openwrt/staging/blogic.git
17 years agoi2c-core: Make some code static
Adrian Bunk [Sat, 13 Oct 2007 21:56:30 +0000 (23:56 +0200)]
i2c-core: Make some code static

After the i2c-isa removal some code can become static.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoi2c: New-style devices can support driver model wakeup flags
David Brownell [Sat, 13 Oct 2007 21:56:29 +0000 (23:56 +0200)]
i2c: New-style devices can support driver model wakeup flags

We need to be able to flag I2C devices, such as RTCs, which can issue wake
events (usually through IRQ lines).  This adds an i2c_board_info.flags bit,
and uses it to initialize the i2c device node.  (And shrinks a few lines
that were overly long.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
17 years agoi2c: Kill struct i2c_device_id
Jean Delvare [Sat, 13 Oct 2007 21:56:29 +0000 (23:56 +0200)]
i2c: Kill struct i2c_device_id

I2C devices do not have any form of ID as PCI or USB devices have.
No driver uses "MODULE_DEVICE_TABLE(i2c, ...)" because it doesn't
make sense. So we can get rid of struct i2c_device_id and the
associated support code.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: Greg KH <greg@kroah.com>
17 years agomissing includes in arch/powerpc/platforms/52xx/lite5200.c
Al Viro [Sat, 13 Oct 2007 18:40:08 +0000 (19:40 +0100)]
missing includes in arch/powerpc/platforms/52xx/lite5200.c

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofallout from elsa setup split
Al Viro [Sat, 13 Oct 2007 18:42:42 +0000 (19:42 +0100)]
fallout from elsa setup split

... and yes, caller wants it to return int.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoJFS: Bio cleanup: Replace missing return statements
Dave Kleikamp [Sat, 13 Oct 2007 17:58:00 +0000 (12:58 -0500)]
JFS: Bio cleanup: Replace missing return statements

commit e30408b2a99cb7b8bf529c7dc2328a19d71894cf ("JFS: fix bio-related
build breakage") removed some "return 0;" statements, rather than
changing them to null returns.

Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Sat, 13 Oct 2007 17:14:25 +0000 (10:14 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  [MIPS] SMP: Fix use of cpumasks.
  [MIPS] Revert "[MIPS] tlbex.c: Cleanup __init usage."
  [MIPS] CFE: Add missing parenthesis.

17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
Linus Torvalds [Sat, 13 Oct 2007 17:13:27 +0000 (10:13 -0700)]
Merge git://git./linux/kernel/git/bart/ide-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6: (27 commits)
  alim15x3: remove redundant m5229_revision check
  sc1200: fix ->dma_base equal zero handling
  cs5520: fix ->dma_base equal zero handling
  sgiioc4: add missing ->dma_base check
  cs5535: add missing ->dma_base check
  ide: remove CONFIG_IDEDMA_IVB config option
  ide: change master/slave IDENTIFY order
  ide: move ide_config_drive_speed() calls to upper layers (take 2)
  pdc202xx_new: check ide_config_drive_speed() return value
  cs5535: check ide_config_drive_speed() return value
  amd74xx/via82cxxx: check ide_config_drive_speed() return value
  au1xxx: fix au1xxx_set_pio_mode()
  icside: use ide_tune_dma()
  ide-pmac: fix PIO setup and enable autotune
  ide-pmac: use ide_tune_dma() (take 2)
  ide-pmac: remove pmac_ide_do_setfeature() (take 2)
  ide-pmac: remove nIEN clearing from pmac_ide_do_setfeature()
  ide-pmac: use __ide_wait_stat()
  ide-pmac: remove extra good status wait from pmac_ide_do_setfeature()
  ide: add __ide_wait_stat() helper
  ...

17 years agoMerge git://git.infradead.org/mtd-2.6
Linus Torvalds [Sat, 13 Oct 2007 17:12:15 +0000 (10:12 -0700)]
Merge git://git.infradead.org/mtd-2.6

* git://git.infradead.org/mtd-2.6: (91 commits)
  [MTD] [NAND] Blackfin on-chip NAND Flash Controller driver
  [MTD] [NOR] fix ctrl-alt-del can't reboot for intel flash bug
  [MTD] [NAND] Fix compiler warning in Alauda driver
  [JFFS2] Remove stray debugging printk
  [JFFS2] Handle dirents on the flash with embedded zero bytes in names.
  [JFFS2] Check for creation of dirents with embedded zero bytes in name.
  [JFFS2] Don't count all 'very dirty' blocks except in debug mode
  [JFFS2] Check whether garbage-collection actually obsoleted its victim.
  [JFFS2] Relax threshold for triggering GC due to dirty blocks.
  [MTD] [OneNAND] Fix typo related with recent commit
  [JFFS2] Trigger garbage collection when very_dirty_list size becomes excessive
  [MTD] [NAND] Avoid deadlock in erase callback; release chip lock first.
  [MTD] [NAND] Resume method for CAFÉ NAND controller
  [MTD] [NAND] Fix PCI ident table for CAFÉ NAND controller.
  [MTD] [NAND] s3c2410: fix arch moves
  [MTD] [OneNAND] fix numerous races
  [MTD] map driver for NOR flash on the Intel Vermilion Range chipset
  [JFFS2] Fix unpoint length
  [MTD] fix CFI point method for discontiguous maps
  [MTD] MAPS: Merge Lubbock and Mainstone drivers into common PXA2xx driver
  ...

