Kiyoshi Ueda [Tue, 11 Dec 2007 22:49:46 +0000 (17:49 -0500)]
blk_end_request: changing xsysace (take 4)
This patch converts xsysace to use blk_end_request interfaces.
Related 'uptodate' arguments are converted to 'error'.
xsysace is a little bit different from "normal" drivers.
xsysace driver has a state machine in it.
It calls end_that_request_first() and end_that_request_last()
from different states. (ACE_FSM_STATE_REQ_TRANSFER and
ACE_FSM_STATE_REQ_COMPLETE, respectively.)
However, those states are consecutive and without any interruption
inbetween.
So we can just follow the standard conversion rule (b) mentioned in
the patch subject "[PATCH 01/30] blk_end_request: add new request
completion interface".
Kiyoshi Ueda [Tue, 11 Dec 2007 22:53:24 +0000 (17:53 -0500)]
blk_end_request: cleanup of request completion (take 4)
This patch merges complete_request() into end_that_request_last()
for cleanup.
complete_request() was introduced by earlier part of this patch-set,
not to break the existing users of end_that_request_last().
Since all users are converted to blk_end_request interfaces and
end_that_request_last() is no longer exported, the code can be
merged to end_that_request_last().
Kiyoshi Ueda [Tue, 11 Dec 2007 22:53:03 +0000 (17:53 -0500)]
blk_end_request: cleanup 'uptodate' related code (take 4)
This patch converts 'uptodate' arguments of no longer exported
interfaces, end_that_request_first/last, to 'error', and removes
internal conversions for it in blk_end_request interfaces.
Also, this patch removes no longer needed end_io_error().
This patch removes the following functions:
o end_that_request_first()
o end_that_request_chunk()
and stops exporting the functions below:
o end_that_request_last()
This patch adds a variant of the interface, blk_end_bidi_request(),
which completes a bidi request.
Bidi request must be completed as a whole, both rq and rq->next_rq
at once. So the interface has 2 arguments for completion size.
As for ->end_io, only rq->end_io is called (rq->next_rq->end_io is not
called). So if special completion handling is needed, the handler
must be set to rq->end_io.
And the handler must take care of freeing next_rq too, since
the interface doesn't care of it if rq->end_io is not NULL.
Kiyoshi Ueda [Tue, 11 Dec 2007 22:51:23 +0000 (17:51 -0500)]
blk_end_request: changing ide-cd (take 4)
This patch converts ide-cd (cdrom_newpc_intr()) to use blk_end_request
interfaces. Related 'uptodate' arguments are converted to 'error'.
In PIO mode, ide-cd (cdrom_newpc_intr()) needs to defer
end_that_request_last() until the device clears DRQ_STAT and raises
an interrupt after end_that_request_first().
So blk_end_request() has to return without completing request
even if no leftover in the request.
ide-cd uses blk_end_request_callback() and a dummy callback function,
which just returns value '1', to tell blk_end_request_callback()
about that.
Kiyoshi Ueda [Tue, 11 Dec 2007 22:51:02 +0000 (17:51 -0500)]
blk_end_request: add callback feature (take 4)
This patch adds a variant of the interface, blk_end_request_callback(),
which has driver callback feature.
Drivers may need to do special works between end_that_request_first()
and end_that_request_last().
For such drivers, blk_end_request_callback() allows it to pass
a callback function which is called between end_that_request_first()
and end_that_request_last().
This interface is only for fallback of other blk_end_request interfaces.
Drivers should avoid their tricky behaviors and use other interfaces
as much as possible.
Currently, only one driver, ide-cd, needs this interface.
So this interface should/will be removed, after the driver removes
such tricky behaviors.
o ide-cd (cdrom_newpc_intr())
In PIO mode, cdrom_newpc_intr() needs to defer end_that_request_last()
until the device clears DRQ_STAT and raises an interrupt after
end_that_request_first().
So end_that_request_first() and end_that_request_last() are called
separately in cdrom_newpc_intr().
This means blk_end_request_callback() has to return without
completing request even if no leftover in the request.
To satisfy the requirement, callback function has return value
so that drivers can tell blk_end_request_callback() to return
without completing request.
Kiyoshi Ueda [Mon, 28 Jan 2008 09:34:20 +0000 (10:34 +0100)]
blk_end_request: changing ide normal caller (take 4)
This patch converts "normal" parts of ide to use blk_end_request
interfaces. Related 'uptodate' arguments are converted to 'error'.
