=2
There is a basic man page for the proposed interface here:
http://ozlabs.org/~cyeoh/cma/process_vm_readv.txt
This has been implemented for x86 and powerpc, other architecture should
mainly (I think) just need to add syscall numbers for the process_vm_readv
and process_vm_writev. There are 32 bit compatibility versions for
64-bit kernels.For arch maintainers there are some simple tests to be able to quickly
verify that the syscalls are working correctly here:http://ozlabs.org/~cyeoh/cma/cma-test-20110718.tgz
Signed-off-by: Chris Yeoh
Cc: Ingo Molnar
Cc: "H. Peter Anvin"
Cc: Thomas Gleixner
Cc: Arnd Bergmann
Cc: Paul Mackerras
Cc: Benjamin Herrenschmidt
Cc: David Howells
Cc: James Morris
Cc:
Cc:
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Fix the wrong use of schedule_hrtimeout_range_clock() in wq_sleep(),
although it is harmless for the syscall mq_timed* now. It was introduced
by 9ca7d8e ("mqueue: Convert message queue timeout to use hrtimers").Signed-off-by: Wanlong Gao
Cc: Carsten Emde
Cc: Thomas Gleixner
Cc: Manfred Spraul
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
The display of the "huge" tag was accidentally removed in 29ea2f698 ("mm:
use walk_page_range() instead of custom page table walking code").Reported-by: Stephen Hemminger
Tested-by: Stephen Hemminger
Reviewed-by: Stephen Wilson
Cc: KOSAKI Motohiro
Cc: Hugh Dickins
Acked-by: David Rientjes
Cc: Lee Schermerhorn
Cc: Alexey Dobriyan
Cc: Christoph Lameter
Cc:
Cc:
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
x86_64 allnoconfig:
In file included from arch/x86/kernel/pci-dma.c:3:
include/linux/dmar.h:248: warning: 'struct acpi_dmar_header' declared inside parameter list
include/linux/dmar.h:248: warning: its scope is only this definition or declaration, which is probably not what you wantCc: Suresh Siddha
Cc: Ingo Molnar
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
Commit 5fd75a7850b5 (dma-mapping: remove unnecessary sync_single_range_*
in dma_map_ops) unified not only the dma_map_ops but also the
corresponding debug_dma_sync_* calls. This led to spurious WARN()ings
like the following because the DMA debug code was no longer able to detect
the DMA buffer base address without the separate offset parameter:WARNING: at lib/dma-debug.c:911 check_sync+0xce/0x446()
firewire_ohci 0000:04:00.0: DMA-API: device driver tries to sync DMA memory it has not allocated [device address=0x00000000cedaa400] [size=1024 bytes]
Call Trace: ...
[] check_sync+0xce/0x446
[] debug_dma_sync_single_for_device+0x39/0x3b
[] ohci_queue_iso+0x4f3/0x77d [firewire_ohci]
...To fix this, unshare the sync_single_* and sync_single_range_*
implementations so that we are able to call the correct debug_dma_sync_*
functions.[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Clemens Ladisch
Cc: FUJITA Tomonori
Cc: Konrad Rzeszutek Wilk
Signed-off-by: Andrew Morton
Signed-off-by: Linus Torvalds -
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (27 commits)
vlan: allow nested vlan_do_receive()
ipv6: fix route lookup in addrconf_prefix_rcv()
bonding: eliminate bond_close race conditions
qlcnic: fix beacon and LED test.
qlcnic: Updated License file
qlcnic: updated reset sequence
qlcnic: reset loopback mode if promiscous mode setting fails.
qlcnic: skip IDC ack check in fw reset path.
i825xx: Fix incorrect dependency for BVME6000_NET
ipv6: fix route error binding peer in func icmp6_dst_alloc
ipv6: fix error propagation in ip6_ufo_append_data()
stmmac: update normal descriptor structure (v2)
stmmac: fix NULL pointer dereference in capabilities fixup (v2)
stmmac: fix a bug while checking the HW cap reg (v2)
be2net: Changing MAC Address of a VF was broken.
be2net: Refactored be_cmds.c file.
bnx2x: update driver version to 1.70.30-0
bnx2x: use FW 7.0.29.0
bnx2x: Enable changing speed when port type is PORT_DA
bnx2x: Fix 54618se LED behavior
... -
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
sparc64: Fix masking and shifting in VIS fpcmp emulation.
sparc32: Correct the return value of memcpy.
sparc32: Remove uses of %g7 in memcpy implementation.
sparc32: Remove non-kernel code from memcpy implementation. -
* 'for-linus' of git://neil.brown.name/md:
md/raid10: Fix bug when activating a hot-spare.
