GitHub/exynos8895/android_kernel_samsung_universal8895.git
15 years agon411: add missing Makefile entry
Herton Ronaldo Krzesinski [Tue, 31 Mar 2009 22:25:37 +0000 (15:25 -0700)]
n411: add missing Makefile entry

There is no entry for n411.c to be built, include one in Makefile.

Signed-off-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Cc: Jaya Kumar <jayakumar.lkml@gmail.com>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoviafb: returns 0 two too early
Roel Kluin [Tue, 31 Mar 2009 22:25:36 +0000 (15:25 -0700)]
viafb: returns 0 two too early

Otherwise this will already return 0 if iteration MAXLOOP-2 occurs in the
first loop.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: Joseph Chan <josephchan@via.com.tw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovesafb: bitwise OR has higher precedence than ?:
Roel Kluin [Tue, 31 Mar 2009 22:25:36 +0000 (15:25 -0700)]
vesafb: bitwise OR has higher precedence than ?:

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Michal Januszewski <michalj@gmail.com>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agouvesafb: bitwise OR has higher precedence than ?:
Roel Kluin [Tue, 31 Mar 2009 22:25:35 +0000 (15:25 -0700)]
uvesafb: bitwise OR has higher precedence than ?:

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Michal Januszewski <michalj@gmail.com>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoarkfb: fix misplaced parentheses
Roel Kluin [Tue, 31 Mar 2009 22:25:34 +0000 (15:25 -0700)]
arkfb: fix misplaced parentheses

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Ondrej Zajicek <santiago@crfreenet.org>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoasiliantfb: fix cmap memory leaks
Andres Salomon [Tue, 31 Mar 2009 22:25:33 +0000 (15:25 -0700)]
asiliantfb: fix cmap memory leaks

- fix cmap leak in removal path
 - fix cmap leak when register_framebuffer fails
 - check return value of fb_alloc_cmap
 - don't continue with driver setup if register_framebuffer fails

[krzysztof.h1@wp.pl: spotted missing iounmap]
[randy.dunlap@oracle.com: move data declaration before any code]
Signed-off-by: Andres Salomon <dilinger@debian.org>
Cc: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrivers/video/omap/hwa742.c: div reaches max_clk_div
Roel Kluin [Tue, 31 Mar 2009 22:25:32 +0000 (15:25 -0700)]
drivers/video/omap/hwa742.c: div reaches max_clk_div

With for(div = 0; div < max_clk_div; div++) { ... } div reaches max_clk_div.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: Joe Perches <joe@perches.com>
Acked-by: Trilok Soni <soni.trilok@gmail.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofbdev: update s1d13xxxfb to differ between revisions and production ids
Kristoffer Ericson [Tue, 31 Mar 2009 22:25:31 +0000 (15:25 -0700)]
fbdev: update s1d13xxxfb to differ between revisions and production ids

The s1d13xxx chip provides two values of identification value: the
Production id (e.g 13506/13505/13806..) and a revision number 0,1,2,3).
Together these can help us to differentiate between similiar setups.

This patch adds the proper way of grabbing both those values and save them
for future reference (in order to decide what functions a card supports,
e.g acceleration).

We also move away from the concept of all s1d13xxx = s1d13806 when we
really support alot more.

[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: simplify s1d13xxxfb_probe()]
Signed-off-by: Kristoffer Ericson <kristoffer.ericson@gmail.com
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofbdev: newport: newport_*wait() return 0 on timeout
Roel Kluin [Tue, 31 Mar 2009 22:25:30 +0000 (15:25 -0700)]
fbdev: newport: newport_*wait() return 0 on timeout

With a postfix decrement t reaches -1 on timeout which results in a
return of 0.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoamifb: check fb_alloc_cmap return value and handle failure properly
Andres Salomon [Tue, 31 Mar 2009 22:25:30 +0000 (15:25 -0700)]
amifb: check fb_alloc_cmap return value and handle failure properly

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoamba-clcd: fix cmap memory leaks
Andres Salomon [Tue, 31 Mar 2009 22:25:29 +0000 (15:25 -0700)]
amba-clcd: fix cmap memory leaks

- fix cmap leak in removal path

- fix cmap leak when register_framebuffer fails

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years ago68328fb: fix cmap memory leaks
Andres Salomon [Tue, 31 Mar 2009 22:25:28 +0000 (15:25 -0700)]
68328fb: fix cmap memory leaks

- fix cmap leak in removal path

- fix cmap leak when register_framebuffer fails

- check return value of fb_alloc_cmap

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotgafb: fix cmap memory leak
Andres Salomon [Tue, 31 Mar 2009 22:25:28 +0000 (15:25 -0700)]
tgafb: fix cmap memory leak

Fix cmap leak when register_framebuffer fails.

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosunxvr500: fix cmap memory leaks
Andres Salomon [Tue, 31 Mar 2009 22:25:27 +0000 (15:25 -0700)]
sunxvr500: fix cmap memory leaks

- fix cmap leak in removal path
 - fix cmap leak when register_framebuffer fails

Signed-off-by: Andres Salomon <dilinger@debian.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovalkyriefb: check fb_alloc_cmap return value and handle failure properly
Andres Salomon [Tue, 31 Mar 2009 22:25:26 +0000 (15:25 -0700)]
valkyriefb: check fb_alloc_cmap return value and handle failure properly

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agostifb: check fb_alloc_cmap return value and handle failure properly
Andres Salomon [Tue, 31 Mar 2009 22:25:26 +0000 (15:25 -0700)]
stifb: check fb_alloc_cmap return value and handle failure properly

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosstfb: check fb_alloc_cmap return value and handle failure properly
Andres Salomon [Tue, 31 Mar 2009 22:25:25 +0000 (15:25 -0700)]
sstfb: check fb_alloc_cmap return value and handle failure properly

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosm501fb: check fb_alloc_cmap return value and handle failure properly
Andres Salomon [Tue, 31 Mar 2009 22:25:24 +0000 (15:25 -0700)]
sm501fb: check fb_alloc_cmap return value and handle failure properly

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@poczta.fm>
Cc: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoskeletonfb: check fb_alloc_cmap return value and handle failure properly
Andres Salomon [Tue, 31 Mar 2009 22:25:23 +0000 (15:25 -0700)]
skeletonfb: check fb_alloc_cmap return value and handle failure properly

Bad example code, no cookie!

