linux-2.6
16 years agoHID: move cypress quirks
Jiri Slaby [Mon, 23 Jun 2008 20:54:08 +0000 (22:54 +0200)] 
HID: move cypress quirks

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: move sunplus quirks
Jiri Slaby [Mon, 23 Jun 2008 19:56:07 +0000 (21:56 +0200)] 
HID: move sunplus quirks

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: move microsoft quirks
Jiri Slaby [Fri, 20 Jun 2008 19:26:11 +0000 (21:26 +0200)] 
HID: move microsoft quirks

Move them from the core code to a separate driver.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: indent switches/cases
Jiri Slaby [Wed, 18 Jun 2008 21:55:41 +0000 (23:55 +0200)] 
HID: indent switches/cases

Bring switch and cases into coding style and save thus some
indentation to make the code tighter.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: add compat support
Jiri Slaby [Fri, 16 May 2008 09:49:22 +0000 (11:49 +0200)] 
HID: add compat support

Add compat option to hid code to allow loading of all modules on
systems which don't allow autoloading because of old userspace.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: move apple quirks
Jiri Slaby [Wed, 18 Jun 2008 21:36:49 +0000 (23:36 +0200)] 
HID: move apple quirks

Move them from the core code to a separate driver.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: move ignore quirks
Jiri Slaby [Fri, 16 May 2008 09:49:20 +0000 (11:49 +0200)] 
HID: move ignore quirks

Move ignore quirks from usbhid-quirks into hid-core code. Also don't output
warning when ENODEV is error code in usbhid and try ordinal input in hidp
when that error is returned.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: move logitech quirks
Jiri Slaby [Fri, 16 May 2008 09:49:19 +0000 (11:49 +0200)] 
HID: move logitech quirks

Move them from the core and input code to a separate driver.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: move usage input mapping to hid.h
Jiri Slaby [Fri, 16 May 2008 09:49:18 +0000 (11:49 +0200)] 
HID: move usage input mapping to hid.h

This mapping are currently used on 2 placces and will be needed by more
quirk drivers, so move them to hid.h to allow them to use it.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: move ids into separate file
Jiri Slaby [Wed, 4 Jun 2008 09:02:56 +0000 (11:02 +0200)] 
HID: move ids into separate file

Move ids from hid-quirks.c into separate file, since it will be needed in
more than one place.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: hid, make parsing event driven
Jiri Slaby [Fri, 16 May 2008 09:49:16 +0000 (11:49 +0200)] 
HID: hid, make parsing event driven

Next step for complete hid bus, this patch includes:
- call parser either from probe or from hid-core if there is no probe.
- add ll_driver structure and centralize some stuff there (open, close...)
- split and merge usb_hid_configure and hid_probe into several functions
  to allow hooks/fixes between them

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoHID: make a bus from hid code
Jiri Slaby [Fri, 16 May 2008 09:49:15 +0000 (11:49 +0200)] 
HID: make a bus from hid code

Make a bus from hid core. This is the first step for converting all the
quirks and separate almost-drivers into real drivers attached to this bus.

It's implemented to change behaviour in very tiny manner, so that no driver
needs to be changed this time.

Also add generic drivers for both usb and bt into usbhid or hidp
respectively which will bind all non-blacklisted device. Those blacklisted
will be either grabbed by special drivers or by nobody if they are broken at
the very rude base.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agomodpost: add support for hid
Jiri Slaby [Mon, 19 May 2008 13:50:01 +0000 (15:50 +0200)] 
modpost: add support for hid

Generate aliases for hid device modules to support autoloading.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
16 years agoMerge branch 'for-2.6.28' of git://linux-nfs.org/~bfields/linux
Linus Torvalds [Tue, 14 Oct 2008 19:31:14 +0000 (12:31 -0700)] 
Merge branch 'for-2.6.28' of git://linux-nfs.org/~bfields/linux

* 'for-2.6.28' of git://linux-nfs.org/~bfields/linux: (59 commits)
  svcrdma: Fix IRD/ORD polarity
  svcrdma: Update svc_rdma_send_error to use DMA LKEY
  svcrdma: Modify the RPC reply path to use FRMR when available
  svcrdma: Modify the RPC recv path to use FRMR when available
  svcrdma: Add support to svc_rdma_send to handle chained WR
  svcrdma: Modify post recv path to use local dma key
  svcrdma: Add a service to register a Fast Reg MR with the device
  svcrdma: Query device for Fast Reg support during connection setup
  svcrdma: Add FRMR get/put services
  NLM: Remove unused argument from svc_addsock() function
  NLM: Remove "proto" argument from lockd_up()
  NLM: Always start both UDP and TCP listeners
  lockd: Remove unused fields in the nlm_reboot structure
  lockd: Add helper to sanity check incoming NOTIFY requests
  lockd: change nlmclnt_grant() to take a "struct sockaddr *"
  lockd: Adjust nlmsvc_lookup_host() to accomodate AF_INET6 addresses
  lockd: Adjust nlmclnt_lookup_host() signature to accomodate non-AF_INET
  lockd: Support non-AF_INET addresses in nlm_lookup_host()
  NLM: Convert nlm_lookup_host() to use a single argument
  svcrdma: Add Fast Reg MR Data Types
  ...

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/lrg/voltage-2.6
Linus Torvalds [Tue, 14 Oct 2008 19:28:55 +0000 (12:28 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/lrg/voltage-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/lrg/voltage-2.6: (26 commits)
  mfd: Fix warning in WM8350
  mfd: Add placeholders for WM8350 client devices
  da903x: add regulator support for DA9030/DA9034
  mfd: Add WM8350 subdevice registration helper
  regulator: Add WM8350 regulator support
  mfd: Add WM8350 interrupt support
  mfd: Add initialisation callback for WM8350
  mfd: Add GPIO pin configuration support for WM8350
  mfd: Add I2C control support for WM8350
  mfd: Core support for the WM8350 AudioPlus PMIC
  mfd: Add WM8350 watchdog register definitions
  mfd: Add WM8350 RTC register definitions
  mfd: Add WM8350 comparator register definitions
  mfd: Add WM8350 PMU register definitions
  mfd: Add WM8350 PMIC register definitions
  mfd: Add WM8350 GPIO register definitions
  mfd: Add WM8350 audio register definitions
  regulator: Export regulator name via sysfs
  regulator: Add WM8400 regulator support
  mfd: Core support for the WM8400 AudioPlus HiFi CODEC and PMU
  ...