The conversion of 'uptodate' to 'error' is done only for the internal
function, __ide_end_request().
ide_end_request() was not changed since it's exported and used
by many ide drivers.
With this patch, blkdev_dequeue_request() in __ide_end_request() is
moved to blk_end_request, since blk_end_request takes care of
dequeueing request like below:
if (!list_empty(&rq->queuelist))
blkdev_dequeue_request(rq);
In the case of ide,
o 'dequeue' variable of __ide_end_request() is 1 only when the request
is still linked to the queue (i.e. rq->queuelist is not empty)
o 'dequeue' variable of __ide_end_request() is 0 only when the request
has already been removed from the queue (i.e. rq->queuelist is empty)
So blk_end_request can handle it correctly although ide always run
thought the code above.
Kiyoshi Ueda [Tue, 11 Dec 2007 22:50:21 +0000 (17:50 -0500)]
blk_end_request: changing cpqarray (take 4)
This patch converts cpqarray to use blk_end_request interfaces.
Related 'ok' arguments are converted to 'error'.
cpqarray is a little bit different from "normal" drivers.
cpqarray directly calls bio_endio() and disk_stat_add()
when completing request. But those can be replaced with
__end_that_request_first().
After the replacement, request completion procedures of
those drivers become like the following:
o end_that_request_first()
o add_disk_randomness()
o end_that_request_last()
This can be converted to __blk_end_request() by following
the rule (b) mentioned in the patch subject
"[PATCH 01/30] blk_end_request: add new request completion interface".
Kiyoshi Ueda [Tue, 11 Dec 2007 22:50:03 +0000 (17:50 -0500)]
blk_end_request: changing cciss (take 4)
This patch converts cciss to use blk_end_request interfaces.
Related 'uptodate' arguments are converted to 'error'.
cciss is a little bit different from "normal" drivers.
cciss directly calls bio_endio() and disk_stat_add()
when completing request. But those can be replaced with
__end_that_request_first().
After the replacement, request completion procedures of
those drivers become like the following:
o end_that_request_first()
o add_disk_randomness()
o end_that_request_last()
This can be converted to blk_end_request() by following
the rule (a) mentioned in the patch subject
"[PATCH 01/30] blk_end_request: add new request completion interface".
Kiyoshi Ueda [Tue, 11 Dec 2007 22:46:10 +0000 (17:46 -0500)]
blk_end_request: changing sx8 (take 4)
This patch converts sx8 to use blk_end_request interfaces.
Related 'uptodate' and 'is_ok' arguments are converted to 'error'.
As a result, the interfaces of internal functions below are changed.
o carm_end_request_queued
o carm_end_rq
o carm_handle_array_info
o carm_handle_scan_chan
o carm_handle_generic
o carm_handle_rw
The 'is_ok' is set at only one place in carm_handle_resp() below:
int is_ok = (status == RMSG_OK);
And the value is propagated to all functions above, and no modification
in other places.
So the actual conversion of the 'is_ok' is done at only one place above.
Kiyoshi Ueda [Tue, 11 Dec 2007 22:42:27 +0000 (17:42 -0500)]
blk_end_request: changing arm (take 4)
This patch converts arm's OMAP mailbox driver to use
blk_end_request interfaces.
If the original code was converted literally, blk_end_request would
be called with '-EIO' because end_that_request_last() were called
with '0' (i.e. failure).
But I think these '0's are bugs in the original code because it's
unlikely that all requests are treated as failure.
(The bugs should have no effect unless these requests have an end_io
callback.)
So I changed them to pass '0' (i.e. success) to blk_end_request.
This patch converts core parts of block layer to use blk_end_request
interfaces. Related 'uptodate' arguments are converted to 'error'.
'dequeue' argument was originally introduced for end_dequeued_request(),
where no attempt should be made to dequeue the request as it's already
dequeued.
However, it's not necessary as it can be checked with
list_empty(&rq->queuelist).
(Dequeued request has empty list and queued request doesn't.)
And it has been done in blk_end_request interfaces.
As a result of this patch, end_queued_request() and
end_dequeued_request() become identical. A future patch will merge
and rename them and change users of those functions.
Kiyoshi Ueda [Tue, 11 Dec 2007 22:41:17 +0000 (17:41 -0500)]
blk_end_request: add/export functions to get request size (take 4)
This patch adds/exports functions to get the size of request in bytes.