Signed-off-by: Andres Salomon <dilinger@debian.org>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovfb: fix memory leaks in removal path
Andres Salomon [Tue, 31 Mar 2009 22:25:22 +0000 (15:25 -0700)]
vfb: fix memory leaks in removal path

We were leaking the cmap memory.

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotridentfb: fix memory leaks in removal path
Andres Salomon [Tue, 31 Mar 2009 22:25:22 +0000 (15:25 -0700)]
tridentfb: fix memory leaks in removal path

We were leaking the cmap memory.

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agotdfxfb: fix memory leaks in removal path
Andres Salomon [Tue, 31 Mar 2009 22:25:21 +0000 (15:25 -0700)]
tdfxfb: fix memory leaks in removal path

We were leaking the cmap memory.

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrivers/video/sgivwfb.c: fix memory leaks in removal path
Andres Salomon [Tue, 31 Mar 2009 22:25:20 +0000 (15:25 -0700)]
drivers/video/sgivwfb.c: fix memory leaks in removal path

We were leaking both the cmap memory and the info struct memory.

Signed-off-by: Andres Salomon <dilinger@debian.org>
Acked-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofbdev: uninline lock_fb_info()
Andrew Morton [Tue, 31 Mar 2009 22:25:19 +0000 (15:25 -0700)]
fbdev: uninline lock_fb_info()

Before:

   text    data     bss     dec     hex filename
   3648    2910      32    6590    19be drivers/video/backlight/backlight.o
   3226    2812      32    6070    17b6 drivers/video/backlight/lcd.o
  30990   16688    8480   56158    db5e drivers/video/console/fbcon.o
  15488    8400      24   23912    5d68 drivers/video/fbmem.o

After:

   text    data     bss     dec     hex filename
   3537    2870      32    6439    1927 drivers/video/backlight/backlight.o
   3131    2772      32    5935    172f drivers/video/backlight/lcd.o
  30876   16648    8480   56004    dac4 drivers/video/console/fbcon.o
  15506    8400      24   23930    5d7a drivers/video/fbmem.o

Cc: Andrea Righi <righi.andrea@gmail.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofbmem: fix fb_info->lock and mm->mmap_sem circular locking dependency
Andrea Righi [Tue, 31 Mar 2009 22:25:18 +0000 (15:25 -0700)]
fbmem: fix fb_info->lock and mm->mmap_sem circular locking dependency

Fix a circular locking dependency in the frame buffer console driver
pushing down the mutex fb_info->lock.

Circular locking dependecies occur calling the blocking
fb_notifier_call_chain() with fb_info->lock held.  Notifier callbacks can
try to acquire mm->mmap_sem, while fb_mmap() acquires the locks in the
reverse order mm->mmap_sem => fb_info->lock.

Tested-by: Andrey Borzenkov <arvidjaar@mail.ru>
Signed-off-by: Andrea Righi <righi.andrea@gmail.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: fix interlaced modes
Krzysztof Helt [Tue, 31 Mar 2009 22:25:17 +0000 (15:25 -0700)]
cirrusfb: fix interlaced modes

Fix calculations of timings for interlaced modes.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: fix threshold register mask for Laguna chips
Krzysztof Helt [Tue, 31 Mar 2009 22:25:17 +0000 (15:25 -0700)]
cirrusfb: fix threshold register mask for Laguna chips

Fix threshold register mask for Laguna chips otherwise some 8bpp modes are
garbled after selecting a 24bpp mode.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: GD5434 (aka SD64) support fixed
Krzysztof Helt [Tue, 31 Mar 2009 22:25:16 +0000 (15:25 -0700)]
cirrusfb: GD5434 (aka SD64) support fixed

Fix handling of the Cirrus Logic GD5434 chip.  Distinguish this chip from the
GD5430.  It allows detecting memory size for both models correctly.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: set MCLK in one place
Krzysztof Helt [Tue, 31 Mar 2009 22:25:15 +0000 (15:25 -0700)]
cirrusfb: set MCLK in one place

A memory clock (MCLK) is set at various places.  Move the setting into one
place.

Set the MCLK only for Zorro cards as the x86 cards should be initialized by
BIOS.

Improve handling of the GD5434 (SD64).

Kill one annoying debug output "virtual offset: ...".

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: add accelerator constant
Krzysztof Helt [Tue, 31 Mar 2009 22:25:15 +0000 (15:25 -0700)]
cirrusfb: add accelerator constant

Add an accelerator constant so almost all Cirrus are recognized as
accelerators by the fbset command.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: fix clock doubling
Krzysztof Helt [Tue, 31 Mar 2009 22:25:14 +0000 (15:25 -0700)]
cirrusfb: fix clock doubling

Cirrus' Alpine and Picasso4 chips uses DAC clock doubling to achieve full
range of pixclock frequencies.

[akpm@linux-foundation.org: fix spelling, use usual comment layout]
Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: use 24bpp instead of 32bpp
Krzysztof Helt [Tue, 31 Mar 2009 22:25:13 +0000 (15:25 -0700)]
cirrusfb: use 24bpp instead of 32bpp

The 32bpp is supported only on the latest Cirrus Logic chips.  Use the
24bpp which is supported at least since Alpine chips (GD543x).

Change 32bpp mode setting to 24bpp mode.  Change acceleration as well.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: GD5446 fixes
Krzysztof Helt [Tue, 31 Mar 2009 22:25:13 +0000 (15:25 -0700)]
cirrusfb: GD5446 fixes

Various fixes to make Cirrus GD5446 chip work.

Another Cirrus chip works with the cirrusfb.  The gd5446 seems very
similar to Alpine chips.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: fix error paths in cirrusfb_xxx_register()
Krzysztof Helt [Tue, 31 Mar 2009 22:25:12 +0000 (15:25 -0700)]
cirrusfb: fix error paths in cirrusfb_xxx_register()

Balance iomap and iounmap and alloc and free calls in case of error druing
device register (probing).