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arjan/linux...
Linus Torvalds [Tue, 14 Oct 2008 19:28:02 +0000 (12:28 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/arjan/linux-2.6-fastboot

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arjan/linux-2.6-fastboot:
  raid, fastboot: hide RAID autodetect option if MD is compiled as a module
  raid: make RAID autodetect default a KConfig option
  warning: fix init do_mounts_md c
  fastboot: make the RAID autostart code print a message just before waiting
  fastboot: make the raid autodetect code wait for all devices to init
  fastboot: Fix bootgraph.pl initcall name regexp
  fastboot: fix issues and improve output of bootgraph.pl
  Add a script to visualize the kernel boot process / time

16 years agoRevert "vt: fix background color on line feed"
Linus Torvalds [Tue, 14 Oct 2008 19:12:02 +0000 (12:12 -0700)] 
Revert "vt: fix background color on line feed"

This reverts commit c9e587abfdec2c2aaa55fab83bcb4972e2f84f9b, and the
subsequent commits that fixed it up:

 - afa9b649 "fbcon: prevent cursor disappearance after switching to 512
   character font"

 - d850a2fa "vt/fbcon: fix background color on line feed"

 - 7fe3915a "vt/fbcon: update scrl_erase_char after 256/512-glyph font
   switch"

by request of Alan Cox. Quoth Alan:
  "Unfortunately it's wrong and its been causing breakages because
   various apps like ncurses expect our previous (and correct)
   behaviour."

Alexander sent out a similar patch.

Requested-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Tested-by: Jan Engelhardt <jengelh@medozas.de>
Cc: Alexander V. Lukyanov <lav@netis.ru>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoocfs2: Make cached block reads the common case.
Joel Becker [Fri, 10 Oct 2008 00:20:34 +0000 (17:20 -0700)] 
ocfs2: Make cached block reads the common case.

ocfs2_read_blocks() currently requires the CACHED flag for cached I/O.
However, that's the common case.  Let's flip it around and provide an
IGNORE_CACHE flag for the special users.  This has the added benefit of
cleaning up the code some (ignore_cache takes on its special meaning
earlier in the loop).

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Kill the last naked wait_on_buffer() for cached reads.
Joel Becker [Fri, 10 Oct 2008 00:20:33 +0000 (17:20 -0700)] 
ocfs2: Kill the last naked wait_on_buffer() for cached reads.

ocfs2's cached buffer I/O goes through ocfs2_read_block(s)().  dir.c had
a naked wait_on_buffer() to wait for some readahead, but it should
use ocfs2_read_block() instead.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Move ocfs2_bread() into dir.c
Joel Becker [Fri, 10 Oct 2008 00:20:32 +0000 (17:20 -0700)] 
ocfs2: Move ocfs2_bread() into dir.c

dir.c is the only place using ocfs2_bread(), so let's make it static to
that file.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Simplify ocfs2_read_block()
Joel Becker [Fri, 10 Oct 2008 00:20:31 +0000 (17:20 -0700)] 
ocfs2: Simplify ocfs2_read_block()

More than 30 callers of ocfs2_read_block() pass exactly OCFS2_BH_CACHED.
Only six pass a different flag set.  Rather than have every caller care,
let's make ocfs2_read_block() take no flags and always do a cached read.
The remaining six places can call ocfs2_read_blocks() directly.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Require an inode for ocfs2_read_block(s)().
Joel Becker [Fri, 10 Oct 2008 00:20:30 +0000 (17:20 -0700)] 
ocfs2: Require an inode for ocfs2_read_block(s)().

Now that synchronous readers are using ocfs2_read_blocks_sync(), all
callers of ocfs2_read_blocks() are passing an inode.  Use it
unconditionally.  Since it's there, we don't need to pass the
ocfs2_super either.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Separate out sync reads from ocfs2_read_blocks()
Joel Becker [Fri, 10 Oct 2008 00:20:29 +0000 (17:20 -0700)] 
ocfs2: Separate out sync reads from ocfs2_read_blocks()

The ocfs2_read_blocks() function currently handles sync reads, cached,
reads, and sometimes cached reads.  We're going to add some
functionality to it, so first we should simplify it.  The uncached,
synchronous reads are much easer to handle as a separate function, so we
instroduce ocfs2_read_blocks_sync().

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Tue, 14 Oct 2008 17:28:49 +0000 (10:28 -0700)] 
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  qlge: Fix page size ifdef test.
  net: Rationalise email address: Network Specific Parts
  dsa: fix compile bug on s390
  netns: mib6 section fixlet
  enic: Fix Kconfig headline description
  de2104x: wrong MAC address fix
  s390: claw compile fixlet
  net: export genphy_restart_aneg
  cxgb3: extend copyrights to 2008
  cxgb3: update driver version
  net/phy: add missing kernel-doc
  pktgen: fix skb leak in case of failure
  mISDN/dsp_cmx.c: fix size checks
  misdn: use nonseekable_open()
  net: fix driver build errors due to missing net/ip6_checksum.h include

16 years agoarch/m68k/mm/kmap.c: introduce missing kfree
Julia Lawall [Mon, 13 Oct 2008 19:59:04 +0000 (21:59 +0200)] 
arch/m68k/mm/kmap.c: introduce missing kfree

Error handling code following a kmalloc should free the allocated data.

The semantic match that finds the problem is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@r exists@
local idexpression x;
statement S;
expression E;
identifier f,l;
position p1,p2;
expression *ptr != NULL;
@@

(
if ((x@p1 = \(kmalloc\|kzalloc\|kcalloc\)(...)) == NULL) S
|
x@p1 = \(kmalloc\|kzalloc\|kcalloc\)(...);
..
if (x == NULL) S
)
<... when != x
     when != if (...) { <+...x...+> }
x->f = E
..>
(
 return \(0\|<+...x...+>\|ptr\);
|
 return@p2 ...;
)

@script:python@
p1 << r.p1;
p2 << r.p2;
@@

print "* file: %s kmalloc %s return %s" % (p1[0].file,p1[0].line,p2[0].line)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agonet/rfkill/rfkill-input.c needs <linux/sched.h>
Geert Uytterhoeven [Mon, 13 Oct 2008 19:59:03 +0000 (21:59 +0200)] 
net/rfkill/rfkill-input.c needs <linux/sched.h>

For some m68k configs, I get:

| net/rfkill/rfkill-input.c: In function 'rfkill_start':
| net/rfkill/rfkill-input.c:208: error: dereferencing pointer to incomplete type

As the incomplete type is `struct task_struct', including <linux/sched.h> fixes
it.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Atari SCSI needs NVRAM
Geert Uytterhoeven [Mon, 13 Oct 2008 19:59:02 +0000 (21:59 +0200)] 
m68k: Atari SCSI needs NVRAM

ERROR: "nvram_read_byte" [drivers/scsi/atari_scsi.ko] undefined!
ERROR: "nvram_check_checksum" [drivers/scsi/atari_scsi.ko] undefined!