17 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm
Linus Torvalds [Sat, 13 Oct 2007 17:02:11 +0000 (10:02 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/avi/kvm

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm: (106 commits)
  KVM: Replace enum by #define
  KVM: Skip pio instruction when it is emulated, not executed
  KVM: x86 emulator: popf
  KVM: x86 emulator: fix src, dst value initialization
  KVM: x86 emulator: jmp abs
  KVM: x86 emulator: lea
  KVM: X86 emulator: jump conditional short
  KVM: x86 emulator: imlpement jump conditional relative
  KVM: x86 emulator: sort opcodes into ascending order
  KVM: Improve emulation failure reporting
  KVM: x86 emulator: pushf
  KVM: x86 emulator: call near
  KVM: x86 emulator: push imm8
  KVM: VMX: Fix exit qualification width on i386
  KVM: Move main vcpu loop into subarch independent code
  KVM: VMX: Move vm entry failure handling to the exit handler
  KVM: MMU: Don't do GFP_NOWAIT allocations
  KVM: Rename kvm_arch_ops to kvm_x86_ops
  KVM: Simplify memory allocation
  KVM: Hoist SVM's get_cs_db_l_bits into core code.
  ...

17 years agoDelete filenames in comments.
Dave Jones [Sat, 13 Oct 2007 01:10:53 +0000 (21:10 -0400)]
Delete filenames in comments.

Since the x86 merge, lots of files that referenced their own filenames
are no longer correct.  Rather than keep them up to date, just delete
them, as they add no real value.

Additionally:
- fix up comment formatting in scx200_32.c
- Remove a credit from myself in setup_64.c from a time when we had no SCM
- remove longwinded history from tsc_32.c which can be figured out from
  git.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofix sparc32 breakage (result of vmlinux.lds.S bug)
Al Viro [Sat, 13 Oct 2007 07:40:24 +0000 (08:40 +0100)]
fix sparc32 breakage (result of vmlinux.lds.S bug)

In commit 4665079cbb2a3e17de82f2ab2940b9f97f37d65e ("[NETNS]: Move some
code into __init section when CONFIG_NET_NS=n") we got a new section -
.exit.text.refok (more of 'let's tell modpost that some bogus calls are
not bogus', a-la text.init.refok).

Unfortunately, the commit in question forgot to add it to TEXT_TEXT,
with rather amusing results.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agofallout from DECLARE_MAC_BUF patches
Al Viro [Sat, 13 Oct 2007 07:30:26 +0000 (08:30 +0100)]
fallout from DECLARE_MAC_BUF patches

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agov4l: copy_to_user() is not a good method name
Al Viro [Sat, 13 Oct 2007 07:25:24 +0000 (08:25 +0100)]
v4l: copy_to_user() is not a good method name

Breaks on any target that has copy_to_user() defined as a non-trivial
macro.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agominimal build fixes for uml (fallout from x86 merge)
Al Viro [Sat, 13 Oct 2007 07:16:04 +0000 (08:16 +0100)]
minimal build fixes for uml (fallout from x86 merge)

 a) include/asm-um/arch can't just point to include/asm-$(SUBARCH) now
 b) arch/{i386,x86_64}/crypto are merged now
 c) subarch-obj needed changes
 d) cpufeature_64.h should pull "cpufeature_32.h", not <asm/cpufeature_32.h>
    since it can be included from asm-um/cpufeature.h
 e) in case of uml-i386 we need CONFIG_X86_32 for make and gcc, but not
    for Kconfig
 f) sysctl.c shouldn't do vdso_enabled for uml-i386 (actually, that one
    should be registered from corresponding arch/*/kernel/*, with ifdef
    going away; that's a separate patch, though).

With that and with Stephen's patch ("[PATCH net-2.6] uml: hard_header fix")
we have uml allmodconfig building both on i386 and amd64.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agonet core: fix kernel-doc for new function parameters
Randy Dunlap [Sat, 13 Oct 2007 04:17:49 +0000 (21:17 -0700)]
net core: fix kernel-doc for new function parameters

Fix networking code kernel-doc for newly added parameters.

Warning(linux-2.6.23-git2//net/core/sock.c:879): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:570): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:594): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:617): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:641): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:667): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:722): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:959): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:1195): No description found for parameter 'dev'
Warning(linux-2.6.23-git2//net/core/dev.c:2105): No description found for parameter 'n'
Warning(linux-2.6.23-git2//net/core/dev.c:3272): No description found for parameter 'net'
Warning(linux-2.6.23-git2//net/core/dev.c:3445): No description found for parameter 'net'
Warning(linux-2.6.23-git2//include/linux/netdevice.h:1301): No description found for parameter 'cpu'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodocbook/mcabook: fixup x86 path/file names
Randy Dunlap [Sat, 13 Oct 2007 04:17:35 +0000 (21:17 -0700)]
docbook/mcabook: fixup x86 path/file names

Fix docbook templates for new x86 path/filenames:
docproc: linux-2.6.23-git2/include/asm-i386/mca_dma.h: No such file or directory

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodocbook/kernel-api: fixup x86 path/file names
Randy Dunlap [Sat, 13 Oct 2007 04:17:23 +0000 (21:17 -0700)]
docbook/kernel-api: fixup x86 path/file names

Fix docbook templates for new x86 path/filenames.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodocbook/kernel-hacking: fixup x86 path/file names
Randy Dunlap [Sat, 13 Oct 2007 04:17:12 +0000 (21:17 -0700)]
docbook/kernel-hacking: fixup x86 path/file names

Fix docbook templates for modified x86 path/filenames.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodocbook/deviceiobook: fixup x86 path/file names
Randy Dunlap [Sat, 13 Oct 2007 04:17:00 +0000 (21:17 -0700)]
docbook/deviceiobook: fixup x86 path/file names

Fix docbook templates for new x86 path/filenames:

docproc: linux-2.6.23-git2/include/asm-i386/io.h: No such file or directory

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh64-2.6
Linus Torvalds [Sat, 13 Oct 2007 16:50:26 +0000 (09:50 -0700)]
Merge git://git./linux/kernel/git/lethal/sh64-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh64-2.6:
  sh64: mach-cayman: Build fixes.
  sh64: Symbol export fixups.
  sh64: linker script tidying and alignment fixups.
  sh64: Set KBUILD_IMAGE to make the rpm target happy.
  sh64: Kill off obsolete linux/blk.h reference.
  sh64: cleanup struct irqaction initializers.
  sh64: Kill off dead gdb stub symbol.
  sh64: alphanumeric display only on Cayman.
  sh64: Add defconfigs for mach-sim and mach-harp.
  sh64: update cayman defconfig.
  sh64: Tidy up Kconfig dependencies.
  sh64: Move consistent DMA routines to arch/sh64/mm/.
  sh64: Some symbol exports and build fixes.
  sh64: mach-sim: Build fixes.
  sh64: mach-harp: Build fixes.
  sh64: Kill off duplicate frame pointer option.
  sh64: Kill off dead ROM-RAM and generic boards.
  sh64: Tidy up includes for Cayman board.
  sh64: Move *_p() I/O routine variants to io.h.

