dev_queue_xmit() needs to dirty fields "state", "q", "bstats" and "qstats"
On x86_64 arch, they currently span three cache lines, involving more
cache line ping pongs than necessary, making longer holding of queue spinlock.
We can reduce this to one cache line, by grouping all read-mostly fields
at the beginning of structure. (Or should I say, all highly modified fields
at the end :) )
Before patch :
offsetof(struct Qdisc, state)=0x38
offsetof(struct Qdisc, q)=0x48
offsetof(struct Qdisc, bstats)=0x80
offsetof(struct Qdisc, qstats)=0x90
sizeof(struct Qdisc)=0xc8
After patch :
offsetof(struct Qdisc, state)=0x80
offsetof(struct Qdisc, q)=0x88
offsetof(struct Qdisc, bstats)=0xa0
offsetof(struct Qdisc, qstats)=0xac
sizeof(struct Qdisc)=0xc0
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To improve manageability, it would be good to be able to disambiguate routes
added by administrator from those added by DHCP client. The only necessary
kernel change is to add value to rtnetlink include file so iproute2 utility
can use it.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/pci/hotplug/fakephp.c: In function 'pci_rescan_bus':
drivers/pci/hotplug/fakephp.c:271: warning: passing argument 1 of 'pci_bus_assign_resources' discards qualifiers from pointer target type
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
- Rename pci_osc_control_set() to acpi_pci_osc_control_set() according
to the other API names in drivers/acpi/pci_root.c.
- Move _OSC related definitions to include/linux/acpi.h because _OSC
related API is implemented in drivers/acpi/pci_root.c now.
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Tested-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Move PCI _OSC management code from drivers/pci/pci-acpi.c to
drivers/acpi/pci_root.c. The benefits are
- We no longer need struct osc_data and its management code (contents
are moved to struct acpi_pci_root). This simplify the code, and we
no longer care about kmalloc() failure.
- We can make pci_acpi_osc_support() be a static function, which is
called only from drivers/acpi/pci_root.c.
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Tested-by: Andrew Patterson <andrew.patterson@hp.com>
Acked-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
If MSI-X interrupt mode is used by the PCI Express port driver, too
many vectors are allocated and it is not ensured that the right
vectors will be used for the right services. Namely, the PCI Express
specification states that both PCI Express native PME and PCI Express
hotplug will always use the same MSI or MSI-X message for signalling
interrupts, which implies that the same vector will be used by both
of them. Also, the VC service does not use interrupts at all.
Moreover, is not clear which of the vectors allocated by
pci_enable_msix() in the current code will be used for PME and
hotplug and which of them will be used for AER if all of these
services are configured.
For these reasons, rework the allocation of interrupts for PCI
Express ports so that if MSI-X are enabled, the right vectors will be
used for the right purposes.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reviewed-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Introduce new function pci_msix_table_size() returning the size of
the MSI-X table of given PCI device or 0 if the device doesn't
support MSI-X.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reviewed-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
The PCI Express port driver uses 'struct pcie_port_service_id' for
matching port service devices and drivers, but this structure
contains fields that duplicate information from the port device
itself (vendor, device, subvendor, subdevice) and fields that are not
used by any existing port service driver (class, class_mask,
drvier_data). Also, both existing port service drivers (AER and
PCIe HP) don't even use the vendor and device fields for device
matching. Therefore 'struct pcie_port_service_id' can be removed
altogether and the only useful members of it (port_type, service) can
be introduced directly into the port service device and port service
driver structures. That simplifies the code quite a bit and reduces
its size.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
The second argument of the ->probe() callback in
struct pcie_port_service_driver is unnecessary and never used.
Remove it.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
The PCI Express port driver should not attempt to register service
devices that require the ability to generate interrupts if generating
interrupts is not possible. Namely, if the port has no interrupt pin
configured and we cannot set up MSI or MSI-X for it, there is no way
it can generate interrupts and in such a case the port services that
rely on interrupts (PME, PCIe HP, AER) should not be enabled for it.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
PCI Express port driver extension, as defined by struct
pcie_port_device_ext in portdrv.h, is allocated and initialized, but
never used (it also is never freed). Extend it to hold the PCI Express
port type as well as the port interrupt mode, change its name and use it
to simplify the code in portdrv_core.c .