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: add imageblit function
Krzysztof Helt [Tue, 31 Mar 2009 22:25:11 +0000 (15:25 -0700)]
cirrusfb: add imageblit function

Add hardware color expansion (imageblit) function.  It roughly doubles
scrolling speed of my Alpine card (GD5430).

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: acceleration improvements
Krzysztof Helt [Tue, 31 Mar 2009 22:25:11 +0000 (15:25 -0700)]
cirrusfb: acceleration improvements

- Fix color expansion for 16bpp and 32bpp modes in the
  cirrusfb_RectFill().

- Make a function with a common blitter code (cirrusfb_set_blitter).

- Add fb_sync function to allow a higher layer synchronize with the
  blitter.

- Kill one redundant blitter reset.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: various Laguna fixes
Krzysztof Helt [Tue, 31 Mar 2009 22:25:10 +0000 (15:25 -0700)]
cirrusfb: various Laguna fixes

- The Laguna GD5465 (AGP) has one register more than non-AGP chips.
  Recognize the AGP version and write a tile control register only on the
  AGP version.  Tested only on an AGP card.

- Bump up RAMDAC frequencies after X11 code.  This allow to drive a flat
  panel resolution 1680x1050 at 16bpp from the 4MB card.

- Fix screen start address overflow bits on Laguna cards (CRT1D
  register).

- Fix exit path in the cirrusfb_pci_register() in case of error.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: check_var improvements
Krzysztof Helt [Tue, 31 Mar 2009 22:25:09 +0000 (15:25 -0700)]
cirrusfb: check_var improvements

Break cirrusfb_decode_var() function into two parts:
cirrusfb_check_pixclock() which can be called from the
cirrusfb_check_var() aand merge rest into the cirrusfb_set_par_foo().
This allows rejecting modes with too high pixclock before before any
change to hardware state (and a console is messed up).

Also, fix RGB field's lengths for 8bpp modes to correct ones so X11 works
with fbdev driver with cirrusfb.

Kill some redundant function calls or register loads.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: Laguna chipset 8bpp fix
Krzysztof Helt [Tue, 31 Mar 2009 22:25:08 +0000 (15:25 -0700)]
cirrusfb: Laguna chipset 8bpp fix

Fix 8bpp mode by adding handling of the Laguna chipsets to various places
and stop trashing a HDR register which probably does not exist on the
Laguna.

Fix compilation warnings about uninitialized variables also.

Finally, all 8bpp, 16bpp and 32bpp modes work on the Laguna chipset.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: various improvements
Krzysztof Helt [Tue, 31 Mar 2009 22:25:08 +0000 (15:25 -0700)]
cirrusfb: various improvements

Various improvements to the code:
- kill a structure with only one
  field: multiplexing and use the
  field directly
- move the cirrusfb_ops structure
  down the file to kill forward
  declarations
- move cirrusfb_init() to kill
  forward declaration
- kill register loads done already
  in the init_vgachip()
- kill assigments done by higher
  layer in the cirrusfb_pan_display()
- do not overwrite line pitch bit in
  the CL_CRT1D register
- kill btype variables if they were
  used only once or twice
- add cpu_relax() in the busy waiting
  loop

The fix to the CL_CRT1D register handling makess the 1024x768 32bpp mode
work.  Previously, only lower resolution modes have worked with 32bpp.

Signed-off-by: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: use 5-6-5 RGB for 16bpp mode
Krzysztof Helt [Tue, 31 Mar 2009 22:25:07 +0000 (15:25 -0700)]
cirrusfb: use 5-6-5 RGB for 16bpp mode

Use the 5-6-5 RGB mode instead of the 5-5-5 mode at 16bpp depth.

It fixes colors in the 16bpp modes on Cirrus Laguna chips.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Geert Uytterhoeven <geert.uytterhoeven@gmail.com>
Cc: Arthur Marsh <arthur.marsh@internode.on.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: do not calculate line length twice
Krzysztof Helt [Tue, 31 Mar 2009 22:25:06 +0000 (15:25 -0700)]
cirrusfb: do not calculate line length twice

A line length is calculated twice: first in the cirrusfb_decode_var() then
in the cirrusfb_set_par_foo().

Use the first calculated value.  A nice side effect is that 32bpp mode
works now.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Geert Uytterhoeven <geert.uytterhoeven@gmail.com>
Cc: Arthur Marsh <arthur.marsh@internode.on.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: add mmio registers for Laguna chipsets
Krzysztof Helt [Tue, 31 Mar 2009 22:25:05 +0000 (15:25 -0700)]
cirrusfb: add mmio registers for Laguna chipsets

The Laguna chipsets use special registers which are available through the
mmio area.  The cirrusfb driver does not use memory mapped registers for
the PCI cards.

Add the memory mapped area for Laguna chipsets and add basic usage of the
special Laguna registers after SVGALIB code.

This gives readable console at 16bpp on the GD-5465 (Laguna AGP).  The
8bpp and 32bpp depths are still broken.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: add Laguna additional overflow register
Krzysztof Helt [Tue, 31 Mar 2009 22:25:04 +0000 (15:25 -0700)]
cirrusfb: add Laguna additional overflow register

Add additional overflow register setting for Laguna chips.

Also, simplify some code in the cirrusfb_pan_display() and
cirrusfb_blank().

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: fix Laguna chipset memory detection and clock setting
Krzysztof Helt [Tue, 31 Mar 2009 22:25:04 +0000 (15:25 -0700)]
cirrusfb: fix Laguna chipset memory detection and clock setting

Fix memory detection and clock setting for Cirrus Laguna chipsets
(GD5464/GD5465).  The changes are done after the Xorg code.

The driver still does not display anything on the GD5465 but it switches
resolutions correctly at least.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocirrusfb: convert printks to dev_foo
Krzysztof Helt [Tue, 31 Mar 2009 22:25:03 +0000 (15:25 -0700)]
cirrusfb: convert printks to dev_foo

Convert all printks to dev_dbg, dev_info or dev_err.  Kill some excessive
debug information and code in the process.