17 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6
Linus Torvalds [Sat, 13 Oct 2007 16:49:04 +0000 (09:49 -0700)]
Merge git://git./linux/kernel/git/lethal/sh-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6: (124 commits)
  sh: allow building for both r2d boards in same binary.
  sh: fix r2d board detection
  sh: Discard .exit.text/.exit.data at runtime.
  sh: Fix up some section alignments in linker script.
  sh: Fix SH-4 DMAC CHCR masking.
  sh: Rip out left-over nommu cond syscall cruft.
  sh: Make kgdb i-cache flushing less inept.
  sh: kgdb section mismatches and tidying.
  sh: cleanup struct irqaction initializers.
  sh: early_printk tidying.
  video: pvr2fb: Add TV (RGB) support to Dreamcast PVR driver.
  sh: Conditionalize gUSA support.
  sh: Follow gUSA preempt changes in __switch_to().
  sh: Tidy up gUSA preempt handling.
  sh: __copy_user() optimizations for small copies.
  sh: clkfwk: Support multi-level clock propagation.
  sh: Fix URAM start address on SH7785.
  sh: Use boot_cpu_data for CPU probe.
  sh: Support extended mode TLB on SH-X3.
  sh: Bump MAX_ACTIVE_REGIONS for SH7785.
  ...

17 years agossb bus needs <linux/io.h>
Geert Uytterhoeven [Sat, 13 Oct 2007 12:31:31 +0000 (14:31 +0200)]
ssb bus needs <linux/io.h>

ssb bus needs <linux/io.h>

  linux/drivers/ssb/main.c: In function 'ssb_ssb_read16':
  linux/drivers/ssb/main.c:518: error: implicit declaration of function 'readw'
  linux/drivers/ssb/main.c: In function 'ssb_ssb_read32':
  linux/drivers/ssb/main.c:526: error: implicit declaration of function 'readl'
  linux/drivers/ssb/main.c: In function 'ssb_ssb_write16':
  linux/drivers/ssb/main.c:534: error: implicit declaration of function 'writew'
  linux/drivers/ssb/main.c: In function 'ssb_ssb_write32':
  linux/drivers/ssb/main.c:542: error: implicit declaration of function 'writel'

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agob43 wireless needs <linux/io.h>
Geert Uytterhoeven [Sat, 13 Oct 2007 12:31:30 +0000 (14:31 +0200)]
b43 wireless needs <linux/io.h>

b43 wireless needs <linux/io.h>

  linux/drivers/net/wireless/b43/pio.h: In function 'b43_pio_write':
  linux/drivers/net/wireless/b43/pio.h:89: error: implicit declaration of function 'mmiowb'

  linux/drivers/net/wireless/b43/phy.c: In function 'b43_phy_write':
  linux/drivers/net/wireless/b43/phy.c:301: error: implicit declaration of function 'mmiowb'

  linuxdrivers/net/wireless/b43/sysfs.c: In function 'b43_attr_interfmode_store':
  linuxdrivers/net/wireless/b43/sysfs.c:147: error: implicit declaration of function 'mmiowb'

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68k: fix net drivers after recent get_stats updates
Geert Uytterhoeven [Sat, 13 Oct 2007 12:31:29 +0000 (14:31 +0200)]
m68k: fix net drivers after recent get_stats updates

m68k: fix net drivers after recent get_stats updates

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agodm: emc_endio returns void
Geert Uytterhoeven [Sat, 13 Oct 2007 12:31:28 +0000 (14:31 +0200)]
dm: emc_endio returns void

emc_endio returns void:
  linux/drivers/md/dm-emc.c: In function 'emc_endio':
  linux/drivers/md/dm-emc.c:58: warning: 'return' with a value, in function returning void

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68k: Export cachectl.h
Matthew Wilcox [Sat, 13 Oct 2007 12:31:27 +0000 (14:31 +0200)]
m68k: Export cachectl.h

libffi in GCC 4.2 needs cachectl.h to do its cache flushing.  But we
don't currently export it.  I believe this patch should do the trick.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68k: ignore restart_syscall
Geert Uytterhoeven [Sat, 13 Oct 2007 12:31:26 +0000 (14:31 +0200)]
m68k: ignore restart_syscall

m68k: ignore restart_syscall, which is not needed on m68k.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68k: Atari keyboard ACIA driver cleanup
Geert Uytterhoeven [Sat, 13 Oct 2007 12:31:25 +0000 (14:31 +0200)]
m68k: Atari keyboard ACIA driver cleanup

m68k: Atari keyboard ACIA driver cleanup:
  - removed dead key autorepeat code
  - removed hardcoded initial keymap

Signed-off-by: Michael Schmitz <schmitz@biophys.uni-duesseldorf.de>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agom68k: Atari input drivers cleanup
Geert Uytterhoeven [Sat, 13 Oct 2007 12:31:24 +0000 (14:31 +0200)]
m68k: Atari input drivers cleanup

m68k: Atari input drivers cleanup:
  - memleak on failed init/register of input devices fixed
  - correct keycodes table (Atari keycodes are almost, but not entirely, equal
    to Linux keycodes).

