Currently a queue can only dispatch up to 4 requests if there are other queues.
This isn't optimal, device can handle more requests, for example, AHCI can
handle 31 requests. I can understand the limit is for fairness, but we could
do a tweak: if the queue still has a lot of slice left, sounds we could
ignore the limit. Test shows this boost my workload (two thread randread of
a SSD) from 78m/s to 100m/s.
Thanks for suggestions from Corrado and Vivek for the patch.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Currently even if BLKTRACESETUP ioctl has failed user must call
BLKTRACETEARDOWN to be shure what all staff was cleaned, which
is contr-intuitive.
Let's setup ioctl make necessery cleanup by it self.
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

merge_bvec_fn() returns bvec->bv_len on success. So we have to check
against this value. But in case of fs_optimization merge we compare
with wrong value. This patch must be included in
 b428cd6da7e6559aca69aa2e3a526037d3f20403
But accidentally i've forgot to add this in the initial patch.
To make things straight let's replace all such checks.
In fact this makes code easy to understand.
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Counters for requests "in flight" and "in driver" are used asymmetrically
in cfq_may_dispatch, and have slightly different meaning.
We split the rq_in_flight counter (was sync_flight) to count both sync
and async requests, in order to use this one, which is more accurate in
some corner cases.
The rq_in_driver counter is coalesced, since individual sync/async counts
are not used any more.
Signed-off-by: Corrado Zoccolo <czoccolo@gmail.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: Fix problem with scatter gather elements in the scsi half of the driver
When support for more than 31 scatter gather elements was added to the block
half of the driver, the SCSI half of the driver was not addressed, and the bump
from 31 to 32 scatter gather elements in the command block itself (not chained)
actually broke the SCSI half of the driver, so that any transfer requiring 32
scatter gather elements wouldn't work.  This fix also increases the max transfer
size and size of the scatter gather table to the limit supported by the controller
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: eliminate unnecessary pointer use in cciss scsi code
An extra level of indirection was being used in some places
for no real reason.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: do not use void pointer for scsi hba data
and get rid of related unnecessary type casting
and delete some superfluous and misleading comments nearby.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: factor out scatter gather chain block mapping code
Rationale is I want to use this code from the scsi half of the
driver.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: fix scatter gather chain block dma direction kludge
The data direction for the chained block of scatter gather
elements should always be PCI_DMA_TODEVICE, but was mistakenly
set to the direction of the data transfer, then a kludge to
fix it was added, in which pci_dma_sync_single_for_device or
pci_dma_sync_single_for_cpu was called.  If the correct direction
is used in the first place, the kludge isn't needed.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: simplify scatter gather code.
Instead of allocating an array of pointers to a structure
containing an SGDescriptor structure, and two other elements
that aren't really used, just allocate SGDescriptor structs.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: factor out scatter gather chain block allocation and freeing
Rationale is that I want to use this code from the scsi half of the
driver.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: detect bad alignment of scsi commands at build time
Incidentally fix some nearby c++ style comments.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

cciss: clarify command list padding calculation
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

CFQ currently applies the same logic of detecting seeky queues and
grouping them together for rotational disks as well as SSDs.
For SSDs, the time to complete a request doesn't depend on the
request location, but only on the size.
This patch therefore changes the criterion to group queues by
request size in case of SSDs, in order to achieve better fairness.
Signed-off-by: Corrado Zoccolo <czoccolo@gmail.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Current seeky detection is based on average seek lenght.
This is suboptimal, since the average will not distinguish between:
* a process doing medium sized seeks
* a process doing some sequential requests interleaved with larger seeks
and even a medium seek can take lot of time, if the requested sector
happens to be behind the disk head in the rotation (50% probability).

Therefore, we change the seeky queue detection to work as follows:
* each request can be classified as sequential if it is very close to
  the current head position, i.e. it is likely in the disk cache (disks
  usually read more data than requested, and put it in cache for
  subsequent reads). Otherwise, the request is classified as seeky.
* an history window of the last 32 requests is kept, storing the
  classification result.
* A queue is marked as seeky if more than 1/8 of the last 32 requests
  were seeky.

This patch fixes a regression reported by Yanmin, on mmap 64k random
reads.
Reported-by: Yanmin Zhang <yanmin_zhang@linux.intel.com>
Signed-off-by: Corrado Zoccolo <czoccolo@gmail.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

On 64 bit builds when CONFIG_BLK_CGROUP=n (the default) this removes 8
bytes of padding from structure io_context and drops its size from 72 to
64 bytes, so needing one fewer cachelines and allowing more objects per
slab in it's kmem_cache.
Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>

----
patch against 2.6.33
compiled & test on x86_64 AMDX2
regards
Richard
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Except for SCSI no device drivers distinguish between physical and
hardware segment limits.  Consolidate the two into a single segment
limit.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

The block layer calling convention is blk_queue_<limit name>.
blk_queue_max_sectors predates this practice, leading to some confusion.
Rename the function to appropriately reflect that its intended use is to
set max_hw_sectors.