[akpm@linux-foundation.org: printk fixes]
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Krzysztof Helt <krzysztof.h1@poczta.fm>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoatyfb: fix header file trailing whitespace
Randy Dunlap [Tue, 31 Mar 2009 22:25:02 +0000 (15:25 -0700)]
atyfb: fix header file trailing whitespace

Fix trailing whitespace because quilt complained about it.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agodrivers/video/uvesafb.c: don't use gfp_any()
Andrew Morton [Tue, 31 Mar 2009 22:25:01 +0000 (15:25 -0700)]
drivers/video/uvesafb.c: don't use gfp_any()

GFP_KERNEL is legal here - we don't need to use gfp_any().

Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc: test before subtraction on unsigned
Roel Kluin [Tue, 31 Mar 2009 22:25:00 +0000 (15:25 -0700)]
rtc: test before subtraction on unsigned

new_alarm is unsigned so test before the subtraction.

[akpm@linux-foundation.org: time-wrapping fix]
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-v3020: coding style cleanup
Mike Rapoport [Tue, 31 Mar 2009 22:24:59 +0000 (15:24 -0700)]
rtc-v3020: coding style cleanup

Signed-off-by: Mike Rapoport <mike@compulab.co.il>
Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-wm8350: retries will reach -1
Roel Kluin [Tue, 31 Mar 2009 22:24:59 +0000 (15:24 -0700)]
rtc-wm8350: retries will reach -1

With a postfix decrement retries will reach -1 rather than 0, so the
warning and error-out will not occur.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc: add EPSON RX8025 support to DS1307 RTC driver
Matthias Fuchs [Tue, 31 Mar 2009 22:24:58 +0000 (15:24 -0700)]
rtc: add EPSON RX8025 support to DS1307 RTC driver

Add support for the EPSON RX8025 RTC.  The date/time registers of this
chip are compatible with the DS1307.

Signed-off-by: Matthias Fuchs <matthias.fuchs@esd-electronics.com>
Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-ds1307: true SMBus compatibility
Ed Swierk [Tue, 31 Mar 2009 22:24:56 +0000 (15:24 -0700)]
rtc-ds1307: true SMBus compatibility

Allow the rtc-ds1307 driver to work with SMBus controllers like nforce2
that do not support i2c block transfers.

Signed-off-by: Ed Swierk <eswierk@aristanetworks.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: BARRE Sebastien <sbarre@sdelcc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: rename p pointer to rtc
dann frazier [Tue, 31 Mar 2009 22:24:55 +0000 (15:24 -0700)]
rtc-parisc: rename p pointer to rtc

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: remove a couple unnecessary variables
dann frazier [Tue, 31 Mar 2009 22:24:55 +0000 (15:24 -0700)]
rtc-parisc: remove a couple unnecessary variables

Signed-off-by: dann frazier <dannf@hp.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: remove unnecessary ret variable
dann frazier [Tue, 31 Mar 2009 22:24:54 +0000 (15:24 -0700)]
rtc-parisc: remove unnecessary ret variable

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: declare rtc_parisc_dev as static
dann frazier [Tue, 31 Mar 2009 22:24:53 +0000 (15:24 -0700)]
rtc-parisc: declare rtc_parisc_dev as static

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: use platform_driver_probe
dann frazier [Tue, 31 Mar 2009 22:24:52 +0000 (15:24 -0700)]
rtc-parisc: use platform_driver_probe

This isn't a hotpluggable device, so call platform_driver_probe
directly in parisc_rtc_init

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: use rtc_valid_tm() in parisc_get_time
dann frazier [Tue, 31 Mar 2009 22:24:52 +0000 (15:24 -0700)]
rtc-parisc: use rtc_valid_tm() in parisc_get_time

Use the return value of rtc_valid_tm() instead of just returning 0.

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: remove struct parisc_rtc
dann frazier [Tue, 31 Mar 2009 22:24:51 +0000 (15:24 -0700)]
rtc-parisc: remove struct parisc_rtc

parisc_rtc now only includes an rtc_device pointer, so let's
just use the rtc_device type directly.

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: remove redundant locking
dann frazier [Tue, 31 Mar 2009 22:24:50 +0000 (15:24 -0700)]
rtc-parisc: remove redundant locking

The RTC subsystem proides ops locking, no need to implement our own

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc-parisc: add a missing include for linux/rtc.h
dann frazier [Tue, 31 Mar 2009 22:24:49 +0000 (15:24 -0700)]
rtc-parisc: add a missing include for linux/rtc.h

Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc: add platform driver for EFI
dann frazier [Tue, 31 Mar 2009 22:24:48 +0000 (15:24 -0700)]
rtc: add platform driver for EFI

Munge Stephane Eranian's efirtc.c code into an rtc platform driver

[akpm@linux-foundation.org: use is_leap_year()]
Signed-off-by: dann frazier <dannf@hp.com>
Cc: Alessandro Zummo <alessandro.zummo@towertech.it>
Cc: stephane eranian <eranian@googlemail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc: convert LEAP_YEAR into an inline
Andrew Morton [Tue, 31 Mar 2009 22:24:48 +0000 (15:24 -0700)]
rtc: convert LEAP_YEAR into an inline

- the LEAP_YEAR macro is buggy - it references its arg multiple times.
  Fix this by turning it into a C function.

- give it a more approriate name

- Move it to rtc.h so that other .c files can use it, instead of copying it.

Cc: dann frazier <dannf@hp.com>
Acked-by: Alessandro Zummo <alessandro.zummo@towertech.it>
Cc: stephane eranian <eranian@googlemail.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agortc: convert wm8350 use new alarm and update operations
Mark Brown [Tue, 31 Mar 2009 22:24:46 +0000 (15:24 -0700)]
rtc: convert wm8350 use new alarm and update operations

These are the only two ioctls so the ioctl() function is also removed.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Cc: Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoautofs4: fix kernel includes
Ian Kent [Tue, 31 Mar 2009 22:24:45 +0000 (15:24 -0700)]
autofs4: fix kernel includes

autofs_dev-ioctl.h is included by both the kernel module and user space tools
and it includes two kernel header files.  Compiles work if the kernel headers
are installed but fail otherwise.

Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoautofs4: fix lookup deadlock
Ian Kent [Tue, 31 Mar 2009 22:24:45 +0000 (15:24 -0700)]
autofs4: fix lookup deadlock

A deadlock can occur when user space uses a signal (autofs version 4 uses
SIGCHLD for this) to effect expire completion.

The order of events is:

Expire process completes, but before being able to send SIGCHLD to it's parent
...

Another process walks onto a different mount point and drops the directory
inode semaphore prior to sending the request to the daemon as it must ...

A third process does an lstat on on the expired mount point causing it to wait
on expire completion (unfortunately) holding the directory semaphore.

The mount request then arrives at the daemon which does an lstat and,
deadlock.

For some time I was concerned about releasing the directory semaphore around
the expire wait in autofs4_lookup as well as for the mount call back.  I
finally realized that the last round of changes in this function made the
expiring dentry and the lookup dentry separate and distinct so the check and
possible wait can be done anywhere prior to the mount call back.  This patch
moves the check to just before the mount call back and inside the directory
inode mutex release.

Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoautofs4: cleanup expire code duplication
Ian Kent [Tue, 31 Mar 2009 22:24:43 +0000 (15:24 -0700)]
autofs4: cleanup expire code duplication

A significant portion of the autofs_dev_ioctl_expire() and
autofs4_expire_multi() functions is duplicated code.  This patch cleans that
up.

Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoecryptfs: use kzfree()
Johannes Weiner [Tue, 31 Mar 2009 22:24:42 +0000 (15:24 -0700)]
ecryptfs: use kzfree()

Use kzfree() instead of memset() + kfree().

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Tyler Hicks <tyhicks@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agopowerpc/fsl_soc: isolate legacy fsl_spi support to mpc832x_rdb boards
Anton Vorontsov [Tue, 31 Mar 2009 22:24:40 +0000 (15:24 -0700)]
powerpc/fsl_soc: isolate legacy fsl_spi support to mpc832x_rdb boards

The advantages of this:
- Don't encourage legacy support;
- Less external symbols, less code to compile-in for !MPC832x_RDB
  platforms.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Kumar Gala <galak@gate.crashing.org>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agopowerpc/83xx: add mmc-spi support via the device tree for MPC8323E-RDB
Anton Vorontsov [Tue, 31 Mar 2009 22:24:39 +0000 (15:24 -0700)]
powerpc/83xx: add mmc-spi support via the device tree for MPC8323E-RDB

- Add gpio-controller node to manage QE GPIO Bank D;
- Add mmc-spi node;
- Modify board file so that it won't use legacy SPI support with the new
  device trees.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Kumar Gala <galak@gate.crashing.org>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agopowerpc: add mmc-spi-slot bindings
Anton Vorontsov [Tue, 31 Mar 2009 22:24:38 +0000 (15:24 -0700)]
powerpc: add mmc-spi-slot bindings

The bindings describes a case where MMC/SD/SDIO slot directly connected to
a SPI bus.  Such setups are widely used on embedded PowerPC boards.

The patch also adds the mmc-spi-slot entry to the OpenFirmware modalias
table.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Kumar Gala <galak@gate.crashing.org>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agospi_mpc83xx: add OF platform driver bindings
Anton Vorontsov [Tue, 31 Mar 2009 22:24:37 +0000 (15:24 -0700)]
spi_mpc83xx: add OF platform driver bindings

Implement full support for OF SPI bindings.  Now the driver can manage its
own chip selects without any help from the board files and/or fsl_soc
constructors.

The "legacy" code is well isolated and could be removed as time goes by.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Kumar Gala <galak@gate.crashing.org>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agospi_mpc83xx: rework chip selects handling
Anton Vorontsov [Tue, 31 Mar 2009 22:24:36 +0000 (15:24 -0700)]
spi_mpc83xx: rework chip selects handling

The main purpose of this patch is to pass 'struct spi_device' to the chip
select handling routines.  This is needed so that we could implement
full-fledged OpenFirmware support for this driver.

While at it, also:
- Replace two {de,activate}_cs routines by single cs_contol().
- Don't duplicate platform data callbacks in mpc83xx_spi struct.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Kumar Gala <galak@gate.crashing.org>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agospi_mpc83xx: fix sparse warnings
Anton Vorontsov [Tue, 31 Mar 2009 22:24:35 +0000 (15:24 -0700)]
spi_mpc83xx: fix sparse warnings

The patch fixes following sparse warnings:

  CHECK   spi_mpc83xx.c
spi_mpc83xx.c:145:1: warning: symbol 'mpc83xx_spi_rx_buf_u8' was not declared. Should it be static?
spi_mpc83xx.c:146:1: warning: symbol 'mpc83xx_spi_rx_buf_u16' was not declared. Should it be static?
spi_mpc83xx.c:147:1: warning: symbol 'mpc83xx_spi_rx_buf_u32' was not declared. Should it be static?
spi_mpc83xx.c:148:1: warning: symbol 'mpc83xx_spi_tx_buf_u8' was not declared. Should it be static?
spi_mpc83xx.c:149:1: warning: symbol 'mpc83xx_spi_tx_buf_u16' was not declared. Should it be static?
spi_mpc83xx.c:150:1: warning: symbol 'mpc83xx_spi_tx_buf_u32' was not declared. Should it be static?
spi_mpc83xx.c:175:32: warning: incorrect type in initializer (different address spaces)
spi_mpc83xx.c:175:32:    expected void *tmp_ptr
spi_mpc83xx.c:175:32:    got unsigned int [noderef] <asn:2>*<noident>
spi_mpc83xx.c:183:26: warning: incorrect type in argument 1 (different address spaces)
spi_mpc83xx.c:183:26:    expected unsigned int [noderef] [usertype] <asn:2>*reg
spi_mpc83xx.c:183:26:    got void *tmp_ptr
spi_mpc83xx.c:184:26: warning: incorrect type in argument 1 (different address spaces)
spi_mpc83xx.c:184:26:    expected unsigned int [noderef] [usertype] <asn:2>*reg
spi_mpc83xx.c:184:26:    got void *tmp_ptr
spi_mpc83xx.c:287:31: warning: incorrect type in initializer (different address spaces)
spi_mpc83xx.c:287:31:    expected void *tmp_ptr
spi_mpc83xx.c:287:31:    got unsigned int [noderef] <asn:2>*<noident>
spi_mpc83xx.c:295:25: warning: incorrect type in argument 1 (different address spaces)
spi_mpc83xx.c:295:25:    expected unsigned int [noderef] [usertype] <asn:2>*reg
spi_mpc83xx.c:295:25:    got void *tmp_ptr
spi_mpc83xx.c:296:25: warning: incorrect type in argument 1 (different address spaces)
spi_mpc83xx.c:296:25:    expected unsigned int [noderef] [usertype] <asn:2>*reg
spi_mpc83xx.c:296:25:    got void *tmp_ptr
spi_mpc83xx.c:486:13: warning: symbol 'mpc83xx_spi_irq' was not declared. Should it be static?

