Any socket recv of less than this ammount will not be offloaded
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add an extra argument to sk_eat_skb, and make it move early copied
packets to the async_wait_queue instead of freeing them.
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Needed to be able to call tcp_cleanup_rbuf in tcp_input.c for I/OAT
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Adds an async_wait_queue and some additional fields to tcp_sock, and a
dma_cookie_t to sk_buff.
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Provides for pinning user space pages in memory, copying to iovecs,
and copying from sk_buffs including fragmented and chained sk_buffs.
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Attempts to allocate per-CPU DMA channels
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Provides an API for offloading memory copies to DMA devices
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, all userspace verbs operations that call into the kernel
are serialized by ib_uverbs_idr_mutex. This can be a scalability
issue for some workloads, especially for devices driven by the ipath
driver, which needs to call into the kernel even for datapath
operations.
Fix this by adding reference counts to the userspace objects, and then
converting ib_uverbs_idr_mutex into a spinlock that only protects the
idrs long enough to take a reference on the object being looked up.
Because remove operations may fail, we have to do a slightly funky
two-step deletion, which is described in the comments at the top of
uverbs_cmd.c.
This also still leaves ib_uverbs_idr_lock as a single lock that is
possibly subject to contention. However, the lock hold time will only
be a single idr operation, so multiple threads should still be able to
make progress, even if ib_uverbs_idr_lock is being ping-ponged.
Surprisingly, these changes even shrink the object code:
add/remove: 23/5 grow/shrink: 4/21 up/down: 633/-693 (-60)
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a call to initialize address handle attributes given a path record.
This is used by the CM, and would be useful for users of UD QPs.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a function to initialize address handle attributes from a work
completion. This functionality is duplicated by both verbs and the CM.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The P_Key is provided into a SIDR REQ in two places, once as a
parameter, and again in the path record. Remove the P_Key as a
parameter and always use the one given in the path record.
This change has no practical effect on ABI functionality.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add enum values for I/O Class values from rev. 10 and rev. 16a SRP
drafts. The values are used to detect targets that implement obsolete
revisions of SRP, so that the initiator can use the old format for
port identifier when connecting to them.
Signed-off-by: Ramachandra K <rkuchimanchi@silverstorm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Move ipath's struct port_info into <rdma/ib_smi.h>, so that it can be
used by mthca to implement client reregister support.
Remove the __attribute__((packed)) because all the members of the struct
are naturally aligned anyway.
Signed-off-by: Leonid Arsh <leonida@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add IB_EVENT_CLIENT_REREGISTER to enum so low-level drivers can
generate "client reregister" events.
Signed-off-by: Leonid Arsh <leonida@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add an LMC cache to struct ib_device, and add a function
ib_get_cached_lmc() to query the cache.
Signed-off-by: Jack Morgenstein <jackm@mellanox.co.il>
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Kernel connection management agent over InfiniBand that connects based
on IP addresses. The agent defines a generic RDMA connection
abstraction to support clients wanting to connect over different RDMA
devices.
The agent also handles RDMA device hotplug events on behalf of clients.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add an address translation service that maps IP addresses to
InfiniBand GID addresses using IPoIB.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Extend matching connection requests to listens in the InfiniBand CM to
include private data checks.
This allows applications to listen on the same service identifier,
with private data directing the request to the appropriate application.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Provide common handling for marshalling data between userspace clients
and kernel InfiniBand drivers.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Reflect the fact that the Cell Broadband Engine supports 64k
pages by adding the bit to the CPU features.
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Attached are two small patches for include/net/ieee80211.h to prepare
for later submission of code to implement a user-space daemon that
supplies 802.11 regulatory information.
The first change adds a bit indicating that 802.11h rules are to be
applied to a channel. As discussed earlier in this list, a single bit
is unlikely to be sufficient; however, at this time I have been unable
to find any regulations implementing differences between 802.11a and
802.11h other than DFS, radar detection and passive scanning. A single
bit is thus sufficient to convey to the driver that these rules should
be obeyed.
The second change adds comments to the freq and max_power fields of
struct ieee80211_channel to indicate the units that are used.