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: init_irq_proc depends on CONFIG_PROC_FS
Geert Uytterhoeven [Mon, 13 Oct 2008 19:59:01 +0000 (21:59 +0200)] 
m68k: init_irq_proc depends on CONFIG_PROC_FS

If CONFIG_PROC_FS is not set, I get:

| arch/m68k/kernel/ints.c:433: error: redefinition of 'init_irq_proc'
| include/linux/interrupt.h:438: error: previous definition of 'init_irq_proc' was here

This was introduced by commit 6168a702ab0be181e5e57a0b2d0e7376f7a47f0b
("Declare init_irq_proc before we use it."), which replaced the #ifdef
protection of the init_irq_proc() call by a static inline dummy if
CONFIG_PROC_FS is not set.

Make init_irq_proc() depend on CONFIG_PROC_FS to fix this.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: remove the dead PCI code
Adrian Bunk [Mon, 13 Oct 2008 19:59:00 +0000 (21:59 +0200)] 
m68k: remove the dead PCI code

This patch removes the no longer used m68k PCI code.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Remove the broken Hades support
Adrian Bunk [Mon, 13 Oct 2008 19:58:59 +0000 (21:58 +0200)] 
m68k: Remove the broken Hades support

This patch removes the Hades support that was marked as BROKEN 5 years ago.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoHP input: kill warnings due to suseconds_t differences
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:58 +0000 (21:58 +0200)] 
HP input: kill warnings due to suseconds_t differences

Kill compiler warnings related to printf() formats in the input drivers for
various HP9000 machines, which are shared between PA-RISC (suseconds_t is int)
and m68k (suseconds_t is long). As both are 32-bit, it's safe to cast to int.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Helge Deller <deller@gmx.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: <asm/pci.h> needs <asm-generic/pci-dma-compat.h>
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:57 +0000 (21:58 +0200)] 
m68k: <asm/pci.h> needs <asm-generic/pci-dma-compat.h>

Several multi-bus subsystems:

| include/linux/ssb/ssb.h: In function 'ssb_dma_mapping_error':
| include/linux/ssb/ssb.h:430: error: implicit declaration of function 'pci_dma_mapping_error'
| include/linux/ssb/ssb.h: In function 'ssb_dma_map_single':
| include/linux/ssb/ssb.h:444: error: implicit declaration of function 'pci_map_single'
| include/linux/ssb/ssb.h: In function 'ssb_dma_unmap_single':
| include/linux/ssb/ssb.h:458: error: implicit declaration of function 'pci_unmap_single'
| include/linux/ssb/ssb.h: In function 'ssb_dma_sync_single_for_cpu':
| include/linux/ssb/ssb.h:475: error: implicit declaration of function 'pci_dma_sync_single_for_cpu'
| include/linux/ssb/ssb.h: In function 'ssb_dma_sync_single_for_device':
| include/linux/ssb/ssb.h:493: error: implicit declaration of function 'pci_dma_sync_single_for_device'

or legacy drivers:

| drivers/net/hp100.c: In function 'pdl_map_data':
| drivers/net/hp100.c:291: error: implicit declaration of function 'pci_map_single'
| drivers/net/hp100.c: In function 'hp100_probe1':
| drivers/net/hp100.c:707: error: implicit declaration of function 'pci_alloc_consistent'
| drivers/net/hp100.c:782: error: implicit declaration of function 'pci_free_consistent'
| drivers/net/hp100.c: In function 'hp100_clean_txring':
| drivers/net/hp100.c:1614: error: implicit declaration of function 'pci_unmap_single'

and

| drivers/scsi/aic7xxx_old.c: In function 'aic7xxx_allocate_scb':
| drivers/scsi/aic7xxx_old.c:2573: error: implicit declaration of function 'pci_alloc_consistent'
| drivers/scsi/aic7xxx_old.c: In function 'aic7xxx_done':
| drivers/scsi/aic7xxx_old.c:2697: error: implicit declaration of function 'pci_unmap_single'
| drivers/scsi/aic7xxx_old.c: In function 'aic7xxx_handle_seqint':
| drivers/scsi/aic7xxx_old.c:4275: error: implicit declaration of function 'pci_map_single'
| drivers/scsi/aic7xxx_old.c: In function 'aic7xxx_free':
| drivers/scsi/aic7xxx_old.c:8460: error: implicit declaration of function 'pci_free_consistent'

rely on PCI DMA operations to be always available.

Add #include <asm-generic/pci-dma-compat.h> to <asm/pci.h> to make them happy.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Add missing dma_sync_single_range_for_{cpu,device}()
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:56 +0000 (21:58 +0200)] 
m68k: Add missing dma_sync_single_range_for_{cpu,device}()

| include/linux/ssb/ssb.h: In function 'ssb_dma_sync_single_range_for_cpu':
| include/linux/ssb/ssb.h:517: error: implicit declaration of function 'dma_sync_single_range_for_cpu'
| include/linux/ssb/ssb.h: In function 'ssb_dma_sync_single_range_for_device':
| include/linux/ssb/ssb.h:538: error: implicit declaration of function 'dma_sync_single_range_for_device'

Add the missing dma_sync_single_range_for_{cpu,device}(), and remove the
`inline' for the non-static function dma_sync_single_for_device().

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Define rtc_lock on Atari
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:55 +0000 (21:58 +0200)] 
m68k: Define rtc_lock on Atari

The nvram and rtc-cmos drivers use the spinlock rtc_lock to protect against
concurrent accesses to the CMOS memory. As m68k doesn't support SMP or preempt
yet, the spinlock calls tend to get optimized away, but not for all
configurations, causing in some rare cases:

| ERROR: "rtc_lock" [drivers/rtc/rtc-cmos.ko] undefined!
| ERROR: "rtc_lock" [drivers/char/nvram.ko] undefined!