31 Oct, 2011
7 commits
-
Signed-off-by: David S. Miller
-
This is a fairly serious bug in RAID10.
When a RAID10 array is degraded and a hot-spare is activated, the
spare does not take up the empty slot, but rather replaces the first
working device.
This is likely to make the array non-functional. It would normally
be possible to recover the data, but that would need care and is not
guaranteed.This bug was introduced in commit
2bb77736ae5dca0a189829fbb7379d43364a9dac
which first appeared in 3.1.Cc: stable@kernel.org
Signed-off-by: NeilBrown -
* 'i2c-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging:
i2c: Functions for byte-swapped smbus_write/read_word_data
i2c-algo-pca: Return standard fault codes
i2c-algo-bit: Return standard fault codes
i2c-algo-bit: Be verbose on bus testing failure
i2c-algo-bit: Let user test buses without failing
i2c/scx200_acb: Fix section mismatch warning in scx200_pci_drv
i2c: I2C_ELEKTOR should depend on HAS_IOPORT -
* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu: (33 commits)
iommu/core: Remove global iommu_ops and register_iommu
iommu/msm: Use bus_set_iommu instead of register_iommu
iommu/omap: Use bus_set_iommu instead of register_iommu
iommu/vt-d: Use bus_set_iommu instead of register_iommu
iommu/amd: Use bus_set_iommu instead of register_iommu
iommu/core: Use bus->iommu_ops in the iommu-api
iommu/core: Convert iommu_found to iommu_present
iommu/core: Add bus_type parameter to iommu_domain_alloc
Driver core: Add iommu_ops to bus_type
iommu/core: Define iommu_ops and register_iommu only with CONFIG_IOMMU_API
iommu/amd: Fix wrong shift direction
iommu/omap: always provide iommu debug code
iommu/core: let drivers know if an iommu fault handler isn't installed
iommu/core: export iommu_set_fault_handler()
iommu/omap: Fix build error with !IOMMU_SUPPORT
iommu/omap: Migrate to the generic fault report mechanism
iommu/core: Add fault reporting mechanism
iommu/core: Use PAGE_SIZE instead of hard-coded value
iommu/core: use the existing IS_ALIGNED macro
iommu/msm: ->unmap() should return order of unmapped page
...Fixup trivial conflicts in drivers/iommu/Makefile: "move omap iommu to
dedicated iommu folder" vs "Rename the DMAR and INTR_REMAP config
options" just happened to touch lines next to each other. -
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp:
amd64_edac: Cleanup return type of amd64_determine_edac_cap()
amd64_edac: Add a fix for Erratum 505
EDAC, MCE, AMD: Simplify NB MCE decoder interface
EDAC, MCE, AMD: Drop local coreid reporting
EDAC, MCE, AMD: Print valid addr when reporting an error
EDAC, MCE, AMD: Print CPU number when reporting the error -
* 'kvm-updates/3.2' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm: (75 commits)
KVM: SVM: Keep intercepting task switching with NPT enabled
KVM: s390: implement sigp external call
KVM: s390: fix register setting
KVM: s390: fix return value of kvm_arch_init_vm
KVM: s390: check cpu_id prior to using it
KVM: emulate lapic tsc deadline timer for guest
x86: TSC deadline definitions
KVM: Fix simultaneous NMIs
KVM: x86 emulator: convert push %sreg/pop %sreg to direct decode
KVM: x86 emulator: switch lds/les/lss/lfs/lgs to direct decode
KVM: x86 emulator: streamline decode of segment registers
KVM: x86 emulator: simplify OpMem64 decode
KVM: x86 emulator: switch src decode to decode_operand()
KVM: x86 emulator: qualify OpReg inhibit_byte_regs hack
KVM: x86 emulator: switch OpImmUByte decode to decode_imm()
KVM: x86 emulator: free up some flag bits near src, dst
KVM: x86 emulator: switch src2 to generic decode_operand()
KVM: x86 emulator: expand decode flags to 64 bits
KVM: x86 emulator: split dst decode to a generic decode_operand()
KVM: x86 emulator: move memop, memopp into emulation context
... -
* 'fbdev-next' of git://github.com/schandinat/linux-2.6: (270 commits)
video: platinumfb: Add __devexit_p at necessary place
drivers/video: fsl-diu-fb: merge diu_pool into fsl_diu_data
drivers/video: fsl-diu-fb: merge diu_hw into fsl_diu_data
drivers/video: fsl-diu-fb: only DIU modes 0 and 1 are supported
drivers/video: fsl-diu-fb: remove unused panel operating mode support
drivers/video: fsl-diu-fb: use an enum for the AOI index
drivers/video: fsl-diu-fb: add several new video modes
drivers/video: fsl-diu-fb: remove broken screen blanking support
drivers/video: fsl-diu-fb: move some definitions out of the header file
drivers/video: fsl-diu-fb: fix some ioctls
video: da8xx-fb: Increased resolution configuration of revised LCDC IP
OMAPDSS: picodlp: add missing #include
fb: fix au1100fb bitrot.