Signed-Off-By: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Currently the kernel blindly halts all the processors and calls the
ibm,suspend-me rtas call. If the firmware is not in the correct
state, we then re-start all the processors and return. It is much
smarter to first check the firmware state, and only if it is waiting,
call the ibm,suspend-me call.
Signed-off-by: Paul Mackerras <paulus@samba.org>
On non partitioned machines we currently set the HV bit in kernel space
only. It turns out we are supposed to maintain the HV bit in both user
and kernel space.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Allocate IOMMU tables local to the relevant node.
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
of_node_to_nid returns -1 if the associativity cannot be found. This
means pcibus_to_cpumask has to be careful not to pass a negative index into
node_to_cpumask.
Since pcibus_to_node could be used a lot, and of_node_to_nid is slow (it
walks a list doing strcmps), lets also cache the node in the
pci_controller struct.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Remove some stale POWER3/POWER4/970 on 32bit kernel support.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Forthcoming machines will extend the FPSCR to 64 bits. We already
had a 64-bit save area for the FPSCR, but we need to use a new form
of the mtfsf instruction. Fortunately this new form is decoded as
an ordinary mtfsf by existing 64-bit processors.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Instead of trying to make PPC64 MSI fit in a Intel-centric MSI layer, a
simple short-term solution is to hook the pci_{en/dis}able_msi() calls
and make a machdep call.
The rest of the MSI functions are superfluous for what is needed at this
time. Many of which can have machdep calls added as needed.
Ben and Michael Ellerman are looking into rewrite the MSI layer to be
more generic. However, in the meantime this works as a interim
solution.
Signed-off-by: Jake Moilanen <moilanen@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Some POWER5+ machines can do 64k hardware pages for normal memory but
not for cache-inhibited pages. This patch lets us use 64k hardware
pages for most user processes on such machines (assuming the kernel
has been configured with CONFIG_PPC_64K_PAGES=y). User processes
start out using 64k pages and get switched to 4k pages if they use any
non-cacheable mappings.
With this, we use 64k pages for the vmalloc region and 4k pages for
the imalloc region. If anything creates a non-cacheable mapping in
the vmalloc region, the vmalloc region will get switched to 4k pages.
I don't know of any driver other than the DRM that would do this,
though, and these machines don't have AGP.
When a region gets switched from 64k pages to 4k pages, we do not have
to clear out all the 64k HPTEs from the hash table immediately. We
use the _PAGE_COMBO bit in the Linux PTE to indicate whether the page
was hashed in as a 64k page or a set of 4k pages. If hash_page is
trying to insert a 4k page for a Linux PTE and it sees that it has
already been inserted as a 64k page, it first invalidates the 64k HPTE
before inserting the 4k HPTE. The hash invalidation routines also use
the _PAGE_COMBO bit, to determine whether to look for a 64k HPTE or a
set of 4k HPTEs to remove. With those two changes, we can tolerate a
mix of 4k and 64k HPTEs in the hash table, and they will all get
removed when the address space is torn down.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Let's not attempt the abolition of mtd->type until/unless it's properly
thought through. And certainly, let's not do it by halves.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Make the definitions of __FD_SET, __FD_CLR and __FD_ISSET independent
from asm/bitops.h and remove the macro magic that tests for __GLIBC__.
Use simple C inline functions instead of set_bit, clear_bit and test_bit.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Converted the locking mutex used for the ACPI hardware
to a spinlock. This change should eliminate all problems
caused by attempting to acquire a semaphore at interrupt
level, and it means that all ACPICA external interfaces
that directly access the ACPI hardware can be safely
called from interrupt level.
Fixed a regression introduced in 20060526 where the ACPI
device initialization could be prematurely aborted with
an AE_NOT_FOUND if a device did not have an optional
_INI method.
Fixed an IndexField issue where a write to the Data
Register should be limited in size to the AccessSize
(width) of the IndexField itself. (BZ 433, Fiodor Suietov)
Fixed problem reports (Valery Podrezov) integrated: - Allow
store of ThermalZone objects to Debug object.
http://bugzilla.kernel.org/show_bug.cgi?id=5369http://bugzilla.kernel.org/show_bug.cgi?id=5370
Fixed problem reports (Fiodor Suietov) integrated: -
acpi_get_table_header() doesn't handle multiple instances
correctly (BZ 364)
Removed four global mutexes that were obsolete and were
no longer being used.