Also introduce a temporary wrapper for backwards compability.  This can
be removed after the merge window is closed.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Add a BLK_ prefix to block layer constants.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

blk_queue_max_hw_sectors is no longer called by any subsystem and can be
removed.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Clarify blk_queue_max_sectors and update documentation.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

There's no need to take css reference here, for the caller
has already called rcu_read_lock() to prevent cgroup from
being removed.
Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Conflicts:
	include/linux/blkdev.h
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

If radix_tree_preload is failed in ima_inode_alloc, we don't need
radix_tree_preload_end because kernel is alread preempt enabled
Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
Signed-off-by: Mimi Zohar <zohar@us.ibm.com>
Signed-off-by: James Morris <jmorris@namei.org>

* 'urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/kyle/parisc-2.6:
  parisc: Set PCI CLS early in boot.

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] Fix broken sn2 build

Set the PCI CLS early in the boot process to prevent
device failures. In pcibios_set_master use the new
pci_cache_line_size instead of a hard-coded value.
Signed-off-by: Carlos O'Donell <carlos@codesourcery.com>
Reviewed-by: Grant Grundler <grundler@google.com>
Signed-off-by: Kyle McMartin <kyle@redhat.com>

* 'for-linus' of git://git.monstr.eu/linux-2.6-microblaze:
  microblaze: Fix out_le32() macro
  microblaze: Fix cache loop function for cache range

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  Revert "block: improve queue_should_plug() by looking at IO depths"

Trailing semicolon causes compilation involving out_le32() to fail.
Signed-off-by: Steven J. Magnani <steve@digidescorp.com>
Signed-off-by: Michal Simek <monstr@monstr.eu>

I create wrong asm code but none test shows that this part of code is wrong.
I am not convinces that were good idea to create asm optimized macros
for caches. The reason is that there is not optimization with previous code
that's why make sense to add old code and do some benchmarking which
functions are faster.
Signed-off-by: Michal Simek <monstr@monstr.eu>

Now that the bio list management stuff is generic, convert pktcdvd to
use bio lists instead of its own private bio list implementation.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Peter Osterlund <petero2@telia.com>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  net: bug fix for vlan + gro issue
  tc35815: Remove a wrong netif_wake_queue() call which triggers BUG_ON
  cdc_ether: new PID for Ericsson C3607w to the whitelist (resubmit)
  IPv6: better document max_addresses parameter
  MAINTAINERS: update mv643xx_eth maintenance status
  e1000: Fix DMA mapping error handling on RX
  iwlwifi: sanity check before counting number of tfds can be free
  iwlwifi: error checking for number of tfds in queue
  iwlwifi: set HT flags after channel in rxon

Traffic (tcp) doesnot start on a vlan interface when gro is enabled.
Even the tcp handshake was not taking place.
This is because, the eth_type_trans call before the netif_receive_skb
in napi_gro_finish() resets the skb->dev to napi->dev from the previously
set vlan netdev interface. This causes the ip_route_input to drop the
incoming packet considering it as a packet coming from a martian source.

I could repro this on 2.6.32.7 (stable) and 2.6.33-rc7.
With this fix, the traffic starts and the test runs fine on both vlan
and non-vlan interfaces.

CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Patrick McHardy <kaber@trash.net>
Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6:
  ACPI: Be in TS_POLLING state during mwait based C-state entry
  ACPI: Fix regression where _PPC is not read at boot even when ignore_ppc=0
  acer-wmi: Respect current backlight level when loading

* 'drm-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm/vmwgfx: Fix queries if no dma buffer thrashing is occuring.
  drm/nv50: fix vram ptes on IGPs to point at stolen system memory
  drm/nv50: fix instmem binding on IGPs to point at stolen system memory
  drm/nv50: improve vram page table construction
  drm/nv50: more efficient clearing of gpu page table entries
  drm/nv50: make nv50_mem_vm_{bind,unbind} operate only on vram
  drm/nouveau: Fix up pre-nv17 analog load detection.

Revert the change made to arch/ia64/sn/kernel/setup.c by commit
204fba4a as it breaks the build.

Fixing the build the b94b0808 way
breaks xpc because genksyms then fails to generate an CRC for
per_cpu____sn_cnodeid_to_nasid because of limitations in the
generic genksyms code.
Signed-off-by: Hedi Berriche <hedi@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>