Add the spinlock to the Atari core code to avoid this.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Michael Schmitz <schmitz@debian.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Remove unused atari_kbd_translate()
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:54 +0000 (21:58 +0200)] 
m68k: Remove unused atari_kbd_translate()

If CONFIG_VT=n, I get:

| arch/m68k/atari/built-in.o: In function `atari_kbd_translate':
| arch/m68k/atari/atakeyb.c:640: undefined reference to `shift_state'

Just remove atari_kbd_translate(), as it's unused.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Michael Schmitz <schmitz@debian.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Modular Amiga keyboard needs key_maps
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:53 +0000 (21:58 +0200)] 
m68k: Modular Amiga keyboard needs key_maps

| ERROR: "key_maps" [drivers/input/keyboard/amikbd.ko] undefined!

Export key_maps in the Amiga core code, as its defined in an autogenerated
file (drivers/char/defkeymap.c)

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Reverse platform MMU logic so Sun 3 is last
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:51 +0000 (21:58 +0200)] 
m68k: Reverse platform MMU logic so Sun 3 is last

Currently Sun 3 support is the first platform option, as the Sun 3 MMU is
incompatible with standard Motorola MMUs. However, this means that
`allmodconfig' enables support for Sun 3, and thus disables support for all
other platforms.

Reverse the logic and move Sun 3 last, so `allmodconfig' enables all
platforms except for Sun 3, increasing compile-coverage.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Add NOTES to init data so its discarded at boot
Roman Zippel [Mon, 13 Oct 2008 19:58:50 +0000 (21:58 +0200)] 
m68k: Add NOTES to init data so its discarded at boot

Add .note.gnu.build-id to init data so it's discarded at boot.

[Andreas Schwab] Use NOTES macro

Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Put .bss at the end of the data section
Roman Zippel [Mon, 13 Oct 2008 19:58:49 +0000 (21:58 +0200)] 
m68k: Put .bss at the end of the data section

Put .bss at the end of the data section

Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: Use new printk() extension %pS to print symbols
Geert Uytterhoeven [Mon, 13 Oct 2008 19:58:48 +0000 (21:58 +0200)] 
m68k: Use new printk() extension %pS to print symbols

This changes the oops and backtrace code to use the new `%pS' printk()
extension to print out symbols rather than manually calling print_symbol.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agom68k: use bcd2bin/bin2bcd
Adrian Bunk [Mon, 13 Oct 2008 19:58:47 +0000 (21:58 +0200)] 
m68k: use bcd2bin/bin2bcd

This patch changes m68k to use the new bcd2bin/bin2bcd functions instead
of the obsolete BCD_TO_BIN/BIN_TO_BCD/BCD2BIN/BIN2BCD macros.

It also remove local bcd2bin/bin2bcd implementations
in favor of the global ones.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years ago8250: Fix lock warning (and possible crash)
Alan Cox [Tue, 14 Oct 2008 10:29:06 +0000 (11:29 +0100)] 
8250: Fix lock warning (and possible crash)

Splitting the 8250 code back up to avoid a clash with the NR_IRQS removal
patch introduced a last minute bug. Put back the additional needed lines
for the old lock init

Signed-off-by: Alan Cox <alan@redhat.com>
[ Ingo also reports that this can cause a spontaneous reboot crash with
  certain configs, and sends in an identical patch ]
Tested-by: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoi2c-viapro: Add support for SMBus Process Call transactions
Prakash Mortha [Tue, 14 Oct 2008 15:30:06 +0000 (17:30 +0200)] 
i2c-viapro: Add support for SMBus Process Call transactions

Add support for SMBus Process Call transactions. These are combined
word write, word read transactions.

Signed-off-by: Prakash Mortha <pmortha@escient.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Restore i2c_smbus_process_call function
Prakash Mortha [Tue, 14 Oct 2008 15:30:06 +0000 (17:30 +0200)] 
i2c: Restore i2c_smbus_process_call function

Restore the i2c_smbus_process_call() as one driver (for the
Micronas MAP5401) will need it soon.

[JD: Update documentation accordingly.]

Signed-off-by: Prakash Mortha <pmortha@escient.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Do earlier driver model init
David Brownell [Tue, 14 Oct 2008 15:30:06 +0000 (17:30 +0200)] 
i2c: Do earlier driver model init

Move I2C driver model init earlier in the boot sequence.

This avoids oopsing in statically linked systems when some
subsystems register I2C drivers in subsys_initcall() code,
but those subsystems are linked (and initialized) before I2C.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Only build Tyan SMBus mux drivers on x86
Jean Delvare [Tue, 14 Oct 2008 15:30:06 +0000 (17:30 +0200)] 
i2c: Only build Tyan SMBus mux drivers on x86

The two Tyan SMBus mux drivers (i2c-amd756-s4882 and i2c-nforce2-s4985)
are only useful on specific x86 motherboards, so there is no point in
letting them be built on other architectures.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Guard against oopses from bad init sequences
David Brownell [Tue, 14 Oct 2008 15:30:05 +0000 (17:30 +0200)] 
i2c: Guard against oopses from bad init sequences

Guard I2C against oopsing because of init sequence problems, by
verifying that i2c_init() has been called before calling any
routines that rely on that initialization.  This specific test
just requires that bus_register(&i2c_bus_type) was called.

Examples of this kind of oopsing come from subystems and drivers
which register I2C drivers in their subsys_initcall code but
which are statically linked before I2C by drivers/Makefile.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Document the implementation details of the /dev interface
Jean Delvare [Tue, 14 Oct 2008 15:30:05 +0000 (17:30 +0200)] 
i2c: Document the implementation details of the /dev interface

I wrote this explanation to answer a question on the i2c mailing list,
and thought it would be good to have in the kernel documentation.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Improve dev-interface documentation
Jean Delvare [Tue, 14 Oct 2008 15:30:05 +0000 (17:30 +0200)] 
i2c: Improve dev-interface documentation

* Clarify some points.
* Point developers to i2c-tools instead of lm_sensors.
* Fix coding style in code examples.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c-parport-light: Don't register a platform device resource
Jean Delvare [Tue, 14 Oct 2008 15:30:04 +0000 (17:30 +0200)] 
i2c-parport-light: Don't register a platform device resource

The i2c-parport-light driver isn't a real platform driver, so it
should not instantiate platform devices with resources. The resource
management system can't cope with colliding resources, and we are
likely to create such a colliding resource.