Signed-off-by: Michael Schmitz <schmitz@biophys.uni-duesseldorf.de>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
17 years agoalim15x3: remove redundant m5229_revision check
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:53 +0000 (17:47 +0200)]
alim15x3: remove redundant m5229_revision check

init_dma_ali15x3() guarantees that hwif->dma_base will never be set for
m5229_revision < 0x20 so remove redundant m5229_revision >= 0x20 check from
init_hwif_common_ali15x3().

While at it remove incorrect comment.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agosc1200: fix ->dma_base equal zero handling
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:53 +0000 (17:47 +0200)]
sc1200: fix ->dma_base equal zero handling

Set hwif->atapi_dma/{ultra,mwdma}_mask and drive->autodma after checking that
->dma_base exists.  If ->dma_base is not set (== PCI BAR4 cannot be reserved)
then DMA hooks shouldn't be initialized or bad things will happen.

OTOH hwif->set_{pio,dma}_mode hooks should be set even if hwif->dma_base == 0.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocs5520: fix ->dma_base equal zero handling
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:52 +0000 (17:47 +0200)]
cs5520: fix ->dma_base equal zero handling

Set hwif->ide_dma_{check,on} and hwif->autodma to 1 after checking that
->dma_base exists.  If ->dma_base is not set (== PCI BAR4 cannot be reserved)
then DMA hooks shouldn't be initialized or bad things will happen.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agosgiioc4: add missing ->dma_base check
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:52 +0000 (17:47 +0200)]
sgiioc4: add missing ->dma_base check

If ->dma_base is not set (== PCI BAR4 cannot be reserved) then DMA hooks
shouldn't be initialized or bad things will happen.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocs5535: add missing ->dma_base check
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:52 +0000 (17:47 +0200)]
cs5535: add missing ->dma_base check

If ->dma_base is not set (== PCI BAR4 cannot be reserved) then DMA hooks
shouldn't be initialized or bad things will happen.

Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide: remove CONFIG_IDEDMA_IVB config option
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:52 +0000 (17:47 +0200)]
ide: remove CONFIG_IDEDMA_IVB config option

Devices which don't set word 93 validation bit should be now handled by
ide-iops.c::ivb_list[] and for debugging purposes cable detection can be
completely overriden with "idex=ata66" parameter.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide: change master/slave IDENTIFY order
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:51 +0000 (17:47 +0200)]
ide: change master/slave IDENTIFY order

Need to probe slave device first to make it release PDIAG-
(this is required for correct device side cable detection).

Based on libata commit f31f0cc2f0b7527072d94d02da332d9bb8d7d94c.

Thanks to Craig for testing this patch.

Cc: Craig Block <chblock3@yahoo.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide: move ide_config_drive_speed() calls to upper layers (take 2)
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:51 +0000 (17:47 +0200)]
ide: move ide_config_drive_speed() calls to upper layers (take 2)

* Convert {ide_hwif_t,ide_pci_device_t}->host_flag to be u16.

* Add IDE_HFLAG_POST_SET_MODE host flag to indicate the need to program
  the host for the transfer mode after programming the device.  Set it
  in au1xxx-ide, amd74xx, cs5530, cs5535, pdc202xx_new, sc1200, pmac
  and via82cxxx host drivers.

* Add IDE_HFLAG_NO_SET_MODE host flag to indicate the need to completely
  skip programming of host/device for the transfer mode ("smart" hosts).
  Set it in it821x host driver and check it in ide_tune_dma().

* Add ide_set_pio_mode()/ide_set_dma_mode() helpers and convert all
  direct ->set_pio_mode/->speedproc users to use these helpers.

* Move ide_config_drive_speed() calls from ->set_pio_mode/->speedproc
  methods to callers.

* Rename ->speedproc method to ->set_dma_mode, make it void and update
  all implementations accordingly.

* Update ide_set_xfer_rate() comments.

* Unexport ide_config_drive_speed().

v2:
* Fix issues noticed by Sergei:
  - export ide_set_dma_mode() instead of moving ->set_pio_mode abuse wrt
    to setting DMA modes from sc1200_set_pio_mode() to do_special()
  - check IDE_HFLAG_NO_SET_MODE in ide_tune_dma()
  - check for (hwif->set_pio_mode) == NULL in ide_set_pio_mode()
  - check for (hwif->set_dma_mode) == NULL in ide_set_dma_mode()
  - return -1 from ide_set_{pio,dma}_mode() if ->set_{pio,dma}_mode == NULL
  - don't set ->set_{pio,dma}_mode on it821x in "smart" mode
  - fix build problem in pmac.c
  - minor fixes in au1xxx-ide.c/cs5530.c/siimage.c
  - improve patch description

Changes in behavior caused by this patch:
- HDIO_SET_PIO_MODE ioctl would now return -ENOSYS for attempts to change
  PIO mode if it821x controller is in "smart" mode
- removal of two debugging printk-s (from cs5530.c and sc1200.c)
- transfer modes 0x00-0x07 passed from user space may be programmed twice on
  the device (not really an issue since 0x00 is not supported correctly by
  any host driver ATM, 0x01 is not supported at all and 0x02-0x07 are invalid)

Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agopdc202xx_new: check ide_config_drive_speed() return value
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:51 +0000 (17:47 +0200)]
pdc202xx_new: check ide_config_drive_speed() return value

Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agocs5535: check ide_config_drive_speed() return value
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:51 +0000 (17:47 +0200)]
cs5535: check ide_config_drive_speed() return value

Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoamd74xx/via82cxxx: check ide_config_drive_speed() return value
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:51 +0000 (17:47 +0200)]
amd74xx/via82cxxx: check ide_config_drive_speed() return value

* Check ide_config_drive_speed() return value.

* While at also call ide_config_drive_speed() if the transfer mode is
  XFER_PIO_SLOW (this case happens iff the transfer mode has already been
  set on the device by ide-proc.c::set_xfer_rate()) and remove redundant
  setting of ->{init,current}_speed.