They are useful because blk_end_request interfaces take bytes
as a completed I/O size instead of sectors.
Kiyoshi Ueda [Tue, 11 Dec 2007 22:40:30 +0000 (17:40 -0500)]
blk_end_request: add new request completion interface (take 4)
This patch adds 2 new interfaces for request completion:
o blk_end_request() : called without queue lock
o __blk_end_request() : called with queue lock held
blk_end_request takes 'error' as an argument instead of 'uptodate',
which current end_that_request_* take.
The meanings of values are below and the value is used when bio is
completed.
0 : success
< 0 : error
Some device drivers call some generic functions below between
end_that_request_{first/chunk} and end_that_request_last().
o add_disk_randomness()
o blk_queue_end_tag()
o blkdev_dequeue_request()
These are called in the blk_end_request interfaces as a part of
generic request completion.
So all device drivers become to call above functions.
To decide whether to call blkdev_dequeue_request(), blk_end_request
uses list_empty(&rq->queuelist) (blk_queued_rq() macro is added for it).
So drivers must re-initialize it using list_init() or so before calling
blk_end_request if drivers use it for its specific purpose.
(Currently, there is no driver which completes request without
re-initializing the queuelist after used it. So rq->queuelist
can be used for the purpose above.)
"Normal" drivers can be converted to use blk_end_request()
in a standard way shown below.
a) end_that_request_{chunk/first}
spin_lock_irqsave()
(add_disk_randomness(), blk_queue_end_tag(), blkdev_dequeue_request())
end_that_request_last()
spin_unlock_irqrestore()
=> blk_end_request()
Jens Axboe [Thu, 15 Nov 2007 08:13:11 +0000 (09:13 +0100)]
SG: Convert SCSI to use scatterlist helpers for sg chaining
Also change scsi_alloc_sgtable() to just return 0/failure, since it
maps to the command passed in. ->request_buffer is now no longer needed,
once drivers are adapted to use scsi_sglist() it can be killed.
If the users sets a new ->bi_bdev on the bio after __bio_clone() has
returned it, the "segment counts valid" flag still remains even though
it may be different with the new target. So don't calculate segment
counts in __bio_clone().
Pete Wyckoff [Tue, 1 Jan 2008 15:23:02 +0000 (10:23 -0500)]
block: allow queue dma_alignment of zero
Let queue_dma_alignment return 0 if it was specifically set to 0.
This permits devices with no particular alignment restrictions to
use arbitrary user space buffers without copying.
Christof Schmitt [Fri, 11 Jan 2008 09:09:43 +0000 (10:09 +0100)]
blktrace: Add blktrace ioctls to SCSI generic devices
Since the SCSI layer uses the request queues from the block layer, blktrace can
also be used to trace the requests to all SCSI devices (like SCSI tape drives),
not only disks. The only missing part is the ioctl interface to start and stop
tracing.
This patch adds the SETUP, START, STOP and TEARDOWN ioctls from blktrace to the
sg device files. With this change, blktrace can be used for SCSI devices like
for disks, e.g.: blktrace -d /dev/sg1 -o - | blkparse -i -
Magnus Damm [Wed, 23 Jan 2008 07:21:18 +0000 (16:21 +0900)]
sh: add spi header and r2d platform data V3
This patch adds the header file asm/spi.h and board specific code for the
r2d board. The header file contains a structure that should be used to
point out a single spi bus. The board specific code for r2d is updated with
such a structure for the new spi_sh_sci driver. The structure contains a
chip select callback plus information about the R9701 rtc chip which is
attached to the spi bus.
Magnus Damm [Fri, 25 Jan 2008 07:04:29 +0000 (16:04 +0900)]
sh: update r7780rp interrupt code
This patch updates the board specific irq code for r7780rp. The new code is
very similar to the other highlander implementations, with the exception that
the r7780rp handles pci interrupts using IRL. To simplify the pci code and
use the same interrupt numbers as r7780mp and r7785rp we hook in to the
cpu specific pci vectors.
The pci interrupts and the push switch all work well with and without this
patch. CF and AX88796 are not ok though and the source of the problem is
unknown at this point. The AX88796 does for not detect it's proper mac
address (IPL gets it right) and the kernel hangs on CF access. As a workaround
this patch removes the CF and the AX88796 from the platform datain case of
r7780rp.