So, better just try to grab the I/O ports we need right at module
initialization time, and bail out if we can't. It has the added
benefit that the module will no longer load if it isn't going to work,
which is definitely more user-friendly.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agohwmon: (dme1737) Convert to a new-style i2c driver
Jean Delvare [Tue, 14 Oct 2008 15:30:04 +0000 (17:30 +0200)] 
hwmon: (dme1737) Convert to a new-style i2c driver

The new-style dme1737 driver implements the optional detect() callback
to cover the use cases of the legacy driver. I don't actually expect
any new-style device for that driver, but as the old i2c API is going
away soon, we have to switch to the new one.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: Juerg Haefliger <juergh@gmail.com>
16 years agohwmon: (dme1737) Be less i2c-centric
Jean Delvare [Tue, 14 Oct 2008 15:30:04 +0000 (17:30 +0200)] 
hwmon: (dme1737) Be less i2c-centric

The dme1737 driver support both LPC (ISA) and SMBus devices. At the
moment it's rather i2c-centric, and LPC variants use a fake i2c_client
for some operations.

In a near future, i2c_client will be allocated by i2c-core rather than
by the device drivers, so non-i2c drivers will not have one. As a
preparation step, change the driver code to no longer assume that
an i2c_client structure is always available. No functional change.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: Juerg Haefliger <juergh@gmail.com>
16 years agoi2c/tps65010: Vibrator hookup to gpiolib
Marek Vasut [Tue, 14 Oct 2008 15:30:04 +0000 (17:30 +0200)] 
i2c/tps65010: Vibrator hookup to gpiolib

All the tps6501{0,1,2,3,4} chips have a signal for hooking up with
a vibrator (for non-auditory cell phone "ring") ... expose that as
one more (output-only) GPIO.

[ dbrownell@users.sourceforge.net: comments; list tps65014 too ]

Signed-off-by: Marek Vasut <marek.vasut@gmail.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c-viapro: Add VX800/VX820 support
Rudolf Marek [Tue, 14 Oct 2008 15:30:03 +0000 (17:30 +0200)] 
i2c-viapro: Add VX800/VX820 support

Thanks to new datasheets published on http://linux.via.com.tw we can now add
support for VX800/VX820 chipsets.

Signed-off-by: Rudolf Marek <r.marek@assembler.cz>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c: Renesas Highlander FPGA SMBus support
Paul Mundt [Tue, 14 Oct 2008 15:30:03 +0000 (17:30 +0200)] 
i2c: Renesas Highlander FPGA SMBus support

This adds support for the SMBus adapter found in the various FPGAs on
the Renesas Highlander platforms. Particularly the R0P7780LC0011RL and
R0P7785LC0011RL FPGAs.

Functionality is fairly restricted, in that only byte and block data
transfers are supported. Normal/fast mode and IRQ/polling are also
supported. Primarily used for various RTCs and thermal sensors.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c-pca-isa: Don't grab arbitrary resources
Rene Herman [Tue, 14 Oct 2008 15:30:03 +0000 (17:30 +0200)] 
i2c-pca-isa: Don't grab arbitrary resources

Grabbing ISA bus resources without anything or anyone telling us we
should can break boot on randconfig/allyesconfig builds by keeping
resources that are in fact owned by different hardware busy and does
as reported by Ingo Molnar.

Generally it's also dangerous to just poke at random I/O ports and
especially those in the range where other old easily confused ISA
hardware might live.

For this specialized I2C bus driver, insist that the user specifies
the resources before grabbing them.

The^WA user of this driver is a one time

echo "options i2c-pca-isa base=0x330 irq=10" >> /etc/modprobe.conf

away from the old behaviour.

Signed-off-by: Rene Herman <rene.herman@gmail.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
16 years agoi2c/isp1301_omap: Convert to a new-style i2c driver, part 2
Jean Delvare [Tue, 14 Oct 2008 15:30:02 +0000 (17:30 +0200)] 
i2c/isp1301_omap: Convert to a new-style i2c driver, part 2

Based on David Brownell's patch for tps65010 and previous work by
Felipe Balbi, this patch finishes converting isp1301_omap to a
new-style i2c driver.

There's definitely room for further drivers cleanups, but these are
out of the scope of this patch.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Tony Lindgren <tony@atomide.com>
16 years agoi2c/isp1301_omap: Convert to a new-style i2c driver, part 1
Felipe Balbi [Tue, 14 Oct 2008 15:30:02 +0000 (17:30 +0200)] 
i2c/isp1301_omap: Convert to a new-style i2c driver, part 1

Based on David Brownell's patch for tps65010, this patch
starts converting isp1301_omap.c to new-style i2c driver.

Signed-off-by: Felipe Balbi <me@felipebalbi.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Tony Lindgren <tony@atomide.com>
16 years agoqlge: Fix page size ifdef test.
Ron Mercer [Tue, 14 Oct 2008 05:55:59 +0000 (22:55 -0700)] 
qlge: Fix page size ifdef test.

This ASIC does support all page sizes. For 4k and 8k page size the TX
control block needs an external scatter gather list.  For page sizes
larger than 8k the max frags is satisfied by the original TX control
block.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agonet: Rationalise email address: Network Specific Parts
Alan Cox [Tue, 14 Oct 2008 02:01:08 +0000 (19:01 -0700)] 
net: Rationalise email address: Network Specific Parts

Clean up the various different email addresses of mine listed in the code
to a single current and valid address. As Dave says his network merges
for 2.6.28 are now done this seems a good point to send them in where
they won't risk disrupting real changes.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agodsa: fix compile bug on s390
Heiko Carstens [Tue, 14 Oct 2008 01:58:48 +0000 (18:58 -0700)] 
dsa: fix compile bug on s390

git commit 45cec1bac0719c904bb5f4405c2937f7e715888c
"dsa: Need to select PHYLIB." causes this build bug on s390:

drivers/built-in.o: In function `phy_stop_interrupts':
/home/heicarst/linux-2.6/drivers/net/phy/phy.c:631: undefined reference to `free_irq'
/home/heicarst/linux-2.6/drivers/net/phy/phy.c:646: undefined reference to `enable_irq'
drivers/built-in.o: In function `phy_start_interrupts':
/home/heicarst/linux-2.6/drivers/net/phy/phy.c:601: undefined reference to `request_irq'
drivers/built-in.o: In function `phy_interrupt':
/home/heicarst/linux-2.6/drivers/net/phy/phy.c:528: undefined reference to `disable_irq_nosync'
drivers/built-in.o: In function `phy_change':
/home/heicarst/linux-2.6/drivers/net/phy/phy.c:674: undefined reference to `enable_irq'
/home/heicarst/linux-2.6/drivers/net/phy/phy.c:692: undefined reference to `disable_irq'

PHYLIB has alread a depend on !S390, however select PHYLIB at DSA overrides
that unfortunately. So add a depend on !S390 to DSA as well.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agonetns: mib6 section fixlet
Alexey Dobriyan [Tue, 14 Oct 2008 01:54:07 +0000 (18:54 -0700)] 
netns: mib6 section fixlet