mx3fb: fix NULL pointer dereference in screen blanking.
video: irq: Remove IRQF_DISABLED
smscufx: change edid data to u8 instead of char
OMAPDSS: DISPC: zorder support for DSS overlays
OMAPDSS: DISPC: VIDEO3 pipeline support
OMAPDSS/OMAP_VOUT: Fix incorrect OMAP3-alpha compatibility setting
video/omap: fix build dependencies
...Fix up conflicts in:
- drivers/staging/xgifb/XGI_main_26.c
Changes to XGIfb_pan_var()
- drivers/video/omap/{lcd_apollon.c,lcd_ldp.c,lcd_overo.c}
Removed (or in the case of apollon.c, merged into the generic
DSS panel in drivers/video/omap2/displays/panel-generic-dpi.c)
30 Oct, 2011
21 commits
-
Reimplemented at least 17 times discounting error mangling cases
where it could be used.Signed-off-by: Jonathan Cameron
Signed-off-by: Jean Delvare -
Adjust i2c-algo-pca to return fault codes compliant with
Documentation/i2c/fault-codes, rather than the undocumented and
vague -EREMOTEIO.Signed-off-by: Jean Delvare
Cc: Wolfram Sang -
Adjust i2c-algo-bit to return fault codes compliant with
Documentation/i2c/fault-codes, rather than the undocumented and
vague -EREMOTEIO.Signed-off-by: Jean Delvare
-
If bus testing fails due to the bus being seen as busy, it might be
helpful for developers to know which line is unexpectedly low.Signed-off-by: Jean Delvare
Reviewed-by: Alex Deucher -
Always failing to register I2C buses when the line testing fails is a
little harsh. While such a failure is definitely a bug in the driver
that exposes the affected I2C bus, things may still work fine if the
missing initialization steps are done later, before the I2C bus is
used. So it seems a better debugging tool to just report the test
failure by default. I introduce bit_test=2 if anyone really misses the
original behavior of bit_test=1.Signed-off-by: Jean Delvare
Reviewed-by: Alex Deucher -
WARNING: drivers/i2c/busses/built-in.o(.data+0x47c8): Section mismatch in reference from the variable scx200_pci_drv to the function .devinit.text:scx200_probe()
The variable scx200_pci_drv references
the function __devinit scx200_probe()
If the reference is valid then annotate the
variable with __init* or __refdata (see linux/init.h) or name the variable:
*driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_consoleSigned-off-by: Harvey Yang
Signed-off-by: Jean Delvare -
On m68k, I get:
drivers/i2c/busses/i2c-elektor.c: In function ‘pcf_isa_init’:
drivers/i2c/busses/i2c-elektor.c:153: error: implicit declaration of function ‘ioport_map’
drivers/i2c/busses/i2c-elektor.c:153: warning: assignment makes pointer from integer without a cast
drivers/i2c/busses/i2c-elektor.c: In function ‘elektor_probe’:
drivers/i2c/busses/i2c-elektor.c:287: error: implicit declaration of function ‘ioport_unmap’Since commit 82ed223c264def2b15ee4bec2e8c3048092ceb5f ("iomap: make IOPORT/PCI
mapping functions conditional"), ioport_map() is only available on platforms
that set HAS_IOPORT.Signed-off-by: Geert Uytterhoeven
Signed-off-by: Jean Delvare -
AMD processors apparently have a bug in the hardware task switching
support when NPT is enabled. If the task switch triggers a NPF, we can
get wrong EXITINTINFO along with that fault. On resume, spurious
exceptions may then be injected into the guest.We were able to reproduce this bug when our guest triggered #SS and the
handler were supposed to run over a separate task with not yet touched
stack pages.Work around the issue by continuing to emulate task switches even in
NPT mode.Signed-off-by: Jan Kiszka
Signed-off-by: Marcelo Tosatti -
Implement sigp external call, which might be required for guests that
issue an external call instead of an emergency signal for IPI.This fixes an issue with "KVM: unknown SIGP: 0x02" when booting
such an SMP guest.Signed-off-by: Christian Ehrhardt
Signed-off-by: Christian Borntraeger
Signed-off-by: Marcelo Tosatti -
KVM common code does vcpu_load prior to calling our arch ioctls and
vcpu_put after we're done here. Via the kvm_arch_vcpu_load/put
callbacks we do load the fpu and access register state into the
