Alan D. Brunelle [Tue, 29 Apr 2008 12:44:19 +0000 (14:44 +0200)]
block: Skip I/O merges when disabled
The block I/O + elevator + I/O scheduler code spend a lot of time trying
to merge I/Os -- rightfully so under "normal" circumstances. However,
if one were to know that the incoming I/O stream was /very/ random in
nature, the cycles are wasted.
This patch adds a per-request_queue tunable that (when set) disables
merge attempts (beyond the simple one-hit cache check), thus freeing up
a non-trivial amount of CPU cycles.
Signed-off-by: Alan D. Brunelle <alan.brunelle@hp.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
This patch changes rq->cmd from the static array to a pointer to
support large commands.
We rarely handle large commands. So for optimization, a struct request
still has a static array for a command. rq_init sets rq->cmd pointer
to the static array.
block: no need to initialize rq->cmd with blk_get_request
blk_get_request initializes rq->cmd (rq_init does) so the users don't
need to do that.
The purpose of this patch is to remove sizeof(rq->cmd) and &rq->cmd,
as a preparation for large command support, which changes rq->cmd from
the static array to a pointer. sizeof(rq->cmd) will not make sense and
&rq->cmd won't work.
block: no need to initialize rq->cmd in prepare_flush_fn hook
The block layer initializes rq->cmd (queue_flush calls rq_init) so
prepare_flush_fn hooks don't need to do that.
The purpose of this patch is to remove sizeof(rq->cmd), as a
preparation for large command support, which changes rq->cmd from the
static array to a pointer. sizeof(rq->cmd) will not make sense.
Adrian Bunk [Tue, 29 Apr 2008 07:49:06 +0000 (09:49 +0200)]
block/blk-barrier.c:blk_ordered_cur_seq() mustn't be inline
This patch fixes the following build error with UML and gcc 4.3:
<-- snip -->
...
CC block/blk-barrier.o
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/blk-barrier.c: In function ‘blk_do_ordered’:
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/blk-barrier.c:57: sorry, unimplemented: inlining failed in call to ‘blk_ordered_cur_seq’: function body not available
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/blk-barrier.c:252: sorry, unimplemented: called from here
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/blk-barrier.c:57: sorry, unimplemented: inlining failed in call to ‘blk_ordered_cur_seq’: function body not available
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/blk-barrier.c:253: sorry, unimplemented: called from here
make[2]: *** [block/blk-barrier.o] Error 1
<-- snip -->
Signed-off-by: Adrian Bunk <bunk@kernel.org> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Adrian Bunk [Tue, 29 Apr 2008 07:49:05 +0000 (09:49 +0200)]
block/elevator.c:elv_rq_merge_ok() mustn't be inline
This patch fixes the following build error with UML and gcc 4.3:
<-- snip -->
...
CC block/elevator.o
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/elevator.c: In function ‘elv_merge’:
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/elevator.c:73: sorry, unimplemented: inlining failed in call to ‘elv_rq_merge_ok’: function body not available
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/elevator.c:103: sorry, unimplemented: called from here
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/elevator.c:73: sorry, unimplemented: inlining failed in call to ‘elv_rq_merge_ok’: function body not available
/home/bunk/linux/kernel-2.6/git/linux-2.6/block/elevator.c:495: sorry, unimplemented: called from here
make[2]: *** [block/elevator.o] Error 1
make[1]: *** [block] Error 2
<-- snip -->
Signed-off-by: Adrian Bunk <bunk@kernel.org> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Mark vget_cycles() as __always_inline, so gcc is never tempted to make
the vsyscall vread_tsc() dive into kernel text, with resulting SIGSEGV.
This was a self-inflicted wound: I've not seen that happen with unhacked
sources; but for debug reasons I'd changed my x86/Makefile to compile
no-unit-at-a-time, and that in conjunction with OPTIMIZE_INLINING=y
ended up with vget_cycles() in kernel text. Perhaps it can happen
in other ways: safer to use __always_inline.
Ian Campbell [Sun, 27 Apr 2008 11:21:11 +0000 (12:21 +0100)]
x86: add more boot protocol documentation
Signed-off-by: Ian Campbell <ijc@hellion.org.uk> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Acked-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
x86: fix warning in "x86: clean up vSMP detection"
The function detect_vsmp_box is a void function in the PCI case.
Change the !PCI stub to void too.
Signed-off-by: Alexander van Heukelum <heukelum@fastmail.fm> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Pavel Emelyanov [Tue, 29 Apr 2008 10:37:41 +0000 (03:37 -0700)]
MAINTAINERS: The socketcan-core list is subscribers-only.
