Drawing on some inspiration from the CFS CPU scheduler design, overhaul
the pending cfq_queue concept list management. Currently CFQ uses a
doubly linked list per priority level for sorting and service uses.
Kill those lists and maintain an rbtree of cfq_queue's, sorted by when
to service them.
This unfortunately means that the ionice levels aren't as strong
anymore, will work on improving those later. We only scale the slice
time now, not the number of times we service. This means that latency
is better (for all priority levels), but that the distinction between
the highest and lower levels aren't as big.
The diffstat speaks for itself.
cfq-iosched.c | 363 +++++++++++++++++---------------------------------
1 file changed, 125 insertions(+), 238 deletions(-)
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
- Move the queue_new flag clear to when the queue is selected
- Only select the non-first queue in cfq_get_best_queue(), if there's
a substantial difference between the best and first.
- Get rid of ->busy_rr
- Only select a close cooperator, if the current queue is known to take
a while to "think".
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
- Implement logic for detecting cooperating processes, so we
choose the best available queue whenever possible.
- Improve residual slice time accounting.
- Remove dead code: we no longer see async requests coming in on
sync queues. That part was removed a long time ago. That means
that we can also remove the difference between cfq_cfqq_sync()
and cfq_cfqq_class_sync(), they are now indentical. And we can
kill the on_dispatch array, just make it a counter.
- Allow a process to go into the current list, if it hasn't been
serviced in this scheduler tick yet.
Possible future improvements including caching the cfqq lookup
in cfq_close_cooperator(), so we don't have to look it up twice.
cfq_get_best_queue() should just use that last decision instead
of doing it again.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
When testing the syslet async io approach, I discovered that CFQ
sometimes didn't perform as well as expected. cfq_should_preempt()
needs to better check for cooperating tasks, so fix that by allowing
preemption of an equal priority queue if the recently queued request
is as good a candidate for IO as the one we are currently waiting for.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Commit 404d5b185b changed the definition
of dev_dbg in the !DEBUG case from being a #define to being a static
inline. There was code in drivers/ps3/vuart.c to do exactly that,
which fails to compile now. This fixes it by removing the redefinition,
as the redefinition is now superfluous.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Remove option of making booke_wdt into a kernel module. This watchdog
cannot be disabled. No point being a kernel module.
Signed-off-by: Dave Jiang <djiang@mvista.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Call of_find_node_by_type with NULL instead of np
so the cpu node does not get put twice.
This was causing kref_put warnings.
Signed-off-by: John Rigby <jrigby@freescale.com>
Acked-by: Sylvain Munaut <tnt@246tNt.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Use DEFINE_SPINLOCK instead of initializing spinlocks to
SPIN_LOCK_UNLOCKED, since DEFINE_SPINLOCK is better for lockdep.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
For dma_alloc_*()
Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
arch/powerpc/sysdev/timer.c:51: error: variable `timer_sysclass' has
initializer but incomplete type
arch/powerpc/sysdev/timer.c:52: error: unknown field `resume' specified in initializer
<etc>
Signed-off-by: Srinivasa Ds <srinivasa@in.ibm.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Just another pass through arch/powerpc for old usages.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Recently, someone fixed a syntax error in the HTDMSOUND driver
introduced 4 years ago.
Unfortunately not by trying to compile this driver for his hardware but
by code inspection - which seems to be a strong indication that there
are no users left for this OSS sound driver.
This patch therefore removes it.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Dan Malek <dan@embeddedalley.com>
Acked-by: Marcelo Tosatti <marcelo@kvack.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch puts enable_kernel_spe into <asm-powerpc/system.h> along with
enable_kernel_altivec etc.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
dt_xlate_reg() looks up the 'reg' property in the specified node
to get the address and size to translate. Add dt_xlate_addr()
which is passed in the address and size to translate.
Signed-off-by: Mark A. Greer <mgreer@mvista.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
A usage of CONFIG_DEVICE_TREE got accidentally truncated; this
fix allows out-of-tree dts files to work.
Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Why create a platform specific board_info structure that is hacked
together, ugly, and dangerous, when we've got a perfectly fine common
board_info structure that is hacked-together, ugly and dangerous.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The arch/ppc/syslib/ppc_sys.c infrastructure does not work well for the
virtex ports. Move the ml300 and ml403 board ports over to use the new
virtex_devices infrastructure.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Acked-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Currently virtex support in mainline make use of the infrastructure in
arch/ppc/syslib/ppc_sys.c for registering common devices on virtex ppc405
platforms. The ppc_sys.c code is not well suited to the dynamic nature of
FPGA designs and makes adding new board ports more complex. This patch
adds a new listing of common devices which does not depend on the ppc_sys.c
infrastructure.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The header files for the ml403 and ml300 are virtually identical, merge
them into a single file.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Reverse dependency order for Xilinx Virtex parts. For these parts, It
makes more sense for boards/chips to specify which features they
provide instead of the features listing the parts they are implemented
in. I think it also makes adding new board ports simpler.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Acked-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Shuffle Kconfig order, making the platform drivers menu depend on the global
option instead of each driver being dependent on it.
Also fix dependency of PPC_PMAC on the G5 one.
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This reverts commit 9414715a7b,
at Olaf Hering's request:
> Paul, please discard this patch. The optional graphics card may have
> also device_type 'serial' if it is in VGA mode.
> I will send an updated patch later.
- s/Device Drivers/Controllers/
- clarify who needs pcilynx
- don't recommend Y for raw1394; M is typically used
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Fix the "attempting to setting" message in ohci1394.
Signed-off-by: Simon Arlott <simon@fire.lp0.eu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Remove the unneeded code that clears, sets and again clears the
rcvPhyPkt bit in the ohci1394 LinkControl register in ohci_initialize().
Signed-off-by: Bernhard Kauer <kauer@os.inf.tu-dresden.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
If posted write failed, an "Unhandled interrupt(s) 0x00000100" message
was logged by mistake.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
eth1394 did not work on buses consisting of S100B...S400B hardware
because it attempted to send GASP packets at S800.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This patch fixes some error handlings in eth1394:
- check return value of kmem_cache_create()
- cleanup resources if hpsb_register_protocol() fails
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> (whitespace)
This patch actually doesn't change anything because there was always 0
== NETDEV_TX_OK returned before.
TODO: Return NETDEV_TX_BUSY in error case and test in different error
conditions.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
RFC 2734 says: "IP-capable nodes may operate with an MTU size larger
than the default [1500 octets], but the means by which a larger MTU is
configured are beyond the scope of this document."
Allow users to set an MTU bigger than 1500.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
highlevel_host_reset no longer has any modular users.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Call only eth1394's own host reset handler from .tx_timeout, not the
reset hooks of all other IEEE 1394 drivers.
A minor drawback of this patch is that ether1394_host_reset by timeout
is not serialized against ether1394_host_reset by bus reset.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Move common code into an extra function. This implicitly adds a missing
node_info->fifo = CSR1212_INVALID_ADDR_SPACE; to .update.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Adjust white space and line wraps. Remove unnecessary parentheses and
braces, unused macros, and some of the more redundant comments.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
There is some common code between ether1394_open and ether1394_add_host
which can be moved to a separate helper function for a slightly smaller
eth1394 driver (-160 bytes on i386.)
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Until now, ieee1394 put an IP-over-1394 capability entry into each new
host's config ROM. As soon as the controller was initialized --- i.e.
right after modprobe ohci1394 --- this entry triggered a hotplug event
which typically caused auto-loading of eth1394.
This irritated or annoyed many users and distributors. Of course they
could blacklist eth1394, but then ieee1394 wrongly advertized IP-over-
1394 capability to the FireWire bus.
Therefore
- remove the offending kernel config option
IEEE1394_CONFIG_ROM_IP1394,
- let eth1394 add the ROM entry by itself, i.e. only after eth1394 was
loaded.
This fixes http://bugzilla.kernel.org/show_bug.cgi?id=7793 .
To emulate the behaviour of older kernels, simply add the following to
to /etc/modprobe.conf:
install ohci1394 /sbin/modprobe eth1394; \
/sbin/modprobe --ignore-install ohci1394
Note, autoloading of eth1394 when an _external_ IP-over-1394 capable
device is discovered is _not_ affected by this patch.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