* Bump driver version.

Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoau1xxx: fix au1xxx_set_pio_mode()
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:51 +0000 (17:47 +0200)]
au1xxx: fix au1xxx_set_pio_mode()

Set transfer mode on the device before programming the host controller for
the new timings (matches what auide_tune_chipset() is doing wrt DMA modes).

Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoicside: use ide_tune_dma()
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:50 +0000 (17:47 +0200)]
icside: use ide_tune_dma()

* Add "good DMA drives" hack for icside to ide-dma.c::ide_find_dma_mode()
  (in the long-term it should be either removed or generalized for all hosts).

* Use ide_tune_dma() in icside.c::icside_dma_check().

  This results in the following changes in behavior:
  - pre-EIDE SWDMA modes are now also respected
  - drive->autodma is checked instead of hwif->autodma
    (doesn't really matter as icside sets both to "1")

* Make ide-dma.c::__ide_dma_good_drive() static and drop "__" prefix.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: fix PIO setup and enable autotune
Benjamin Herrenschmidt [Sat, 13 Oct 2007 15:47:50 +0000 (17:47 +0200)]
ide-pmac: fix PIO setup and enable autotune

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: use ide_tune_dma() (take 2)
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:50 +0000 (17:47 +0200)]
ide-pmac: use ide_tune_dma() (take 2)

* Add missing initialization of hwif->autodma and drive->autodma to
  pmac_ide_setup_dma().

* Use ide_tune_dma() in pmac_ide_dma_check().

While at it:

* Fix pmac_ide_dma_check() return value if DMA mode is not programmed
  (should be "-1" otherwise ide_set_dma() will try to enable DMA).

* Remove unnecessary drive->using_dma fiddling (->dma_off_quietly is always
  called before ide_set_dma() call and ide_set_dma() calls ->ide_dma_on if
  ->ide_dma_check returns "0").

v2:
* No reason to blacklist all ide_floppy devices and the old code was always
  enabling DMA anyway (without even programming controller/device if the
  device was ide_floppy).

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: remove pmac_ide_do_setfeature() (take 2)
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:50 +0000 (17:47 +0200)]
ide-pmac: remove pmac_ide_do_setfeature() (take 2)

Use ide_config_drive_speed() instead of pmac_ide_do_setfeature() and remove
the latter, also  ide-iops.c::__ide_wait_stat() could be static again.

Since for IDE PMAC host driver IDE_CONTROL_REG is always true, device's
->quirk_list is always zero and ->ide_dma_host_{on,off} are nops than
the only changes in behavior are:

* if PIO mode is set then ->dma_off_queitly is called to disable DMA

* if setting transfer mode fails ide_dump_status() is called to dump status

v2:
* IDE PMAC controllers allow separate PIO and DMA timings and PPC userland
  depends on this fact, and calls "hdparm -p" without calling "hdparm -d".

  Therefore to compensate for DMA being disabled by ide_config_drive_speed()
  for PIO modes:

  - add IDE_HFLAG_SET_PIO_MODE_KEEP_DMA flag and set it in PMAC host driver

  - add handling of the new flag to ide-io.c::do_special()

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: remove nIEN clearing from pmac_ide_do_setfeature()
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:49 +0000 (17:47 +0200)]
ide-pmac: remove nIEN clearing from pmac_ide_do_setfeature()

Upper layers are responsible for controlling nIEN so don't clear nIEN after
command execution in pmac_ide_do_setfeature().

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: use __ide_wait_stat()
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:49 +0000 (17:47 +0200)]
ide-pmac: use __ide_wait_stat()

* Use __ide_wait_stat() instead of wait_for_ready() in pmac_ide_do_setfeature().

While at it do following changes to match __ide_wait_stat() call in
ide_config_drive_speed():

* Wait WAIT_CMD time (20 sec) instead of 2 sec for device to clear BUSY_STAT.

* Check DRQ_STAT bit (shouldn't be set for good device status).

Also remove no longer needed wait_for_ready() from ide-iops.c.

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: remove extra good status wait from pmac_ide_do_setfeature()
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:49 +0000 (17:47 +0200)]
ide-pmac: remove extra good status wait from pmac_ide_do_setfeature()

Don't check for good device status before executing the command in
pmac_ide_do_setfeature() (ide_config_drive_speed() doesn't do this).

It is a job of upper layers to guarantee that the device is ready to
accept new command before we get here.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide: add __ide_wait_stat() helper
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:49 +0000 (17:47 +0200)]
ide: add __ide_wait_stat() helper

* Split off checking of the status register from ide_wait_stat() to
  __ide_wait_stat() helper.

* Use the new helper in ide_config_drive_speed().  The only change in the
  functionality is that the function now fails if after 20 sec (WAIT_CMD)
  device is still busy (BUSY_STAT bit is set) while previously instead of
  failing the function continued with checking for the correct device status
  (which would give the device additional 10 usec to clear BUSY_STAT bit).

* Remove stale comment for ide_config_drive_speed().

* Remove duplicate comment for ide_wait_stat() from <linux/ide.h>.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: remove pmac_ide_{m,u}dma_enable() (take 2)
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:49 +0000 (17:47 +0200)]
ide-pmac: remove pmac_ide_{m,u}dma_enable() (take 2)

* Fix pmac_ide_dma_check() to use pmac_ide_tune_chipset() and remove no longer
  necessary pmac_ide_{m,u}dma_enable().

* While at it remove some dead code from pmac_ide_dma_check() (leftovers from
  conversion to use ide_max_dma_mode()).

There should be no functionality changes caused by this patch.

v2:
* Fix compile by replacing "id" with "drive->id" in pmac_ide_dma_check()
  (Noticed by Ben).