Additionally, remove the redundant interrupt_mode member of struct
pcie_device defined in include/linux/pcieport_if.h .
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Close-to-open cache consistency rules really only require us to flush out
writes on calls to close(), and require us to revalidate attributes on the
very last close of the file.
Currently we appear to be doing a lot of extra attribute revalidation
and cache flushes.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This fixes a regression against FreeBSD servers as reported by Tomas
Kasparek. Apparently when using RPC over a TCP socket, the FreeBSD servers
don't ever react to the client closing the socket, and so commit
e06799f958 (SUNRPC: Use shutdown() instead of
close() when disconnecting a TCP socket) causes the setup to hang forever
whenever the client attempts to close and then reconnect.
We break the deadlock by adding a 'linger2' style timeout to the socket,
after which, the client will abort the connection using a TCP 'RST'.
The default timeout is set to 15 seconds. A subsequent patch will put it
under user control by means of a systctl.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
When a port's link is down (except to driver restart) and the port is
configured for auto sensing, we try to sense port link type (Ethernet
or InfiniBand) in order to determine how to initialize the port. If
the port type needs to be changed, all mlx4 for the device interfaces
are unregistered and then registered again with the new port
types. Sensing is done with intervals of 3 seconds.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Clean up: Enable the use of const arguments in higher level svc_ APIs
by adding const to the arguments of the helper functions in svc_xprt.h
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Add /proc/fs/nfsd/pool_stats to export to userspace various
statistics about the operation of rpc server thread pools.
This patch is based on a forward-ported version of
knfsd-add-pool-thread-stats which has been shipping in the SGI
"Enhanced NFS" product since 2006 and which was previously
posted:
http://article.gmane.org/gmane.linux.nfs/10375
It has also been updated thus:
* moved EXPORT_SYMBOL() to near the function it exports
* made the new struct struct seq_operations const
* used SEQ_START_TOKEN instead of ((void *)1)
* merged fix from SGI PV 990526 "sunrpc: use dprintk instead of
printk in svc_pool_stats_*()" by Harshula Jayasuriya.
* merged fix from SGI PV 964001 "Crash reading pool_stats before
nfsds are started".
Signed-off-by: Greg Banks <gnb@sgi.com>
Signed-off-by: Harshula Jayasuriya <harshula@sgi.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Avoid overloading the CPU scheduler with enormous load averages
when handling high call-rate NFS loads. When the knfsd bottom half
is made aware of an incoming call by the socket layer, it tries to
choose an nfsd thread and wake it up. As long as there are idle
threads, one will be woken up.
If there are lot of nfsd threads (a sensible configuration when
the server is disk-bound or is running an HSM), there will be many
more nfsd threads than CPUs to run them. Under a high call-rate
low service-time workload, the result is that almost every nfsd is
runnable, but only a handful are actually able to run. This situation
causes two significant problems:
1. The CPU scheduler takes over 10% of each CPU, which is robbing
the nfsd threads of valuable CPU time.
2. At a high enough load, the nfsd threads starve userspace threads
of CPU time, to the point where daemons like portmap and rpc.mountd
do not schedule for tens of seconds at a time. Clients attempting
to mount an NFS filesystem timeout at the very first step (opening
a TCP connection to portmap) because portmap cannot wake up from
select() and call accept() in time.
Disclaimer: these effects were observed on a SLES9 kernel, modern
kernels' schedulers may behave more gracefully.
The solution is simple: keep in each svc_pool a counter of the number
of threads which have been woken but have not yet run, and do not wake
any more if that count reaches an arbitrary small threshold.
Testing was on a 4 CPU 4 NIC Altix using 4 IRIX clients, each with 16
synthetic client threads simulating an rsync (i.e. recursive directory
listing) workload reading from an i386 RH9 install image (161480
regular files in 10841 directories) on the server. That tree is small
enough to fill in the server's RAM so no disk traffic was involved.
This setup gives a sustained call rate in excess of 60000 calls/sec
before being CPU-bound on the server. The server was running 128 nfsds.
Profiling showed schedule() taking 6.7% of every CPU, and __wake_up()
taking 5.2%. This patch drops those contributions to 3.0% and 2.2%.
Load average was over 120 before the patch, and 20.9 after.