  LD      net/ipv6/ipv6.o
WARNING: net/ipv6/ipv6.o(.text+0xd8): Section mismatch in reference from the function inet6_net_init() to the function .init.text:ipv6_init_mibs()

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoenic: Fix Kconfig headline description
Roland Dreier [Tue, 14 Oct 2008 01:53:05 +0000 (18:53 -0700)] 
enic: Fix Kconfig headline description

I don't think the enic driver has anything to do with Mark Everett
(http://en.wikipedia.org/wiki/A_Man_Called_E).  Fix the Kconfig
description.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agode2104x: wrong MAC address fix
Martin Langer [Tue, 14 Oct 2008 01:49:38 +0000 (18:49 -0700)] 
de2104x: wrong MAC address fix

The de2104x returns sometimes a wrong MAC address. The wrong one is
like the original one, but it comes with an one byte shift. I found
this bug on an older alpha ev5 cpu. More details are available in Gentoo
bugreport #240718.

It seems the hardware is sometimes a little bit too slow for an
immediate access. This patch solves the problem by introducing a small
udelay.

Signed-off-by: Martin Langer <martin-langer@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agos390: claw compile fixlet
Alexey Dobriyan [Tue, 14 Oct 2008 01:48:43 +0000 (18:48 -0700)] 
s390: claw compile fixlet

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agonet: export genphy_restart_aneg
Adrian Bunk [Tue, 14 Oct 2008 01:48:09 +0000 (18:48 -0700)] 
net: export genphy_restart_aneg

This patch fixes the following build error caused by
commit ed94493fb38a665cebcf750dfabe8a6dd13e136f
(mv643xx_eth: convert to phylib):

<--  snip  -->

...
  Building modules, stage 2.
  MODPOST 1280 modules
ERROR: "genphy_restart_aneg" [drivers/net/mv643xx_eth.ko] undefined!
...
make[2]: *** [__modpost] Error 1

<--  snip  -->

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agocxgb3: extend copyrights to 2008
Divy Le Ray [Tue, 14 Oct 2008 01:47:30 +0000 (18:47 -0700)] 
cxgb3: extend copyrights to 2008

Update copyright banner to 2008.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agocxgb3: update driver version
Divy Le Ray [Tue, 14 Oct 2008 01:47:02 +0000 (18:47 -0700)] 
cxgb3: update driver version

Add a field to the driver versioning info.
Update version to 1.1.0.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agonet/phy: add missing kernel-doc
Randy Dunlap [Tue, 14 Oct 2008 01:46:22 +0000 (18:46 -0700)] 
net/phy: add missing kernel-doc

Fix kernel-doc warning, missing description:

Warning(lin2627-g3-kdocfixes//drivers/net/phy/mdio_bus.c:63): No description found for parameter 'd'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agopktgen: fix skb leak in case of failure
Ilpo Järvinen [Tue, 14 Oct 2008 01:43:59 +0000 (18:43 -0700)] 
pktgen: fix skb leak in case of failure

Seems that skb goes into void unless something magic happened
in pskb_expand_head in case of failure.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agomISDN/dsp_cmx.c: fix size checks
Adrian Bunk [Tue, 14 Oct 2008 01:42:55 +0000 (18:42 -0700)] 
mISDN/dsp_cmx.c: fix size checks

The checks for ensuring that the array indices are inside the range
were flipped.

Reported-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agomisdn: use nonseekable_open()
Andrew Morton [Tue, 14 Oct 2008 01:42:07 +0000 (18:42 -0700)] 
misdn: use nonseekable_open()

The driver just sets ->llseek to NULL.  It should also clear FMODE_LSEEK to
tell the VFS that seeks are not supported.

Pointed out by Christoph Hellwig.

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agonet: fix driver build errors due to missing net/ip6_checksum.h include
Kamalesh Babulal [Tue, 14 Oct 2008 01:41:01 +0000 (18:41 -0700)] 
net: fix driver build errors due to missing net/ip6_checksum.h include

  2.6.27-git2 kernel build fails with allyesconfig on powerpc with
build error

<introduced by commit 01f2e4ead2c51226ed1283ef6a8388ca6f4cff8f>

CC    drivers/net/enic/enic_main.o
drivers/net/enic/enic_main.c: In function â\80\98enic_queue_wq_skb_tsoâ\80\99:
drivers/net/enic/enic_main.c:576: error: implicit declaration of function â\80\98csum_ipv6_magicâ\80\99
make[3]: *** [drivers/net/enic/enic_main.o] Error 1

<introduced by commit c4e84bde1d595d857d3c74b49b9c45cc770df792>

drivers/net/qlge/qlge_main.c: In function â\80\98ql_tsoâ\80\99:
drivers/net/qlge/qlge_main.c:1862: error: implicit declaration of function â\80\98csum_ipv6_magicâ\80\99
make[3]: *** [drivers/net/qlge/qlge_main.o] Error 1

<introduced by commit 95252236e73e789dd186ce796a2abc60b3a61ebe>

drivers/net/jme.c: In function â\80\98jme_tx_tsoâ\80\99:
drivers/net/jme.c:1784: error: implicit declaration of function â\80\98csum_ipv6_magicâ\80\99
make[2]: *** [drivers/net/jme.o] Error 1

Signed-off-by: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
16 years agoocfs2: Refactor xattr list and remove ocfs2_xattr_handler().
Tao Ma [Thu, 9 Oct 2008 15:06:14 +0000 (23:06 +0800)] 
ocfs2: Refactor xattr list and remove ocfs2_xattr_handler().

According to Christoph Hellwig's advice, we really don't need
a ->list to handle one xattr's list. Just a map from index to
xattr prefix is enough. And I also refactor the old list method
with the reference from fs/xfs/linux-2.6/xfs_xattr.c and the
xattr list method in btrfs.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Calculate EA hash only by its suffix.
Tao Ma [Thu, 9 Oct 2008 15:06:13 +0000 (23:06 +0800)] 
ocfs2: Calculate EA hash only by its suffix.