When I posted a copy_to_user fixes, the list daemon refused to accept
the Cc: <this-list>, because I was not a subscriber.
I found, that other lists with such a feature are marked respectively
in the MAINTAINERS file.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Acked-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de> Signed-off-by: David S. Miller <davem@davemloft.net>
Philip Craig [Tue, 29 Apr 2008 10:35:10 +0000 (03:35 -0700)]
netfilter: nf_conntrack: padding breaks conntrack hash on ARM
commit 0794935e "[NETFILTER]: nf_conntrack: optimize hash_conntrack()"
results in ARM platforms hashing uninitialised padding. This padding
doesn't exist on other architectures.
Fix this by replacing NF_CT_TUPLE_U_BLANK() with memset() to ensure
everything is initialised. There were only 4 bytes that
NF_CT_TUPLE_U_BLANK() wasn't clearing anyway (or 12 bytes on ARM).
Signed-off-by: Philip Craig <philipc@snapgear.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Timo Teras [Tue, 29 Apr 2008 10:32:25 +0000 (03:32 -0700)]
ipv4: Update MTU to all related cache entries in ip_rt_frag_needed()
Add struct net_device parameter to ip_rt_frag_needed() and update MTU to
cache entries where ifindex is specified. This is similar to what is
already done in ip_rt_redirect().
Signed-off-by: Timo Teras <timo.teras@iki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
David L Stevens [Tue, 29 Apr 2008 10:23:22 +0000 (03:23 -0700)]
net: Add compat support for getsockopt (MCAST_MSFILTER)
This patch adds support for getsockopt for MCAST_MSFILTER for
both IPv4 and IPv6. It depends on the previous setsockopt patch,
and uses the same method.
Signed-off-by: David L Stevens <dlstevens@us.ibm.com> Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by: David S. Miller <davem@davemloft.net>
David L Stevens [Tue, 29 Apr 2008 10:23:00 +0000 (03:23 -0700)]
net: Several cleanups for the setsockopt compat support.
1) added missing "__user" for kgsr and kgf pointers
2) verify read for only GROUP_FILTER_SIZE(0). The group_filter
structure definition (via RFC) includes space for one source
in the source list array, but that source need not be present.
So, sizeof(group_filter) > GROUP_FILTER_SIZE(0). Fixed
the user read-check for minimum length to use the smaller size.
3) remove unneeded "&" for gf_slist addresses
Signed-off-by: David L Stevens <dlstevens@us.ibm.com> Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by: David S. Miller <davem@davemloft.net>
ipvs: fix oops in backup for fwmark conn templates
Fixes bug http://bugzilla.kernel.org/show_bug.cgi?id=10556
where conn templates with protocol=IPPROTO_IP can oops backup box.
Result from ip_vs_proto_get() should be checked because
protocol value can be invalid or unsupported in backup. But
for valid message we should not fail for templates which use
IPPROTO_IP. Also, add checks to validate message limits and
connection state. Show state NONE for templates using IPPROTO_IP.
Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off-by: David S. Miller <davem@davemloft.net>
Bodo Stroesser [Tue, 29 Apr 2008 10:18:13 +0000 (03:18 -0700)]
bridge: kernel panic when unloading bridge module
There is a race condition when unloading bridge and netfilter.
The problem happens if __fake_rtable is in use by a skb
coming in, while someone starts to unload bridge.ko.
br_netfilter_fini() is called at the beginning of unload
in br_deinit() while skbs still are being forwarded and
transferred to local ip stack. Thus there is a possibility
of the __fake_rtable pointer not being removed in a skb that
goes up to ip stack. This results in a kernel panic, as
ip_rcv() calls the input-function of __fake_rtable, which
is NULL.
Moving the call of br_netfilter_fini() to the end of
br_deinit() solves the problem.
Signed-off-by: Bodo Stroesser <bstroesser@fujitsu-siemens.com> Signed-off-by: Stephen Hemminger <stephen.hemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
When device is added to bridge its refcnt is incremented (in new_nbp()), but if
error occurs during further br_add_if() operations this counter is not
decremented back. Fix it by adding dev_put() call in the error path.
Signed-off-by: Volodymyr G Lukiianyk <volodymyrgl@gmail.com> Signed-off-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: David S. Miller <davem@davemloft.net>
netfilter: {nfnetlink,ip,ip6}_queue: fix skb_over_panic when enlarging packets
While reinjecting *bigger* modified versions of IPv6 packets using
libnetfilter_queue, things work fine on a 2.6.24 kernel (2.6.22 too)
but I get the following on recents kernels (2.6.25, trace below is
against today's net-2.6 git tree):
Looking at the code, I ended up in nfq_mangle() function (called by
nfqnl_recv_verdict()) which performs a call to skb_copy_expand() due to
the increased size of data passed to the function. AFAICT, it should ask
for 'diff' instead of 'diff - skb_tailroom(e->skb)'. Because the
resulting sk_buff has not enough space to support the skb_put(skb, diff)
call a few lines later, this results in the call to skb_over_panic().