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Restructured, flattened, and simplified the internal
interfaces for namespace object evaluation - resulting
in smaller code, less CPU stack use, and fewer
interfaces. (With assistance from Mikhail Kouzmich)
Fixed a problem with the CopyObject operator where the
first parameter was not typed correctly for the parser,
interpreter, compiler, and disassembler. Caused various
errors and unexpected behavior.
Fixed a problem where a ShiftLeft or ShiftRight of
more than 64 bits produced incorrect results with some
C compilers. Since the behavior of C compilers when
the shift value is larger than the datatype width is
apparently not well defined, the interpreter now detects
this condition and simply returns zero as expected in all
such cases. (BZ 395)
Fixed problem reports (Valery Podrezov) integrated: -
Update String-to-Integer conversion to match ACPI 3.0A spec
http://bugzilla.kernel.org/show_bug.cgi?id=5329
Allow interpreter to handle nested method declarations
http://bugzilla.kernel.org/show_bug.cgi?id=5361
Fixed problem reports (Fiodor Suietov) integrated: -
acpi_terminate() doesn't free debug memory allocation
list objects (BZ 355) - After Core Subsystem
shutdown, acpi_subsystem_status() returns AE_OK (BZ 356) -
acpi_os_unmap_memory() for RSDP can be invoked inconsistently
(BZ 357) - Resource Manager should return AE_TYPE for
non-device objects (BZ 358) - Incomplete cleanup branch
in AcpiNsEvaluateRelative (BZ 359) - Use acpi_os_free()
instead of ACPI_FREE in acpi_rs_set_srs_method_data (BZ 360)
- Incomplete cleanup branch in acpi_ps_parse_aml (BZ 361) -
Incomplete cleanup branch in acpi_ds_delete_walk_state (BZ 362)
- acpi_get_table_header returns AE_NO_ACPI_TABLES until DSDT
is loaded (BZ 365) - Status of the Global Initialization
Handler call not used (BZ 366) - Incorrect object parameter
to Global Initialization Handler (BZ 367)
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Replaced the acpi_os_queue_for_execution() with a new
interface named acpi_os_execute(). The major difference is
that the new interface does not have a Priority parameter,
this appeared to be useless and has been replaced by
a Type parameter. The Type tells the OS what type of
execution is being requested, such as global lock handler,
notify handler, GPE handler, etc. This allows the host
to queue and execute the request as appropriate for the
request type, possibly using different work queues and
different priorities for the various request types. This
enables fixes for multithreading deadlock problems such as
http://bugzilla.kernel.org/show_bug.cgi?id=5534
(Alexey Starikovskiy and Bob Moore)
Fixed a possible memory leak associated with the
support for the so-called "implicit return" ACPI
extension. Reported by FreeBSD (Fiodor Suietov)
http://bugzilla.kernel.org/show_bug.cgi?id=6514
Fixed a problem with the Load() operator where a table
load from an operation region could overwrite an internal
table buffer by up to 7 bytes and cause alignment faults
on IPF systems. (With assistance from Luming Yu)
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Removed a device initialization optimization introduced in
20051216 where the _STA method was not run unless an _INI
was also present for the same device. This optimization
could cause problems because it could allow _INI methods
to be run within a not-present device subtree (If a
not-present device had no _INI, _STA would not be run,
the not-present status would not be discovered, and the
children of the device would be incorrectly traversed.)
Implemented a new _STA optimization where namespace
subtrees that do not contain _INI are identified and
ignored during device initialization. Selectively running
_STA can significantly improve boot time on large machines
(with assistance from Len Brown.)
Implemented support for the device initialization case
where the returned _STA flags indicate a device not-present
but functioning. In this case, _INI is not run, but the
device children are examined for presence, as per the
ACPI specification.
Implemented an additional change to the IndexField support
in order to conform to MS behavior. The value written to
the Index Register is not simply a byte offset, it is a
byte offset in units of the access width of the parent
Index Field. (Fiodor Suietov)
Defined and deployed a new OSL interface,
acpi_os_validate_address(). This interface is called during
the creation of all AML operation regions, and allows
the host OS to exert control over what addresses it will
allow the AML code to access. Operation Regions whose
addresses are disallowed will cause a runtime exception
when they are actually accessed (will not affect or abort
table loading.)