According to Christoph Hellwig's advice, the hash value of EA
is only calculated by its suffix.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Move trusted and user attribute support into xattr.c
Mark Fasheh [Tue, 7 Oct 2008 21:52:59 +0000 (14:52 -0700)] 
ocfs2: Move trusted and user attribute support into xattr.c

Per Christoph Hellwig's suggestion - don't split these up. It's not like we
gained much by having the two tiny files around.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Uninline ocfs2_xattr_name_hash()
Mark Fasheh [Tue, 7 Oct 2008 21:31:42 +0000 (14:31 -0700)] 
ocfs2: Uninline ocfs2_xattr_name_hash()

This is too big to be inlined.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Don't check for NULL before brelse()
Mark Fasheh [Tue, 7 Oct 2008 21:25:16 +0000 (14:25 -0700)] 
ocfs2: Don't check for NULL before brelse()

This is pointless as brelse() already does the check.

Signed-off-by: Mark Fasheh
16 years agoocfs2: use smaller counters in ocfs2_remove_xattr_clusters_from_cache
Mark Fasheh [Tue, 7 Oct 2008 19:50:46 +0000 (12:50 -0700)] 
ocfs2: use smaller counters in ocfs2_remove_xattr_clusters_from_cache

i and b_len don't really need to be u64's. Xattr extent lengths should be
limited by the VFS, and then the size of our on-disk length field.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Documentation update for user_xattr / nouser_xattr mount options
Mark Fasheh [Tue, 7 Oct 2008 18:09:24 +0000 (11:09 -0700)] 
ocfs2: Documentation update for user_xattr / nouser_xattr mount options

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: make la_debug_mutex static
Mark Fasheh [Tue, 7 Oct 2008 18:02:04 +0000 (11:02 -0700)] 
ocfs2: make la_debug_mutex static

It can also be moved into ocfs2_la_debug_read().

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Remove pointless !!
Mark Fasheh [Mon, 6 Oct 2008 23:16:08 +0000 (16:16 -0700)] 
ocfs2: Remove pointless !!

ocfs2_stack_supports_plocks() doesn't need this to properly return a zero or
one value.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add empty bucket support in xattr.
Tao Ma [Fri, 19 Sep 2008 14:17:41 +0000 (22:17 +0800)] 
ocfs2: Add empty bucket support in xattr.

As Mark mentioned, it may be time-consuming when we remove the
empty xattr bucket, so this patch try to let empty bucket exist
in xattr operation. The modification includes:
1. Remove the functin of bucket and extent record deletion during
   xattr delete.
2. In xattr set:
   1) Don't clean the last entry so that if the bucket is empty,
      the hash value of the bucket is the hash value of the entry
      which is deleted last.
   2) During insert, if we meet with an empty bucket, just use the
      1st entry.
3. In binary search of xattr bucket, use the bucket hash value(which
   stored in the 1st xattr entry) to find the right place.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2/xattr.c: Fix a bug when inserting xattr.
Tao Ma [Fri, 19 Sep 2008 14:16:34 +0000 (22:16 +0800)] 
ocfs2/xattr.c: Fix a bug when inserting xattr.

During the process of xatt insertion, we use binary search
to find the right place and "low" is set to it. But when
there is one xattr which has the same name hash as the inserted
one, low is the wrong value. So set it to the right position.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add xattr mount option in ocfs2_show_options()
Sunil Mushran [Fri, 5 Sep 2008 18:29:14 +0000 (11:29 -0700)] 
ocfs2: Add xattr mount option in ocfs2_show_options()

Patch adds check for [no]user_xattr in ocfs2_show_options() that completes
the list of all mount options.

Signed-off-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Switch over to JBD2.
Joel Becker [Thu, 4 Sep 2008 03:03:41 +0000 (20:03 -0700)] 
ocfs2: Switch over to JBD2.

ocfs2 wants JBD2 for many reasons, not the least of which is that JBD is
limiting our maximum filesystem size.

It's a pretty trivial change.  Most functions are just renamed.  The
only functional change is moving to Jan's inode-based ordered data mode.
It's better, too.

Because JBD2 reads and writes JBD journals, this is compatible with any
existing filesystem.  It can even interact with JBD-based ocfs2 as long
as the journal is formated for JBD.

We provide a compatibility option so that paranoid people can still use
JBD for the time being.  This will go away shortly.

[ Moved call of ocfs2_begin_ordered_truncate() from ocfs2_delete_inode() to
  ocfs2_truncate_for_delete(). --Mark ]

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add the 'inode64' mount option.
Joel Becker [Thu, 4 Sep 2008 03:03:40 +0000 (20:03 -0700)] 
ocfs2: Add the 'inode64' mount option.

Now that ocfs2 limits inode numbers to 32bits, add a mount option to
disable the limit.  This parallels XFS.  64bit systems can handle the
larger inode numbers.

[ Added description of inode64 mount option in ocfs2.txt. --Mark ]

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Limit inode allocation to 32bits.
Joel Becker [Thu, 4 Sep 2008 03:03:39 +0000 (20:03 -0700)] 
ocfs2: Limit inode allocation to 32bits.

ocfs2 inode numbers are block numbers.  For any filesystem with less
than 2^32 blocks, this is not a problem.  However, when ocfs2 starts
using JDB2, it will be able to support filesystems with more than 2^32
blocks.  This would result in inode numbers higher than 2^32.

The problem is that stat(2) can't handle those numbers on 32bit
machines.  The simple solution is to have ocfs2 allocate all inodes
below that boundary.

The suballoc code is changed to honor an optional block limit.  Only the
inode suballocator sets that limit - all other allocations stay unlimited.

The biggest trick is to grow the inode suballocator beneath that limit.
There's no point in allocating block groups that are above the limit,
then rejecting their elements later on.  We want to prevent the inode
allocator from ever having block groups above the limit.  This involves
a little gyration with the local alloc code.  If the local alloc window
is above the limit, it signals the caller to try the global bitmap but
does not disable the local alloc file (which can be used for other
allocations).

[ Minor cleanup - removed an ML_NOTICE comment. --Mark ]

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Resolve deadlock in ocfs2_xattr_free_block.
Tao Ma [Fri, 29 Aug 2008 01:00:19 +0000 (09:00 +0800)] 
ocfs2: Resolve deadlock in ocfs2_xattr_free_block.

In ocfs2_xattr_free_block, we take a cluster lock on xb_alloc_inode while we
have a transaction open. This will deadlock the downconvert thread, so fix
it.

We can clean up how xattr blocks are removed while here - this patch also
moves the mechanism of releasing xattr block (including both value, xattr
tree and xattr block) into this function.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: bug-fix for journal extend in xattr.
Tao Ma [Mon, 1 Sep 2008 00:45:18 +0000 (08:45 +0800)] 
ocfs2: bug-fix for journal extend in xattr.