The patch below asks for allocation of a copy with enough space for
mangled packet and the same amount of headroom as old sk_buff. While
looking at how the regression appeared (e2b58a67), I noticed the same
pattern in ipq_mangle_ipv6() and ipq_mangle_ipv4(). The patch corrects
those locations too.
Tested with bigger reinjected IPv6 packets (nfqnl_mangle() path), things
are ok (2.6.25 and today's net-2.6 git tree).
Signed-off-by: Arnaud Ebalard <arno@natisbad.org> Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Pavel Emelyanov [Tue, 29 Apr 2008 10:15:35 +0000 (03:15 -0700)]
netfilter: x_tables: fix net namespace leak when reading /proc/net/xxx_tables_names
The seq_open_net() call should be accompanied with seq_release_net() one.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org> Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
John Heffner [Tue, 29 Apr 2008 10:13:02 +0000 (03:13 -0700)]
tcp: Allow send-limited cwnd to grow up to max_burst when gso disabled
This changes the logic in tcp_is_cwnd_limited() so that cwnd may grow
up to tcp_max_burst() even when sk_can_gso() is false, or when
sysctl_tcp_tso_win_divisor != 0.
Signed-off-by: John Heffner <johnwheffner@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
The RapidIO system size will auto probe in RIO setup. The route table
and rionet_active in rionet.c are changed to be allocated dynamically
according to the size of the system.
Signed-off-by: Zhang Wei <wei.zhang@freescale.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
[RAPIDIO] Add OF-tree support to RapidIO controller driver
This initializes the RapidIO controller driver using addresses and
interrupt numbers obtained from the firmware device tree, rather than
using hardcoded constants.
Signed-off-by: Zhang Wei <wei.zhang@freescale.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
The original RapidIO driver suppose there is only one mpc85xx RIO controller
in system. So, some data structures are defined as mpc85xx_rio global, such
as 'regs_win', 'dbell_ring', 'msg_tx_ring'. Now, I changed them to mport's
private members. And you can define multi RIO OF-nodes in dts file for multi
RapidIO controller in one processor, such as PCI/PCI-Ex host controllers in
Freescale's silicon. And the mport operation function declaration should be
changed to know which RapidIO controller is target.
Signed-off-by: Zhang Wei <wei.zhang@freescale.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
block: add dma alignment and padding support to blk_rq_map_kern
This patch adds bio_copy_kern similar to
bio_copy_user. blk_rq_map_kern uses bio_copy_kern instead of
bio_map_kern if necessary.
bio_copy_kern uses temporary pages and the bi_end_io callback frees
these pages. bio_copy_kern saves the original kernel buffer at
bio->bi_private it doesn't use something like struct bio_map_data to
store the information about the caller.
As ps3disk is a ppc64-only driver, sector_t equals to unsigned long, and the
cast is not needed.
Reuse in another (possibly 32-bit) driver is protected by the safety net called
`compiler warning' (with the cast, it may silently truncate to 32-bit).
If sector_t ever changes, we will get a compiler warning as well (with the
cast, we won't).
This requires moving rq_init() from get_request() to blk_alloc_request().
The upside is that we can now require an rq_init() from any path that
wishes to hand the request to the block layer.
rq_init() will be exported for the code that uses struct request
without blk_get_request.
This is a preparation for large command support, which needs to
initialize struct request in a proper way (that is, just doing a
memset() will not work).
acpi_ev_disable_gpe() has an optimization where it doesn't disable
a GPE that it "doesn't have to". Unfortunately, it can get tricked
by AML that scribbles on register state behind its back. So when asked
to disable a GPE, simply do it -- a redundant register write
in the common case is a fair price to pay to be bomb-proof
for the rare cases.
http://bugzilla.kernel.org/show_bug.cgi?id=6217
Signed-off-by: Damián Viano <des@debian.org> Acked-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
PNPACPI: move _CRS/_PRS warnings closer to the action
Move warnings about _CRS and _PRS problems to the place where we
actually make the ACPI calls. Then we don't have to pass around
acpi_status values any more than necessary.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
PNP: add pnp_new_resource() to find a new unset pnp_resource
This encapsulates the code to locate a new pnp_resource of the
desired type. Currently this uses the pnp_resource_table, but
it will soon change to find a resource in a linked list.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Signed-off-by: Len Brown <len.brown@intel.com>
Save the ISAPNP config register index in the struct pnp_resource.