Defined and deployed a new OSL interface,
acpi_os_validate_interface(). This interface allows the host OS
to match the various "optional" interface/behavior strings
for the _OSI predefined control method as appropriate
(with assistance from Bjorn Helgaas.)
Restructured and corrected various problems in the
exception handling code paths within DsCallControlMethod
and DsTerminateControlMethod in dsmethod (with assistance
from Takayoshi Kochi.)
Modified the Linux source converter to ignore quoted string
literals while converting identifiers from mixed to lower
case. This will correct problems with the disassembler
and other areas where such strings must not be modified.
The ACPI_FUNCTION_* macros no longer require quotes around
the function name. This allows the Linux source converter
to convert the names, now that the converter ignores
quoted strings.
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Implemented header file support for the following
additional ACPI tables: ASF!, BOOT, CPEP, DBGP, MCFG, SPCR,
SPMI, TCPA, and WDRT. With this support, all current and
known ACPI tables are now defined in the ACPICA headers and
are available for use by device drivers and other software.
Implemented support to allow tables that contain ACPI
names with invalid characters to be loaded. Previously,
this would cause the table load to fail, but since
there are several known cases of such tables on
existing machines, this change was made to enable
ACPI support for them. Also, this matches the
behavior of the Microsoft ACPI implementation.
https://bugzilla.novell.com/show_bug.cgi?id=147621
Fixed a couple regressions introduced during the memory
optimization in the 20060317 release. The namespace
node definition required additional reorganization and
an internal datatype that had been changed to 8-bit was
restored to 32-bit. (Valery Podrezov)
Fixed a problem where a null pointer passed to
acpi_ut_delete_generic_state() could be passed through
to acpi_os_release_object which is unexpected. Such
null pointers are now trapped and ignored, matching
the behavior of the previous implementation before the
deployment of acpi_os_release_object(). (Valery Podrezov,
Fiodor Suietov)
Fixed a memory mapping leak during the deletion of
a SystemMemory operation region where a cached memory
mapping was not deleted. This became a noticeable problem
for operation regions that are defined within frequently
used control methods. (Dana Meyers)
Reorganized the ACPI table header files into two main
files: one for the ACPI tables consumed by the ACPICA core,
and another for the miscellaneous ACPI tables that are
consumed by the drivers and other software. The various
FADT definitions were merged into one common section and
three different tables (ACPI 1.0, 1.0+, and 2.0)
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Implemented the use of a cache object for all internal
namespace nodes. Since there are about 1000 static nodes
in a typical system, this will decrease memory use for
cache implementations that minimize per-allocation overhead
(such as a slab allocator.)
Removed the reference count mechanism for internal
namespace nodes, since it was deemed unnecessary. This
reduces the size of each namespace node by about 5%-10%
on all platforms. Nodes are now 20 bytes for the 32-bit
case, and 32 bytes for the 64-bit case.
Optimized several internal data structures to reduce
object size on 64-bit platforms by packing data within
the 64-bit alignment. This includes the frequently used
ACPI_OPERAND_OBJECT, of which there can be ~1000 static
instances corresponding to the namespace objects.
Added two new strings for the predefined _OSI method:
"Windows 2001.1 SP1" and "Windows 2006".
Split the allocation tracking mechanism out to a separate
file, from utalloc.c to uttrack.c. This mechanism appears
to be only useful for application-level code. Kernels may
wish to not include uttrack.c in distributions.
Removed all remnants of the obsolete ACPI_REPORT_* macros
and the associated code. (These macros have been replaced
by the ACPI_ERROR and ACPI_WARNING macros.)
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
For a legacy ATA controller, libata registers two separate host sets.
There was no connection between the two hosts making it impossible to
traverse all ports related to the controller. This patch adds
host_set->next which points to the second host_set and makes
ata_pci_remove_one() remove all associated host_sets.
* On device removal, all ports hanging off the device are properly
detached. Prior to this patch, ports on the first host_set weren't
detached casuing oops on driver unloading.
* On device removal, both host_sets are properly freed
This will also be used by new power management code to suspend and
resume all ports of a controller. host_set/port representation will
be improved to handle legacy controllers better and this host_set
linking will go away with it.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