This patch is a forward-ported version of knfsd-avoid-nfsd-overload
which has been shipping in the SGI "Enhanced NFS" product since 2006.
It has been posted before:
http://article.gmane.org/gmane.linux.nfs/10374
Signed-off-by: Greg Banks <gnb@sgi.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
If a filesystem being written to via NFS returns a short write count
(as opposed to an error) to nfsd, nfsd treats that as a success for
the entire write, rather than the short count that actually succeeded.
For example, given a 8192 byte write, if the underlying filesystem
only writes 4096 bytes, nfsd will ack back to the nfs client that all
8192 bytes were written. The nfs client does have retry logic for
short writes, but this is never called as the client is told the
complete write succeeded.
There are probably other ways it could happen, but in my case it
happened with a fuse (filesystem in userspace) filesystem which can
rather easily have a partial write.
Here is a patch to properly return the short write count to the
client.
Signed-off-by: David Shaw <dshaw@jabberwocky.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
As part of reducing the scope of the client_mutex, and in order to
remove the need for mutexes from the callback code (so that callbacks
can be done as asynchronous rpc calls), move manipulations of the
file_hashtable under the recall_lock.
Update the relevant comments while we're here.
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Cc: Alexandros Batsakis <batsakis@netapp.com>
Reviewed-by: Benny Halevy <bhalevy@panasas.com>
Delegreturn is enough a special case for preprocess_stateid_op to
warrant just open-coding it in delegreturn.
There should be no change in behavior here; we're just reshuffling code.
Thanks to Yang Hongyang for catching a critical typo.
Reviewed-by: Yang Hongyang <yanghy@cn.fujitsu.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
The base versions handle constant folding now, none of these headers
are exported to userspace, so the __ prefixed versions are not
necessary.
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Impact: build fix
Fix:
arch/x86/kvm/x86.o: In function `dma_debug_add_bus':
(.text+0x0): multiple definition of `dma_debug_add_bus'
dma_debug_add_bus() should be a static inline function.
Cc: Joerg Roedel <joerg.roedel@amd.com>
LKML-Reference: <20090317120112.GP6159@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: better performance for if branch tracer
Use an array to count the hit and misses of a conditional instead
of using another conditional. This cuts down on saturation of branch
predictions and increases performance of modern pipelined architectures.
Signed-off-by: Witold Baryluk <baryluk@smp.if.uj.edu.pl>
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
This patch adds a new function called ring_buffer_set_clock that
allows a tracer to assign its own clock source to the buffer.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Impact: cleanup
Clean up #ifdefs and replace them with helper functions.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Impact: cleanup/sanitization
Start from a sane state while enabling dma and interrupt-remapping, by
clearing the previous recorded faults and disabling previously
enabled queued invalidation and interrupt-remapping.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Impact: new interfaces (not yet used)
Routines for disabling queued invalidation and interrupt remapping.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Impact: interface augmentation (not yet used)
Enable fault handling flow for intr-remapping aswell. Fault handling
code now shared by both dma-remapping and intr-remapping.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Since creating a device node is normally an operation requiring special
privilege, Igor Zhbanov points out that it is surprising (to say the
least) that a client can, for example, create a device node on a
filesystem exported with root_squash.
So, make sure CAP_MKNOD is among the capabilities dropped when an nfsd
thread handles a request from a non-root user.
Reported-by: Igor Zhbanov <izh1979@gmail.com>
Cc: stable@kernel.org
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
It depends on L3 support from 2.4 kernel (CONFIG_L3) that never got
merged into mainline. Since there's no way to use it on any of
supported machines (iPaq h3100 or h3600), better drop it for now.
It can be reimplemented later using ASoC infrastructure (there's
already a driver for uda1341 codec in mainline, so only CPU and machine
parts need to be written).
Signed-off-by: Dmitry Artamonow <mad_soft@inbox.ru>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Impact: allow architectures to monitor busses for dma mem leakage
This patch adds checking code to detect if a device has pending DMA
operations when it is about to be unbound from its device driver.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
This adds a function to dump the DMA mappings that the debugging code is
aware of -- either for a single device, or for _all_ devices.
This can be useful for debugging -- sticking a call to it in the DMA
page fault handler, for example, to see if the faulting address _should_
be mapped or not, and hence work out whether it's IOMMU bugs we're
seeing, or driver bugs.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