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: remove control register messing from pmac_ide_dma_check()
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:49 +0000 (17:47 +0200)]
ide-pmac: remove control register messing from pmac_ide_dma_check()

pmac_ide_do_setfeature() contains matching nIEN setting/clearing so this
Device Control register messing in pmac_ide_dma_check() is totally unnecessary.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: fix set_timings_mdma()
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:48 +0000 (17:47 +0200)]
ide-pmac: fix set_timings_mdma()

* Move adjusting of cycle time for devices providing explicit DMA cycle time
  from pmac_ide_mdma_enable() to set_timings_mdma().

* Remove no longer needed drive_cycle_time argument.

* BUG() if unsupported speed argument value is passed (shouldn't happen).

* Matching access/recovery timings always exist so remove redundant check.

* Make set_timings_mdma() void.

* Update pmac_ide_tune_chipset()'s comment.

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: pmac_ide_tune_chipset() fixes
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:48 +0000 (17:47 +0200)]
ide-pmac: pmac_ide_tune_chipset() fixes

* Don't check check for pmif == NULL (it should never be NULL if we got here).

* Make a local copy of the timings and set the pmif->timings[] only after
  setting the transfer mode on the device (otherwise SELECT_DRIVE() call in
  pmac_ide_do_setfeature() will program new timings before the transfer mode
  is set on the device - this was pointed out by Sergei).  This change makes
  pmac_ide_tune_chipset() behavior match this of pmac_ide_{m,u}dma_enable().

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide-pmac: don't check kauai_lookup_timing() return value
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:48 +0000 (17:47 +0200)]
ide-pmac: don't check kauai_lookup_timing() return value

kauai_lookup_timing() should always return non-zero return value:

* BUG() in kauai_lookup_timing() if the timing info cannot be found.

* Remove code checking for zero return value from all callers.

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide: unexport ide_acpi_set_state
Adrian Bunk [Sat, 13 Oct 2007 15:47:48 +0000 (17:47 +0200)]
ide: unexport ide_acpi_set_state

This patch removes the unused EXPORT_SYMBOL_GPL(ide_acpi_set_state)

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoide_platform: set hwif->chipset
Bartlomiej Zolnierkiewicz [Sat, 13 Oct 2007 15:47:47 +0000 (17:47 +0200)]
ide_platform: set hwif->chipset

We need to set hwif->chipset or IDE PCI host drivers may try to claim
our ide_hwifs[] slot.

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
17 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
David Woodhouse [Sat, 13 Oct 2007 13:58:23 +0000 (14:58 +0100)]
Merge branch 'master' of git://git./linux/kernel/git/torvalds/linux-2.6

17 years agoMerge Linux 2.6.23
David Woodhouse [Sat, 13 Oct 2007 13:43:54 +0000 (14:43 +0100)]
Merge Linux 2.6.23

17 years ago[MTD] [NAND] Blackfin on-chip NAND Flash Controller driver
Bryan Wu [Tue, 2 Oct 2007 20:56:05 +0000 (13:56 -0700)]
[MTD] [NAND] Blackfin on-chip NAND Flash Controller driver

This is the driver for latest Blackfin on-chip nand flash controller

 - use nand_chip and mtd_info common nand driver interface
 - provide both PIO and dma operation
 - compiled with ezkit bf548 configuration
 - use hardware 1-bit ECC
 - tested with YAFFS2 and can mount YAFFS2 filesystem as rootfs

ChangeLog from try#1
 - use hweight32() instead of count_bits()
 - replace bf54x with bf5xx and BF54X with BF5XX
 - compare against plat->page_size in 2 cases when enable hardware ECC

ChangeLog from try#2
 - passed nand_test suites
 - use cpu_relax() instead of busy wait loop
 - some coding style issue pointed out by Andrew

Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [NOR] fix ctrl-alt-del can't reboot for intel flash bug
Kevin Hao [Tue, 2 Oct 2007 20:56:04 +0000 (13:56 -0700)]
[MTD] [NOR] fix ctrl-alt-del can't reboot for intel flash bug

When we press ctrl-alt-del,kernel_restart_prepare will invoke
cfi_intelext_reboot which will set flash to read array mode, but later
when device_shutdown is invoked which may put current work queue to
sleep and other process may be scheduled to running and programming
flash in not FL_READY mode again. So we can't boot up if this flash is
used for bootloader.

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [NAND] Fix compiler warning in Alauda driver
akpm@linux-foundation.org [Thu, 11 Oct 2007 07:02:06 +0000 (00:02 -0700)]
[MTD] [NAND] Fix compiler warning in Alauda driver

drivers/mtd/nand/alauda.c: In function 'alauda_bounce_read':
drivers/mtd/nand/alauda.c:412: warning: comparison of distinct pointer types lacks a cast

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Remove stray debugging printk
David Woodhouse [Sat, 13 Oct 2007 13:29:39 +0000 (14:29 +0100)]
[JFFS2] Remove stray debugging printk

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Handle dirents on the flash with embedded zero bytes in names.
David Woodhouse [Sat, 13 Oct 2007 10:35:58 +0000 (11:35 +0100)]
[JFFS2] Handle dirents on the flash with embedded zero bytes in names.

In three places: summary scan, normal scan, REF_PRISTINE GC.

Just truncate at the NUL, since that was the correct thing to do in the
only case where this (inexplicable) breakage has been seen.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Check for creation of dirents with embedded zero bytes in name.
David Woodhouse [Sat, 13 Oct 2007 10:33:50 +0000 (11:33 +0100)]
[JFFS2] Check for creation of dirents with embedded zero bytes in name.

I have no idea how this happened, but OLPC trac #4184 suggests that it
did. Catch it early.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Don't count all 'very dirty' blocks except in debug mode
David Woodhouse [Sat, 13 Oct 2007 10:32:16 +0000 (11:32 +0100)]
[JFFS2] Don't count all 'very dirty' blocks except in debug mode

... where we'll actually print the count in a debug message.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Check whether garbage-collection actually obsoleted its victim.
David Woodhouse [Sat, 13 Oct 2007 10:31:23 +0000 (11:31 +0100)]
[JFFS2] Check whether garbage-collection actually obsoleted its victim.