Magnus Damm [Thu, 24 Jan 2008 09:45:45 +0000 (18:45 +0900)]
sh: remove consistent alloc stuff from the machine vector
Now with the voyagergx cruft gone and the dreamcast using declared
coherent memory for pci there are no users of the consistent alloc and
free functions pointers in the machine vector.
So this little patch simply removes these function pointers from the macvec.
Magnus Damm [Thu, 24 Jan 2008 09:35:10 +0000 (18:35 +0900)]
sh: declared coherent memory support V2
This patch adds declared coherent memory support to the sh architecture. All
functions are based on the x86 implementation. Header files are adjusted to
use the new functions instead of the former consistent_alloc() code.
This version includes the few changes what were included in the fix patch
together with modifications based on feedback from Paul.
Paul Mundt [Tue, 22 Jan 2008 00:57:27 +0000 (09:57 +0900)]
sh: Kill off -traditional for linker script.
Some of Sam's new work in the kbuild queue depend on ## concatenation
within the linker script, which doesn't work when -traditional is
enabled. -traditional is a legacy remnant anyways, and we no longer
require it for anything, so kill it off completely.
Adrian McMenamin [Mon, 21 Jan 2008 09:52:48 +0000 (18:52 +0900)]
cdrom: Add support for Sega Dreamcast GD-ROM.
This patch adds support for the GD-Rom drive, SEGA's proprietary
implementation of an IDE CD Rom for the SEGA Dreamcast. This driver
implements Sega's Packet Interface (SPI) - at least partially. It will
also read disks in SEGA's propreitary GD format.
Unlike previous drivers (which were never in mainline) this uses DMA and
not PIO to read disks. It is a new driver, not a refactoring of old
drivers.
Magnus Damm [Tue, 15 Jan 2008 03:47:53 +0000 (12:47 +0900)]
sh: remove voyagergx
This patch removes redundant irq handling code together with unused
consistent alloc code. R2D uart setup code is changed to use
sm501-regs.h and unused header files are removed.
Magnus Damm [Tue, 15 Jan 2008 03:44:44 +0000 (12:44 +0900)]
sh: r2d - add lcd planel timings to sm501 platform data
This patch adds timings to drive a Sharp LQ104V1DG21 lcd panel that can
be hooked up to R2D-1 or R2D-PLUS. The sm501fb driver should leave the
pins FPEN and VBIASEN alone, and this patch instructs the driver to do
so by not setting flags flags for these pins.
This patch works best together with the patch posted to the
linux-fbdev-devel list "sm501fb: control panel pin usage with platform
data flags", but this patch can be merged independently.
This patch removes interrupt priority tables from the intc code.
Optimal priority assignment varies with embedded application anyway,
so keeping the interrupt priority tables together with cpu-specific
code doesn't make sense.
The function intc_set_priority() should be used instead to set the
desired interrupt priority level.
Andrew Murray [Tue, 8 Jan 2008 04:27:23 +0000 (13:27 +0900)]
sh: Update SE7712 PCLK definition.
This patch provides a correct value for CONFIG_SH_PCLK_FREQ for the
SH7712 solution engine when used with the board's default factory
settings. This results in the board running at its maximum CPU clock
rate (200 MHz).
Paul Mundt [Wed, 26 Dec 2007 09:37:16 +0000 (18:37 +0900)]
sh: Fix get_user()/put_user() build error.
Fixes the build error caused by -Werror on gcc 3.x compilers:
arch/sh/kernel/signal_32.c: In function `sys_sigaction':
arch/sh/kernel/signal_32.c:66: warning: initialization discards qualifiers from pointer target type
arch/sh/kernel/signal_32.c:67: warning: initialization discards qualifiers from pointer target type
arch/sh/kernel/signal_32.c:69: warning: initialization discards qualifiers from pointer target type
arch/sh/kernel/signal_32.c:70: warning: initialization discards qualifiers from pointer target type
Paul Mundt [Tue, 18 Dec 2007 00:43:15 +0000 (09:43 +0900)]
sh: Always use CONFIG_HZ for HZ.
Currently the wdt forces HZ=1000 and sidesteps CONFIG_HZ completely. This
is a remnant from when HZ was hardcoded and before CONFIG_HZ was
introduced. Additionally, not all of the timers have this requirement
these days, so it's also an artificial limitation. Just kill it off and
use CONFIG_HZ directly.