processor, which saves us moving the state on every SIE exit the
kernel handles. However this breaks register setting from userspace,
because of the following sequence:
1a. vcpu load stores userspace register content
1b. vcpu load loads guest register content
2. kvm_arch_vcpu_ioctl_set_fpu/sregs updates saved guest register content
3a. vcpu put stores the guest registers and overwrites the new content
3b. vcpu put loads the userspace register set againThis patch loads the new guest register state into the cpu, so that the correct
(new) set of guest registers will be stored in step 3a.Signed-off-by: Carsten Otte
Signed-off-by: Christian Borntraeger
Signed-off-by: Marcelo Tosatti -
This patch fixes the return value of kvm_arch_init_vm in case a memory
allocation goes wrong.Signed-off-by: Carsten Otte
Signed-off-by: Christian Borntraeger
Signed-off-by: Marcelo Tosatti -
We use the cpu id provided by userspace as array index here. Thus we
clearly need to check it first. Ooops.CC:
Signed-off-by: Carsten Otte
Signed-off-by: Christian Borntraeger
Signed-off-by: Marcelo Tosatti -
commit 2425717b27eb (net: allow vlan traffic to be received under bond)
broke ARP processing on vlan on top of bonding.+-------+
eth0 --| bond0 |---bond0.103
eth1 --| |
+-------+52870.115435: skb_gro_reset_offset
Reviewed-by: Jiri Pirko
Signed-off-by: David S. Miller -
The route lookup to find a previously auto-configured route for a prefixes used
to use rt6_lookup(), with the prefix from the RA used as an address. However,
that kind of lookup ignores routing tables, the prefix length and route flags,
so when there were other matching routes, even in different tables and/or with
a different prefix length, the wrong route would be manipulated.Now, a new function "addrconf_get_prefix_route()" is used for the route lookup,
which searches in RT6_TABLE_PREFIX and takes the prefix-length and route flags
into account.Signed-off-by: Andreas Hofmeister
Signed-off-by: David S. Miller -
This patch resolves two sets of race conditions.
Mitsuo Hayasaka reported the
first, as follows:The bond_close() calls cancel_delayed_work() to cancel delayed works.
It, however, cannot cancel works that were already queued in workqueue.
The bond_open() initializes work->data, and proccess_one_work() refers
get_work_cwq(work)->wq->flags. The get_work_cwq() returns NULL when
work->data has been initialized. Thus, a panic occurs.He included a patch that converted the cancel_delayed_work calls
in bond_close to flush_delayed_work_sync, which eliminated the above
problem.His patch is incorporated, at least in principle, into this
patch. In this patch, we use cancel_delayed_work_sync in place of
flush_delayed_work_sync, and also convert bond_uninit in addition to
bond_close.This conversion to _sync, however, opens new races between
bond_close and three periodically executing workqueue functions:
bond_mii_monitor, bond_alb_monitor and bond_activebackup_arp_mon.The race occurs because bond_close and bond_uninit are always
called with RTNL held, and these workqueue functions may acquire RTNL to
perform failover-related activities. If bond_close or bond_uninit is
waiting in cancel_delayed_work_sync, deadlock occurs.These deadlocks are resolved by having the workqueue functions
acquire RTNL conditionally. If the rtnl_trylock() fails, the functions
reschedule and return immediately. For the cases that are attempting to
perform link failover, a delay of 1 is used; for the other cases, the
normal interval is used (as those activities are not as time critical).Additionally, the bond_mii_monitor function now stores the delay
in a variable (mimicing the structure of activebackup_arp_mon).Lastly, all of the above renders the kill_timers sentinel moot,
and therefore it has been removed.Tested-by: Mitsuo Hayasaka
Signed-off-by: Jay Vosburgh
Signed-off-by: David S. Miller -
o Updated version number to 5.0.25
o Do not hold onto RESETTING_BIT for entire duration of LED/ beacon test.