In OLPC trac #4184 we found a case where a corrupted node didn't
actually get obsoleted when we tried to garbage-collect it. So we wrote
out many million copies of it, in repeated attempts to obsolete it,
until the flash became full. Don't Do That.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[JFFS2] Relax threshold for triggering GC due to dirty blocks.
David Woodhouse [Sat, 13 Oct 2007 10:29:07 +0000 (11:29 +0100)]
[JFFS2] Relax threshold for triggering GC due to dirty blocks.

Instead of matching resv_blocks_gcmerge, which is only about 3, instead
match resv_blocks_gctrigger, which includes a proportion of the total
device size.

These ought to become tunable from userspace, at some point.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years ago[MTD] [OneNAND] Fix typo related with recent commit
Kyungmin Park [Wed, 10 Oct 2007 04:48:14 +0000 (13:48 +0900)]
[MTD] [OneNAND] Fix typo related with recent commit

Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
17 years agoKVM: Replace enum by #define
Avi Kivity [Wed, 10 Oct 2007 12:03:16 +0000 (14:03 +0200)]
KVM: Replace enum by #define

Easier for existence test (#ifdef) in userspace.

Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Skip pio instruction when it is emulated, not executed
Avi Kivity [Sat, 15 Sep 2007 14:34:36 +0000 (17:34 +0300)]
KVM: Skip pio instruction when it is emulated, not executed

If we defer updating rip until pio instructions are executed, we have a
problem with reset:  a pio reset updates rip, and when the instruction
completes we skip the emulated instruction, pointing rip somewhere completely
unrelated.

Fix by updating rip when we see decode the instruction, not after emulation.

Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: popf
Nitin A Kamble [Sat, 15 Sep 2007 07:45:05 +0000 (10:45 +0300)]
KVM: x86 emulator: popf

Implement emulation of instruction:
    popf
    opcode:  0x9d

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: fix src, dst value initialization
Nitin A Kamble [Sat, 15 Sep 2007 07:43:33 +0000 (10:43 +0300)]
KVM: x86 emulator: fix src, dst value initialization

Some operand fetches are less than the machine word size and can result in
stale bits if used together with operands of different sizes.

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: jmp abs
Nitin A Kamble [Sat, 15 Sep 2007 07:41:26 +0000 (10:41 +0300)]
KVM: x86 emulator: jmp abs

Implement emulation of instruction:
    jump absolute r/m
    opcode: 0xff /4

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: lea
Nitin A Kamble [Sat, 15 Sep 2007 07:35:36 +0000 (10:35 +0300)]
KVM: x86 emulator: lea

Implement emulation of instruction
    lea r16/r32, m
    opcode:  0x8d:

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: X86 emulator: jump conditional short
Nitin A Kamble [Sat, 15 Sep 2007 07:25:41 +0000 (10:25 +0300)]
KVM: X86 emulator: jump conditional short

Implement emulation of more jump conditional instructions
    jcc shortrel
    opcodes: 0x70 - 0x7f

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: imlpement jump conditional relative
Nitin A Kamble [Sat, 15 Sep 2007 07:23:07 +0000 (10:23 +0300)]
KVM: x86 emulator: imlpement jump conditional relative

Implement emulation of instruction:
    jump conditional rel
    opcodes: 0x0f 0x80 - 0x0f 0x8f

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: sort opcodes into ascending order
Nitin A Kamble [Sat, 15 Sep 2007 07:13:07 +0000 (10:13 +0300)]
KVM: x86 emulator: sort opcodes into ascending order

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Improve emulation failure reporting
Avi Kivity [Wed, 12 Sep 2007 10:21:09 +0000 (13:21 +0300)]
KVM: Improve emulation failure reporting

Report failed opcodes from all locations.

Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: pushf
Nitin A Kamble [Wed, 29 Aug 2007 01:22:47 +0000 (18:22 -0700)]
KVM: x86 emulator: pushf

Implement emulation of instruction
pushf
opcode: 0x9c

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: call near
Nitin A Kamble [Wed, 29 Aug 2007 01:08:37 +0000 (18:08 -0700)]
KVM: x86 emulator: call near

Implement emulation of instruction
opcode: 0xe8
call (near)

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: x86 emulator: push imm8
Nitin A Kamble [Wed, 29 Aug 2007 00:58:52 +0000 (17:58 -0700)]
KVM: x86 emulator: push imm8

Implement the instruction

     push imm8
     opcode: 0x6a

Signed-off-by: Nitin A Kamble <nitin.a.kamble@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: VMX: Fix exit qualification width on i386
He, Qing [Wed, 12 Sep 2007 06:18:28 +0000 (14:18 +0800)]
KVM: VMX: Fix exit qualification width on i386

According to Intel Software Developer's Manual, Vol. 3B, Appendix H.4.2,
exit qualification should be of natural width. However, current code
uses u64 as the data type for this register, which occasionally
introduces invalid value to VMExit handling logics. This patch fixes
this bug.

I have tested Windows and Linux guest on i386 host, and they can boot
successfully with this patch.

Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Move main vcpu loop into subarch independent code
Avi Kivity [Mon, 10 Sep 2007 15:10:54 +0000 (18:10 +0300)]
KVM: Move main vcpu loop into subarch independent code

This simplifies adding new code as well as reducing overall code size.

Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: VMX: Move vm entry failure handling to the exit handler
Avi Kivity [Mon, 10 Sep 2007 14:27:03 +0000 (17:27 +0300)]
KVM: VMX: Move vm entry failure handling to the exit handler

This will help moving the main loop to subarch independent code.

Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: MMU: Don't do GFP_NOWAIT allocations
Avi Kivity [Mon, 10 Sep 2007 08:28:17 +0000 (11:28 +0300)]
KVM: MMU: Don't do GFP_NOWAIT allocations

Before preempt notifiers, kvm needed to allocate memory with GFP_NOWAIT so
as not to have to enable preemption and take a heavyweight exit.  On oom, we'd
fall back to a GFP_KERNEL allocation.