Paul Mundt [Mon, 10 Dec 2007 06:50:28 +0000 (15:50 +0900)]
sh: Encode L1/L2 cache shape in auxvt.
This adds in the L1I/L1D/L2 cache shape support to their respective
entries in the ELF auxvt, based on the Alpha implementation. We use
this on the userspace libc side for calculating a tightly packed
SHMLBA amongst other things.
Thomas Betker [Fri, 30 Nov 2007 09:22:10 +0000 (18:22 +0900)]
sh: Fix up uImage target entry point.
This patch changes the uImage target so that it generates a wrapped
compressed vmlinux, rather than a wrapped zImage. The previous version
matched the ARM, this version matches the PPC. However I would question
how useful a self decompressing image is with a boot loader which does
decompression, so I think this is more useful. I also feel it matches
the descrition in the help text ("Compressed kernel image") better.
Stuart Menefy [Fri, 30 Nov 2007 09:16:23 +0000 (18:16 +0900)]
sh: get_user fixes and nommu consolidation.
When a get_user(to, from++) is called the pointer increment is performed
after its first usage, in the specific after the __add_ok invokation.
This causes a wrong get_user return value, putting a wrong character
in the destination variable. This patch solves the problem using a new
temporary pointer.
Additionally this reworks the use of the register banks, allowing for
consolidation between the MMU and nommu implementations.
Stuart Menefy [Fri, 30 Nov 2007 08:06:36 +0000 (17:06 +0900)]
sh: Preparation for uncached jumps through PMB.
Presently most of the 29-bit physical parts do P1/P2 segmentation
with a 1:1 cached/uncached mapping, jumping between the two to
control the caching behaviour. This provides the basic infrastructure
to maintain this behaviour on 32-bit physical parts that don't map
P1/P2 at all, using a shiny new linker section and corresponding
fixmap entry.
Stuart Menefy [Fri, 30 Nov 2007 07:12:36 +0000 (16:12 +0900)]
sh: GUSA atomic rollback support.
This implements kernel-level atomic rollback built on top of gUSA,
as an alternative non-IRQ based atomicity method. This is generally
a faster method for platforms that are lacking the LL/SC pairs that
SH-4A and later use, and is only supportable on legacy cores.
Paul Mundt [Wed, 28 Nov 2007 10:14:23 +0000 (19:14 +0900)]
rtc: rtc-sh: Split out the CPU defs to asm/cpu/.
With all of the different CPU types this was getting a but unwieldly.
Since sh64 is now integrated, we don't have to worry about multiple
architectures caring about the header definitions.
Split out the defs for each asm/cpu/ to make rtc-sh slightly less
visually offensive.
Paul Mundt [Wed, 28 Nov 2007 06:56:27 +0000 (15:56 +0900)]
sh: CCR1->CCR renaming for SH-2 parts.
Avoid namespace collision with a CCR1 definition. The general
SH code always expects CCR anyways, so there's no point in keeping
the CCR1 naming around.
Fixes up synclink collisions:
drivers/char/pcmcia/synclink_cs.c:283:1: warning: "CCR1" redefined
In file included from include/asm/cache.h:13,
from include/asm/processor_32.h:15,
from include/asm/processor.h:60,
from include/linux/prefetch.h:14,
from include/linux/list.h:8,
from include/linux/module.h:9,
from drivers/char/pcmcia/synclink_cs.c:38:
include/asm/cpu/cache.h:21:1: warning: this is the location of the previous definition
Stuart Menefy [Mon, 26 Nov 2007 12:29:09 +0000 (21:29 +0900)]
sh: Provide a 29/32-bit physical hint for bootloaders.
Shoves a magic word in to the empty_zero_page section for the
bootloader to work out whether to start the kernel in 29-bit
or 32-bit mode.
[ Renesas CPUs already take care of the initial PMB mappings entirely
in hardware and decide on 29-bit/32-bit physical depending on which
pin powered up the CPU, so this is mostly for ST parts. -- PFM ].
Paul Mundt [Mon, 26 Nov 2007 10:54:02 +0000 (19:54 +0900)]
sh: Add SH7263 CPU support.
This adds support for the SH7263 (SH-2A) CPU.
This particular CPU is a superset of SH7203, adding some additional
peripheral blocks and hooking up additional (reserved on SH7203)
vectors in the INTC block.
projects / linux.git / log