Instead, just checking for RESETTING_BIT not set before sending config_led
command down to card.o Take rtnl_lock instead of RESETTING_BIT for beacon test while sending
config_led command down to make sure interface cannot be brought up/ down.o Allocate and free resources if interface is down before
sending the config_led command. This is to make sure config_led
command sending doesn't fail.o Clear QLCNIC_LED_ENABLE bit if beacon/ LED test fails to start.
Signed-off-by: Sucheta Chakraborty
Signed-off-by: Anirban Chakraborty
Signed-off-by: David S. Miller -
Updated qlcnic's license file.
Signed-off-by: Sritej Velaga
Signed-off-by: Anirban Chakraborty
Signed-off-by: David S. Miller -
Signed-off-by: Sony Chacko
Signed-off-by: Anirban Chakraborty
Signed-off-by: David S. Miller -
If promiscous mode setting fails, reset loopback mode setting in firmware.
Signed-off-by: Sucheta Chakraborty
Signed-off-by: Anirban Chakraborty
Signed-off-by: David S. Miller -
In fw reset path, we should consider any change in device state as an
ack from the other driver. When that happens, we don't have to wait for
an explicit ack.Signed-off-by: Sritej Velaga
Signed-off-by: Anirban Chakraborty
Signed-off-by: David S. Miller
29 Oct, 2011
4 commits
-
* git://github.com/rustyrussell/linux:
lguest: move process freezing before pending signals check
lguest: don't allow KVM-detection cpuid.
lguest: Allow running under paravirt-enabled KVM. -
* 'devicetree/merge' of git://git.secretlab.ca/git/linux-2.6:
ARM: mark empty gpio.h files empty
gpio: Fix ARM versatile-express build failure
of: include errno.h -
* 'spi/next' of git://git.secretlab.ca/git/linux-2.6:
drivercore: Add helper macro for platform_driver boilerplate
spi: irq: Remove IRQF_DISABLED
OMAP: SPI: Fix the trying to free nonexistent resource error
spi/spi-ep93xx: add module.h include
spi/tegra: fix compilation error in spi-tegra.c
spi: spi-dw: fix all sparse warnings
spi/spi-pl022: Call pl022_dma_remove(pl022) only if enable_dma is true
spi/spi-pl022: calculate_effective_freq() must set rate t allocate more sg than required.
spi/spi-pl022: Use GFP_ATOMIC for allocation from tasklet
spi/spi-pl022: Resolve formatting issues -
* 'gpio/next' of git://git.secretlab.ca/git/linux-2.6:
h8300: Move gpio.h to gpio-internal.h
gpio: pl061: add DT binding support
gpio: fix build error in include/asm-generic/gpio.h
gpiolib: Ensure struct gpio is always defined
irq: Add EXPORT_SYMBOL_GPL to function of irq generic-chip
gpio-ml-ioh: Use NUMA_NO_NODE not GFP_KERNEL
gpio-pch: Use NUMA_NO_NODE not GFP_KERNEL
gpio: langwell: ensure alternate function is cleared
gpio-pch: Support interrupt function
gpio-pch: Save register value in suspend()
gpio-pch: modify gpio_nums and mask
gpio-pch: support ML7223 IOH n-Bus
gpio-pch: add spinlock in suspend/resume processing
gpio-pch: Delete invalid "restore" code in suspend()
gpio-ml-ioh: Fix suspend/resume issue
gpio-ml-ioh: Support interrupt function
gpio-ml-ioh: Delete unnecessary code
gpio/mxc: add chained_irq_enter/exit() to mx3_gpio_irq_handler()
gpio/nomadik: use genirq core to track enablement
gpio/nomadik: disable clocks when unused