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Kumar Gala <galak@gate.crashing.org>
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoramfs: add support for "mode=" mount option
Wu Fengguang [Tue, 31 Mar 2009 22:24:34 +0000 (15:24 -0700)]
ramfs: add support for "mode=" mount option

Addresses http://bugzilla.kernel.org/show_bug.cgi?id=12843

"I use ramfs instead of tmpfs for /tmp because I don't use swap on my
laptop.  Some apps need 1777 mode for /tmp directory, but ramfs does not
support 'mode=' mount option."

Reported-by: Avan Anishchuk <matimatik@gmail.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agolis3: SPI transport layer
Daniel Mack [Tue, 31 Mar 2009 22:24:33 +0000 (15:24 -0700)]
lis3: SPI transport layer

Make use of the new abstraction layer and add a new transport layer for
spi.  Works fine on a PXA based board.

Signed-off-by: Daniel Mack <daniel@caiaq.de>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Eric Piel <eric.piel@tremplin-utc.net>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agolis3: solve dependency between core and ACPI
Daniel Mack [Tue, 31 Mar 2009 22:24:32 +0000 (15:24 -0700)]
lis3: solve dependency between core and ACPI

This solves the dependency between lis3lv02d.[ch] and ACPI specific
methods.  It introduces a ->bus_priv pointer to the device struct which is
casted to 'struct acpi_device' in the ACIP layer.  Changed hp_accel.c
accordingly.

Signed-off-by: Daniel Mack <daniel@caiaq.de>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Eric Piel <eric.piel@tremplin-utc.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agolis3: reorder functions to make forward decl obsolete
Daniel Mack [Tue, 31 Mar 2009 22:24:31 +0000 (15:24 -0700)]
lis3: reorder functions to make forward decl obsolete

Move lis3lv02d_init_device() down so that the forward declaration of
lis3lv02d_add_fs() becomes unnecessary.

Signed-off-by: Daniel Mack <daniel@caiaq.de>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Eric Piel <eric.piel@tremplin-utc.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohp_accel: axis conversion for hp compaq 8710w
Luca Cappa [Tue, 31 Mar 2009 22:24:31 +0000 (15:24 -0700)]
hp_accel: axis conversion for hp compaq 8710w

I have a laptop HP Compaq 8710W, I compiled into my kernel the LIS3LV02DL
and HP_ACCEL module drivers.  While loading it cannot recognize the laptop
model, so i am sending the necessary information to update the database of
axis orientations.

>When the laptop is horizontal the position reported is about 0 for X and Y
>and a positive value for Z
Yes, it is about 0,0,1000, the actual reading says: (-17,-26,1018);

> If the left side is elevated, X increases (becomes positive)
Yes, X goes toward to positive 1000.

>If the front side (where the touchpad is) is elevated, Y decreases (becomes negative)
No, Y goes toward to positive 1000.

>If the laptop is put upside-down, Z becomes negative
Yes, the laptop on a table Z gives 1000, and if upsidedown the Z reads
-1000.

So in few words the Y axis is inverted.

Cc: Eric Piel <eric.piel@tremplin-utc.net>
Signed-off-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohp_accel: add two more axis information
Pavel Machek [Tue, 31 Mar 2009 22:24:29 +0000 (15:24 -0700)]
hp_accel: add two more axis information

Add two more laptops to whitelist.

Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Pavel Machek <pavel@ucw.cz>
Cc: Daniel Mack <daniel@caiaq.de>
Cc: Eric Piel <eric.piel@tremplin-utc.net>
Cc: Vladimir Botka <vbotka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohwmon: Add LTC4215 driver
Ira Snyder [Tue, 31 Mar 2009 22:24:29 +0000 (15:24 -0700)]
hwmon: Add LTC4215 driver

Add Linux support for the Linear Technology LTC4215 Hot Swap controller
I2C monitoring interface.

I have tested the driver with my board, and it appears to work fine.  With
the power supplies disabled, it reads 11.93V input, 1.93V output, no
current and no power.  With the supplies enabled, it reads 11.93V input,
11.98V output, no current, no power.  I'm not drawing any current at the
moment, so this is reasonable.  The value in the sense register never
reads anything except 0, so I expect to get zero from the current and
power calculations.

I didn't attempt to support changing any of the chip's settings or
enabling the FET.  I'm not sure even how to do that and still fit within
the hwmon framework.  :)

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Cc: Jean Delvare <khali@linux-fr.org>
Cc: "Mark M. Hoffman" <mhoffman@lightlink.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohwmon: LM95241 driver
Davide Rizzo [Tue, 31 Mar 2009 22:24:27 +0000 (15:24 -0700)]
hwmon: LM95241 driver

An hwmon driver for the National Semiconductor LM95241 triple temperature
sensors chip

Signed-off-by: Davide Rizzo <elpa-rizzo@gmail.com>
Cc: Jean Delvare <khali@linux-fr.org>
Cc: "Mark M. Hoffman" <mhoffman@lightlink.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohp_accel: adev is poor name of exported symbol
Pavel Machek [Tue, 31 Mar 2009 22:24:26 +0000 (15:24 -0700)]
hp_accel: adev is poor name of exported symbol

As Andrew noted, adev is pretty poor name for symbol being exported.
Rename it to lis3.