In ocfs2_extend_trans, when we can't extend the current
transaction, it will commit current transaction and restart
a new one. So if the previous credits we have allocated aren't
used(the block isn't dirtied before our extend), we will not
have enough credits for any future operation(it will cause jbd
complain and bug out). So check this and re-extend it.

Signed-off-by: Tao Ma <tao.ma@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Change ocfs2_get_*_extent_tree() to ocfs2_init_*_extent_tree()
Joel Becker [Fri, 22 Aug 2008 19:46:09 +0000 (12:46 -0700)] 
ocfs2: Change ocfs2_get_*_extent_tree() to ocfs2_init_*_extent_tree()

The original get/put_extent_tree() functions held a reference on
et_root_bh.  However, every single caller already has a safe reference,
making the get/put cycle irrelevant.

We change ocfs2_get_*_extent_tree() to ocfs2_init_*_extent_tree().  It
no longer gets a reference on et_root_bh.  ocfs2_put_extent_tree() is
removed.  Callers now have a simpler init+use pattern.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Comment struct ocfs2_extent_tree_operations.
Joel Becker [Fri, 22 Aug 2008 00:11:10 +0000 (17:11 -0700)] 
ocfs2: Comment struct ocfs2_extent_tree_operations.

struct ocfs2_extent_tree_operations provides methods for the different
on-disk btrees in ocfs2.  Describing what those methods do is probably a
good idea.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Make ocfs2_extent_tree the first-class representation of a tree.
Joel Becker [Thu, 21 Aug 2008 02:36:33 +0000 (19:36 -0700)] 
ocfs2: Make ocfs2_extent_tree the first-class representation of a tree.

We now have three different kinds of extent trees in ocfs2: inode data
(dinode), extended attributes (xattr_tree), and extended attribute
values (xattr_value).  There is a nice abstraction for them,
ocfs2_extent_tree, but it is hidden in alloc.c.  All the calling
functions have to pick amongst a varied API and pass in type bits and
often extraneous pointers.

A better way is to make ocfs2_extent_tree a first-class object.
Everyone converts their object to an ocfs2_extent_tree() via the
ocfs2_get_*_extent_tree() calls, then uses the ocfs2_extent_tree for all
tree calls to alloc.c.

This simplifies a lot of callers, making for readability.  It also
provides an easy way to add additional extent tree types, as they only
need to be defined in alloc.c with a ocfs2_get_<new>_extent_tree()
function.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Add an insertion check to ocfs2_extent_tree_operations.
Joel Becker [Thu, 21 Aug 2008 01:32:45 +0000 (18:32 -0700)] 
ocfs2: Add an insertion check to ocfs2_extent_tree_operations.

A couple places check an extent_tree for a valid inode.  We move that
out to add an eo_insert_check() operation.  It can be called from
ocfs2_insert_extent() and elsewhere.

We also have the wrapper calls ocfs2_et_insert_check() and
ocfs2_et_sanity_check() ignore NULL ops.  That way we don't have to
provide useless operations for xattr types.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Create specific get_extent_tree functions.
Joel Becker [Thu, 21 Aug 2008 00:44:24 +0000 (17:44 -0700)] 
ocfs2: Create specific get_extent_tree functions.

A caller knows what kind of extent tree they have.  There's no reason
they have to call ocfs2_get_extent_tree() with a NULL when they could
just as easily call a specific function to their type of extent tree.

Introduce ocfs2_dinode_get_extent_tree(),
ocfs2_xattr_tree_get_extent_tree(), and
ocfs2_xattr_value_get_extent_tree().  They only take the necessary
arguments, calling into the underlying __ocfs2_get_extent_tree() to do
the real work.

__ocfs2_get_extent_tree() is the old ocfs2_get_extent_tree(), but
without needing any switch-by-type logic.

ocfs2_get_extent_tree() is now a wrapper around the specific calls.  It
exists because a couple alloc.c functions can take et_type.  This will
go later.

Another benefit is that ocfs2_xattr_value_get_extent_tree() can take a
struct ocfs2_xattr_value_root* instead of void*.  This gives us
typechecking where we didn't have it before.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Determine an extent tree's max_leaf_clusters in an et_op.
Joel Becker [Thu, 21 Aug 2008 00:31:10 +0000 (17:31 -0700)] 
ocfs2: Determine an extent tree's max_leaf_clusters in an et_op.

Provide an optional extent_tree_operation to specify the
max_leaf_clusters of an ocfs2_extent_tree.  If not provided, the value
is 0 (unlimited).

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Use struct ocfs2_extent_tree in ocfs2_num_free_extents().
Joel Becker [Thu, 21 Aug 2008 00:09:42 +0000 (17:09 -0700)] 
ocfs2: Use struct ocfs2_extent_tree in ocfs2_num_free_extents().

ocfs2_num_free_extents() re-implements the logic of
ocfs2_get_extent_tree().  Now that ocfs2_get_extent_tree() does not
allocate, let's use it in ocfs2_num_free_extents() to simplify the code.

The inode validation code in ocfs2_num_free_extents() is not needed.
All callers are passing in pre-validated inodes.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Provide the get_root_el() method to ocfs2_extent_tree_operations.
Joel Becker [Thu, 21 Aug 2008 00:19:50 +0000 (17:19 -0700)] 
ocfs2: Provide the get_root_el() method to ocfs2_extent_tree_operations.

The root_el of an ocfs2_extent_tree needs to be calculated from
et->et_object.  Make it an operation on et->et_ops.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Make 'private' into 'object' on ocfs2_extent_tree.
Joel Becker [Wed, 20 Aug 2008 23:57:27 +0000 (16:57 -0700)] 
ocfs2: Make 'private' into 'object' on ocfs2_extent_tree.

The 'private' pointer was a way to store off xattr values, which don't
live at a set place in the bh.  But the concept of "the object
containing the extent tree" is much more generic.  For an inode it's the
struct ocfs2_dinode, for an xattr value its the value.  Let's save off
the 'object' at all times.  If NULL is passed to
ocfs2_get_extent_tree(), 'object' is set to bh->b_data;

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
16 years agoocfs2: Make ocfs2_extent_tree get/put instead of alloc.
Joel Becker [Wed, 20 Aug 2008 23:48:35 +0000 (16:48 -0700)] 
ocfs2: Make ocfs2_extent_tree get/put instead of alloc.

Rather than allocating a struct ocfs2_extent_tree, just put it on the
stack.  Fill it with ocfs2_get_extent_tree() and drop it with
ocfs2_put_extent_tree().  Now the callers don't have to ENOMEM, yet
still safely ref the root_bh.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>