With preemption notifiers, we can do a GFP_KERNEL allocation, and perform
the heavyweight exit only if the kernel decides to put us to sleep.

Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Rename kvm_arch_ops to kvm_x86_ops
Christian Ehrhardt [Sun, 9 Sep 2007 12:41:59 +0000 (15:41 +0300)]
KVM: Rename kvm_arch_ops to kvm_x86_ops

This patch just renames the current (misnamed) _arch namings to _x86 to
ensure better readability when a real arch layer takes place.

Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Simplify memory allocation
Laurent Vivier [Thu, 30 Aug 2007 12:56:21 +0000 (14:56 +0200)]
KVM: Simplify memory allocation

The mutex->splinlock convertion alllows us to make some code simplifications.
As we can keep the lock longer, we don't have to release it and then
have to check if the environment has not been modified before re-taking it. We
can remove kvm->busy and kvm->memory_config_version.

Signed-off-by: Laurent Vivier <Laurent.Vivier@bull.net>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Hoist SVM's get_cs_db_l_bits into core code.
Rusty Russell [Wed, 5 Sep 2007 15:21:32 +0000 (01:21 +1000)]
KVM: Hoist SVM's get_cs_db_l_bits into core code.

SVM gets the DB and L bits for the cs by decoding the segment.  This
is in fact the completely generic code, so hoist it for kvm-lite to use.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Keep control regs in sync
Rusty Russell [Wed, 5 Sep 2007 15:20:38 +0000 (01:20 +1000)]
KVM: Keep control regs in sync

We don't update the vcpu control registers in various places.  We
should do so.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Clean up unloved invlpg emulation
Rusty Russell [Sun, 9 Sep 2007 11:12:54 +0000 (14:12 +0300)]
KVM: Clean up unloved invlpg emulation

invlpg shouldn't fetch the "src" address, since it may not be valid,
however SVM's "solution" which neuters emulation of all group 7
instruction is horrible and breaks kvm-lite.  The simplest fix is to
put a special check in for invlpg.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Remove the unused invlpg member of struct kvm_arch_ops.
Rusty Russell [Sun, 9 Sep 2007 11:10:57 +0000 (14:10 +0300)]
KVM: Remove the unused invlpg member of struct kvm_arch_ops.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Set the ET flag in CR0 after initializing FX
Amit Shah [Sat, 25 Aug 2007 08:35:52 +0000 (11:35 +0300)]
KVM: Set the ET flag in CR0 after initializing FX

This was missed when moving stuff around in fbc4f2e

Fixes Solaris guests and bug #1773613

Signed-off-by: Amit Shah <amit.shah@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: enable in-kernel APIC INIT/SIPI handling
He, Qing [Mon, 3 Sep 2007 14:07:41 +0000 (17:07 +0300)]
KVM: enable in-kernel APIC INIT/SIPI handling

This patch enables INIT/SIPI handling using in-kernel APIC by
introducing a ->mp_state field to emulate the SMP state transition.

[avi: remove smp_processor_id() warning]

Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Xin Li <xin.b.li@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: round robin for APIC lowest priority delivery mode
He, Qing [Mon, 3 Sep 2007 14:01:36 +0000 (17:01 +0300)]
KVM: round robin for APIC lowest priority delivery mode

Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: deliver PIC interrupt only to vcpu0
Qing He [Mon, 17 Sep 2007 06:47:13 +0000 (14:47 +0800)]
KVM: deliver PIC interrupt only to vcpu0

This patch changes the PIC interrupts delivery. Now it is only delivered
to vcpu0 when either condition is met (on vcpu0):
  1. local APIC is hardware disabled
  2. LVT0 is unmasked and configured to delivery mode ExtInt

It fixes the 2x faster wall clock on x86_64 and SMP i386 Linux guests

Signed-off-by: Eddie (Yaozu) Dong <eddie.dong@intel.com>
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: disable tpr/cr8 sync when in-kernel APIC is used
He, Qing [Thu, 30 Aug 2007 09:04:26 +0000 (17:04 +0800)]
KVM: disable tpr/cr8 sync when in-kernel APIC is used

Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Migrate lapic hrtimer when vcpu moves to another cpu
Eddie Dong [Mon, 3 Sep 2007 13:15:12 +0000 (16:15 +0300)]
KVM: Migrate lapic hrtimer when vcpu moves to another cpu

This reduces overhead by accessing cachelines from the wrong node, as well
as simplifying locking.

[Qing: fix for inactive or expired one-shot timer]

Signed-off-by: Yaozu (Eddie) Dong <Eddie.Dong@intel.com>
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: Keep track of missed timer irq injections
Eddie Dong [Mon, 3 Sep 2007 13:56:58 +0000 (16:56 +0300)]
KVM: Keep track of missed timer irq injections

APIC timer IRQ is set every time when a certain period
expires at host time, but the guest may be descheduled
at that time and thus the irq be overwritten by later fire.
This patch keep track of firing irq numbers and decrease
only when the IRQ is injected to guest or buffered in
APIC.

Signed-off-by: Yaozu (Eddie) Dong <Eddie.Dong@intel.com>
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
17 years agoKVM: VMX: Use shadow TPR/cr8 for 64-bits guests
Yang, Sheng [Wed, 12 Sep 2007 10:03:11 +0000 (18:03 +0800)]
KVM: VMX: Use shadow TPR/cr8 for 64-bits guests

This patch enables TPR shadow of VMX on CR8 access. 64bit Windows using
CR8 access TPR frequently. The TPR shadow can improve the performance of
access TPR by not causing vmexit.

Signed-off-by: Sheng Yang <sheng.yang@intel.com>
Signed-off-by: Yaozu (Eddie) Dong <eddie.dong@intel.com>
Signed-off-by: Qing He <qing.he@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>