Signed-off-by: Pavel Machek <pavel@ucw.cz>
Cc: Eric Piel <eric.piel@tremplin-utc.net>
Cc: Vladimir Botka <vbotka@suse.cz>
Cc: <Quoc.Pham@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agohp_accel: small documentation updates
Pavel Machek [Tue, 31 Mar 2009 22:24:25 +0000 (15:24 -0700)]
hp_accel: small documentation updates

Fix english in Documentation, add "how to test" description.

Signed-off-by: Pavel Machek <pavel@suse.cz>
Cc: Eric Piel <eric.piel@tremplin-utc.net>
Cc: Vladimir Botka <vbotka@suse.cz>
Cc: <Quoc.Pham@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: make tty use keyed wakeups
Davide Libenzi [Tue, 31 Mar 2009 22:24:24 +0000 (15:24 -0700)]
epoll keyed wakeups: make tty use keyed wakeups

Introduce keyed event wakeups inside the TTY code.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: make eventfd use keyed wakeups
Davide Libenzi [Tue, 31 Mar 2009 22:24:23 +0000 (15:24 -0700)]
epoll keyed wakeups: make eventfd use keyed wakeups

Introduce keyed event wakeups inside the eventfd code.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: teach epoll about hints coming with the wakeup key
Davide Libenzi [Tue, 31 Mar 2009 22:24:22 +0000 (15:24 -0700)]
epoll keyed wakeups: teach epoll about hints coming with the wakeup key

Use the events hint now sent by some devices, to avoid unnecessary wakeups
for events that are of no interest for the caller.  This code handles both
devices that are sending keyed events, and the ones that are not (and
event the ones that sometimes send events, and sometimes don't).

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: make sockets use keyed wakeups
Davide Libenzi [Tue, 31 Mar 2009 22:24:21 +0000 (15:24 -0700)]
epoll keyed wakeups: make sockets use keyed wakeups

Add support for event-aware wakeups to the sockets code.  Events are
delivered to the wakeup target, so that epoll can avoid spurious wakeups
for non-interesting events.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: introduce new *_poll() wakeup macros
Davide Libenzi [Tue, 31 Mar 2009 22:24:20 +0000 (15:24 -0700)]
epoll keyed wakeups: introduce new *_poll() wakeup macros

Introduce new wakeup macros that allow passing an event mask to the wakeup
targets.  They exactly mimic their non-_poll() counterpart, with the added
event mask passing capability.  I did add only the ones currently
requested, avoiding the _nr() and _all() for the moment.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll keyed wakeups: add __wake_up_locked_key() and __wake_up_sync_key()
Davide Libenzi [Tue, 31 Mar 2009 22:24:20 +0000 (15:24 -0700)]
epoll keyed wakeups: add __wake_up_locked_key() and __wake_up_sync_key()

This patchset introduces wakeup hints for some of the most popular (from
epoll POV) devices, so that epoll code can avoid spurious wakeups on its
waiters.

The problem with epoll is that the callback-based wakeups do not, ATM,
carry any information about the events the wakeup is related to.  So the
only choice epoll has (not being able to call f_op->poll() from inside the
callback), is to add the file* to a ready-list and resolve the real events
later on, at epoll_wait() (or its own f_op->poll()) time.  This can cause
spurious wakeups, since the wake_up() itself might be for an event the
caller is not interested into.

The rate of these spurious wakeup can be pretty high in case of many
network sockets being monitored.

By allowing devices to report the events the wakeups refer to (at least
the two major classes - POLLIN/POLLOUT), we are able to spare useless
wakeups by proper handling inside the epoll's poll callback.

Epoll will have in any case to call f_op->poll() on the file* later on,
since the change to be done in order to have the full event set sent via
wakeup, is too invasive for the way our f_op->poll() system works (the
full event set is calculated inside the poll function - there are too many
of them to even start thinking the change - also poll/select would need
change too).

Epoll is changed in a way that both devices which send event hints, and
the ones that don't, are correctly handled.  The former will gain some
efficiency though.

As a general rule for devices, would be to add an event mask by using
key-aware wakeup macros, when making up poll wait queues.  I tested it
(together with the epoll's poll fix patch Andrew has in -mm) and wakeups
for the supported devices are correctly filtered.

Test program available here:

http://www.xmailserver.org/epoll_test.c

This patch:

Nothing revolutionary here.  Just using the available "key" that our
wakeup core already support.  The __wake_up_locked_key() was no brainer,
since both __wake_up_locked() and __wake_up_locked_key() are thin wrappers
around __wake_up_common().

The __wake_up_sync() function had a body, so the choice was between
borrowing the body for __wake_up_sync_key() and calling it from
__wake_up_sync(), or make an inline and calling it from both.  I chose the
former since in most archs it all resolves to "mov $0, REG; jmp ADDR".

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: William Lee Irwin III <wli@movementarian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoeventfd: improve support for semaphore-like behavior
Davide Libenzi [Tue, 31 Mar 2009 22:24:18 +0000 (15:24 -0700)]
eventfd: improve support for semaphore-like behavior

People started using eventfd in a semaphore-like way where before they
were using pipes.

That is, counter-based resource access.  Where a "wait()" returns
immediately by decrementing the counter by one, if counter is greater than
zero.  Otherwise will wait.  And where a "post(count)" will add count to
the counter releasing the appropriate amount of waiters.  If eventfd the
"post" (write) part is fine, while the "wait" (read) does not dequeue 1,
but the whole counter value.