We need this because it is important to write ISAPNP configuration
back to the same registers we read it from. For example, if we
read valid regions from memory descriptors 0, 1, and 3, we'd
better write them back to the same registers, without compressing
them to descriptors 0, 1, and 2.
This was previously guaranteed by using the index into the
pnp_resource_table array as the ISAPNP config register index.
However, I am removing those fixed-size arrays, so we need to
save the ISAPNP register index elsewhere.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Signed-off-by: Len Brown <len.brown@intel.com>
In some places, we need to get the struct pnp_resource, not just
the struct resource, because ISAPNP needs to store the register
index in the pnp_resource.
I don't like pnp_get_pnp_resource() and hope that it is temporary,
but we need it for a little while.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Signed-off-by: Len Brown <len.brown@intel.com>
There are no remaining references to the PNP_MAX_* constants or
the pnp_resource_table structure outside of the PNP core. Make
them private to the PNP core.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Signed-off-by: Len Brown <len.brown@intel.com>
Remove some PNP_MAX_* uses. The pnp_resource_table isn't
dynamic yet, but with pnp_get_resource(), we can start moving
away from the table size constants.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Signed-off-by: Len Brown <len.brown@intel.com>
PNP: check for conflicts with all resources, not just earlier ones
This patch removes a use of "idx" in pnp_check_port() and similar
functions, in preparation for replacing idx with a pointer to the
resource itself.
I split this out because it changes the behavior slightly: we used
to check for conflicts only with earlier resources, e.g., we checked
resource 2 against resources 0 and 1 but not against 3, 4, etc. Now
we will check against all resources except 2.
Since resources are assigned in ascending order, the old behavior
was probably safe, but I don't like to depend on that ordering.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
This adds a pnp_get_resource() that works the same way as
platform_get_resource(). This will enable us to consolidate
many pnp_resource_table references in one place, which will
make it easier to make the table dynamic.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
PNP: factor pnp_init_resource_table() and pnp_clean_resource_table()
Move the common part of pnp_init_resource_table() and
pnp_clean_resource_table() into a new pnp_init_resource().
This reduces a little code duplication and will be
useful later to initialize an individual resource.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Signed-off-by: Len Brown <len.brown@intel.com>
These are no longer used in the PNP core either, so we can just remove
them completely.
It's possible that there are out-of-tree drivers that use these
interfaces. They should be changed to either (1) use PNP quirks
to work around broken hardware or firmware, or (2) use the sysfs
interfaces to control resource usage from userspace.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
PNP: remove pnp_resource_table from internal pnp_clean_resource_table interface
This changes pnp_clean_resource_table() to take a pnp_dev pointer
rather than a pnp_resource_table pointer. This reduces the visibility
of pnp_resource_table and removes an opportunity for error in the
caller.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
Add pnp_init_resources(struct pnp_dev *) to replace
pnp_init_resource_table(), which takes a pointer to the
pnp_resource_table itself. Passing only the pnp_dev * reduces
the possibility for error in the caller and removes the
pnp_resource_table implementation detail from the interface.
Even though pnp_init_resource_table() is exported, I did not
export pnp_init_resources() because it is used only by the PNP
core.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
Stop passing around struct pnp_resource_table pointers. In most cases,
the caller doesn't need to know how the resources are stored inside
the struct pnp_dev.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
PNP: remove pnp_resource_table from internal get/set interfaces
When we call protocol->get() and protocol->set() methods, we currently
supply pointers to both the pnp_dev and the pnp_resource_table even
though the pnp_resource_table should always be the one associated with
the pnp_dev.
This removes the pnp_resource_table arguments to make it clear that
these methods only operate on the specified pnp_dev.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
Add debug output to resource option registration functions (enabled
by CONFIG_PNP_DEBUG). This uses dev_printk, so I had to add pnp_dev
arguments at the same time.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
PNPACPI: hoist dma_flags() out of pnpacpi_parse_allocated_dmaresource()
Hoist dma_flags() out of pnpacpi_parse_allocated_dmaresource() into its
caller. This makes pnpacpi_parse_allocated_dmaresource() more similar
to pnpbios_parse_allocated_dmaresource().
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com> Acked-By: Rene Herman <rene.herman@gmail.com> Signed-off-by: Len Brown <len.brown@intel.com>
projects / linux-2.6-omap-h63xx.git / log