Currently we cap request allocations at q->nr_requests, but we allow a
batching io context to allocate up to 32 more (default setting).  This
can flood the queue with request allocations, with only a few batching
processes.  The real fix would be to limit the number of batchers, but
as that isn't currently tracked, I suggest we just cap the maximum
number of allocated requests to eg 50% over the limit.

This was observed in real life, users typically see this as vmstat bo
numbers going off the wall with seconds of no queueing afterwards.
Behaviour this bursty is not beneficial.
Signed-off-by: NJens Axboe <axboe@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

insert a missing bio_put when writting the md superblock.

Without this we have a steady growth in the "bio" slab.
Signed-off-by: NNeil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Merge head 'upstream-20050628-1' of rsync://rsync.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev

This patch removes unnecessary ata_port_start/stop() calls from
ahci_port_start/stop().
Signed-off-by: NTejun Heo <htejun@gmail.com>

This patch lengthens the delay between DET setting and clearing for
COMMRESET from 400us to 1ms.  I couldn't find any requiremen regarding
the duration of COMMRESET in SATA I/II specs but AHCI-1.1 10.4.2
states that it should be at least 1ms.
Signed-off-by: NTejun Heo <htejun@gmail.com>

Contributions from Alan Cox and maximilian attems.

drivers/block/cfq-iosched.c: In function 'cfq_put_queue':
drivers/block/cfq-iosched.c:303: sorry, unimplemented: inlining failed in call to 'cfq_pending_requests': function body not available
drivers/block/cfq-iosched.c:1080: sorry, unimplemented: called from here
drivers/block/cfq-iosched.c: In function '__cfq_may_queue':
drivers/block/cfq-iosched.c:1955: warning: the address of 'cfq_cfqq_must_alloc_slice', will always evaluate as 'true'
make[1]: *** [drivers/block/cfq-iosched.o] Error 1
make: *** [drivers/block/cfq-iosched.o] Error 2

Cc: Jeff Garzik <jgarzik@pobox.com>
Cc: Jens Axboe <axboe@suse.de>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

"Allied Telesis RE1000 Plus" is not supported at all
by kernel 2.6.12.

Signed-off-by: komurojun-mbn@nifty.com

I fleshed out libata.tmpl a bit while I was taking notes.

The amd8111e driver directly assigns the DMA mask to the dma_mask
member of the struct pci_dev instead of using pci_set_dma_mask(). This
makes the call to pci_dma_supported() redundant as pci_set_dma_mask()
does this check.

I do not own this device so I only compile-tested this patch.
Signed-off-by: NTobias Klauser <tklauser@nuerscht.ch>

For boards that invert the SMC91x IRQ line (maybe an FPGA inverts it),
the set_irq_type() call can't assume IRQT_RISING.  These particular
boards currently use OMAP-specific calls to change the trigger type,
but the boards break when set_irq_type() stops being a NOP.
Signed-off-by: NDavid Brownell <dbrownell@users.sourceforge.net>

Don't auto-configure yenta sockets for PCMCIA devices if it is connected to
the root PCI bus on the x86 or x86_64 architectures.  Previously, this was
handled by the "ioport_resource"/"iomem_resource" check a few lines below,
but with the new ACPI-based resource handling this doesn't catch all cases
any longer.

pci-yenta-cardbus-fix.patch and this patch should solve the initialization
time trouble.  However, the ACPI-based PCI resource handling is badly
broken, IMHO:

- many resources of devices don't show up in the resource trees (
  /proc/iomem and /proc/ioports) any longer. This means that PCMCIA, but
  also possibly other subsystems (ISA, PnP, ...) do not know which resources
  it cannot use.

- verify_root_windows() should fail if there are no iomem _or_ ioport
  resources, not only if there are no iomem _and_ ioport resources.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Provide a "modalias" entry in sysfs for PCMCIA devices.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

The PCMCIA card services layer is never setting the i/o map attributes when
SS_CAP_STATIC_MAP is specified.  Net result, sockets' set_io_map() calls
always see requests with most flags clear, meaning 8 bit access.

For hardware that always autosizes, that won't matter; and all current
STATIC_MAP drivers ignore those attributes.  A new driver (for at91rm9200)
suffers badly from this, since this forces everything into 8 bit mode and
that breaks both (a) cards requiring 16 bit access, and (b) ide-cs; but of
course 8-bit cards work OK (as does accessing card attributes).

So this patch arranges to pass the attributes down, matching the behavior
for non-static mappings (using the first/only I/O window).
Signed-off-by: NDavid Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.de>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

register_chrdev() can return errors (negative) other then -EBUSY, so check
for any negative error code.
Signed-off-by: NRandy Dunlap <rdunlap@xenotime.net>
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Fix example hash generator in Documentation.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

- properly bail out in set_cis_map if call to socket driver's set_mem_map
  failed
- don't abort do_mem_probe cycle if one entry failed (!CONFIG_PCMCIA_PROBE)
- don't do iomem probing in chunks larger than 0x800000 (1 << 23) as
  yenta_socket and vrc4173_cardu.c fail to set_mem_map for windows equal to
  or larger than (1 << 24).
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

PCMCIA needs CRC32.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

The size of pointers may differ between (userspace) modpost and (kernelspace)
modules -- so fix mod_devicetable.h to reflect this possibility.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Remove the IRQ_INFO2_VALID flag in synclink_cs -- I overlooked it when
removing all other users in PCMCIA drivers for 2.6.11.  Thanks to Marcelo
Tosatti for noticing it.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

      randy_dunlap <rdunlap@xenotime.net>

Convert deprecated check_region() calls to request/release region.
Add return value check on one request_region().

I suspect that it may do an extra release_region(), which should
generate a warning message from the kernel.
Signed-off-by: Nrandy_dunlap <rdunlap@xenotime.net>
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Properly wait for the class refcount to reach zero.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Clean up ds.c
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Clean up cs.c
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Move the 16-bit PCMICA resource handling from pcmcia_core.o to pcmcia.o.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Rename some functions in drivers/pcmcia/ to show they belong to the PCMCIA
subsystem.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Remove unnecessary includes in ds.c and pcmcia_ioctl.c
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Merge struct pcmcia_bus_socket into struct pcmcia_socket.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

make pcmcia_bus_socket->state a bitfield, and rename it pcmcia_state to
prepare for struct pcmcia_bus_socket integration into struct pcmcia_socket.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

struct pcmcia_callback isn't needed for each socket, one is enough for all
sockets.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Move all PCMCIA_IOCTL-related code to a different file.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>

From: Richard Purdie <rpurdie@rpsys.net>

The pcmcia-move-pcmcia-ioctl-to-a-separate-file patch was corrupted in -mm2
causing this problem.
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Add a new config option to control the building of the PCMCIA IOCTL. Currently,
it is not yet made public, though the help text is there already.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowksi.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Automatically mark the parent PCI-PCI bridge windows as resources available
for PCMCIA usage.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Make adjust_io and adjust_memory independent of adjust_t to allow for IO
resources > x86's IO_SPACE_LIMIT.
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>