The problem with eventfd is that a read() on the fd returns and wipes the
whole counter, making the use of it as semaphore a little bit more
cumbersome.  You can do a read() followed by a write() of COUNTER-1, but
IMO it's pretty easy and cheap to make this work w/out extra steps.  This
patch introduces a new eventfd flag that tells eventfd to only dequeue 1
from the counter, allowing simple read/write to make it behave like a
semaphore.  Simple test here:

http://www.xmailserver.org/eventfd-sem.c

To be back-compatible with earlier kernels, userspace applications should
probe for the availability of this feature via

#ifdef EFD_SEMAPHORE
fd = eventfd2 (CNT, EFD_SEMAPHORE);
if (fd == -1 && errno == EINVAL)
<fallback>
#else
<fallback>
#endif

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: <linux-api@vger.kernel.org>
Tested-by: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: use real type instead of void *
Tony Battersby [Tue, 31 Mar 2009 22:24:16 +0000 (15:24 -0700)]
epoll: use real type instead of void *

eventpoll.c uses void * in one place for no obvious reason; change it to
use the real type instead.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: clean up ep_modify
Tony Battersby [Tue, 31 Mar 2009 22:24:15 +0000 (15:24 -0700)]
epoll: clean up ep_modify

ep_modify() doesn't need to set event.data from within the ep->lock
spinlock as the comment suggests.  The only place event.data is used is
ep_send_events_proc(), and this is protected by ep->mtx instead of
ep->lock.  Also update the comment for mutex_lock() at the top of
ep_scan_ready_list(), which mentions epoll_ctl(EPOLL_CTL_DEL) but not
epoll_ctl(EPOLL_CTL_MOD).

ep_modify() can also use spin_lock_irq() instead of spin_lock_irqsave().

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: remove unnecessary xchg
Tony Battersby [Tue, 31 Mar 2009 22:24:15 +0000 (15:24 -0700)]
epoll: remove unnecessary xchg

xchg in ep_unregister_pollwait() is unnecessary because it is protected by
either epmutex or ep->mtx (the same protection as ep_remove()).

If xchg was necessary, it would be insufficient to protect against
problems: if multiple concurrent calls to ep_unregister_pollwait() were
possible then a second caller that returns without doing anything because
nwait == 0 could return before the waitqueues are removed by the first
caller, which looks like it could lead to problematic races with
ep_poll_callback().

So remove xchg and add comments about the locking.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: remember the event if epoll_wait returns -EFAULT
Tony Battersby [Tue, 31 Mar 2009 22:24:14 +0000 (15:24 -0700)]
epoll: remember the event if epoll_wait returns -EFAULT

If epoll_wait returns -EFAULT, the event that was being returned when the
fault was encountered will be forgotten.  This is not a big deal since
EFAULT will happen only if a buggy userspace program passes in a bad
address, in which case what happens later usually doesn't matter.
However, it is easy to remember the event for later, and this patch makes
a simple change to do that.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: don't use current in irq context
Tony Battersby [Tue, 31 Mar 2009 22:24:13 +0000 (15:24 -0700)]
epoll: don't use current in irq context

ep_call_nested() (formerly ep_poll_safewake()) uses "current" (without
dereferencing it) to detect callback recursion, but it may be called from
irq context where the use of current is generally discouraged.  It would
be better to use get_cpu() and put_cpu() to detect the callback recursion.

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: remove debugging code
Davide Libenzi [Tue, 31 Mar 2009 22:24:12 +0000 (15:24 -0700)]
epoll: remove debugging code

Remove debugging code from epoll.  There's no need for it to be included
into mainline code.

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: fix epoll's own poll (update)
Davide Libenzi [Tue, 31 Mar 2009 22:24:11 +0000 (15:24 -0700)]
epoll: fix epoll's own poll (update)

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoepoll: fix epoll's own poll
Davide Libenzi [Tue, 31 Mar 2009 22:24:10 +0000 (15:24 -0700)]
epoll: fix epoll's own poll

Fix a bug inside the epoll's f_op->poll() code, that returns POLLIN even
though there are no actual ready monitored fds.  The bug shows up if you
add an epoll fd inside another fd container (poll, select, epoll).

The problem is that callback-based wake ups used by epoll does not carry
(patches will follow, to fix this) any information about the events that
actually happened.  So the callback code, since it can't call the file*
->poll() inside the callback, chains the file* into a ready-list.

So, suppose you added an fd with EPOLLOUT only, and some data shows up on
the fd, the file* mapped by the fd will be added into the ready-list (via
wakeup callback).  During normal epoll_wait() use, this condition is
sorted out at the time we're actually able to call the file*'s
f_op->poll().

Inside the old epoll's f_op->poll() though, only a quick check
!list_empty(ready-list) was performed, and this could have led to
reporting POLLIN even though no ready fds would show up at a following
epoll_wait().  In order to correctly report the ready status for an epoll
fd, the ready-list must be checked to see if any really available fd+event
would be ready in a following epoll_wait().

Operation (calling f_op->poll() from inside f_op->poll()) that, like wake
ups, must be handled with care because of the fact that epoll fds can be
added to other epoll fds.

Test code:

/*
 *  epoll_test by Davide Libenzi (Simple code to test epoll internals)
 *  Copyright (C) 2008  Davide Libenzi
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 *  Davide Libenzi <davidel@xmailserver.org>
 *
 */

#include <sys/types.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <signal.h>
#include <limits.h>
#include <poll.h>
#include <sys/epoll.h>
#include <sys/wait.h>

#define EPWAIT_TIMEO (1 * 1000)
#ifndef POLLRDHUP
#define POLLRDHUP 0x2000
#endif

#define EPOLL_MAX_CHAIN 100L

#define EPOLL_TF_LOOP (1 << 0)

struct epoll_test_cfg {
long size;
long flags;
};

static int xepoll_create(int n) {
int epfd;

if ((epfd = epoll_create(n)) == -1) {
perror("epoll_create");
exit(2);
}

return epfd;
}

static void xepoll_ctl(int epfd, int cmd, int fd, struct epoll_event *evt) {
if (epoll_ctl(epfd, cmd, fd, evt) < 0) {
perror("epoll_ctl");
exit(3);
}
}

static void xpipe(int *fds) {
if (pipe(fds)) {
perror("pipe");
exit(4);
}
}

static pid_t xfork(void) {
pid_t pid;

if ((pid = fork()) == (pid_t) -1) {
perror("pipe");
exit(5);
}

return pid;
}

static int run_forked_proc(int (*proc)(void *), void *data) {
int status;
pid_t pid;

if ((pid = xfork()) == 0)
exit((*proc)(data));
if (waitpid(pid, &status, 0) != pid) {
perror("waitpid");
return -1;
}

return WIFEXITED(status) ? WEXITSTATUS(status): -2;
}

static int check_events(int fd, int timeo) {
struct pollfd pfd;

fprintf(stdout, "Checking events for fd %d\n", fd);
memset(&pfd, 0, sizeof(pfd));
pfd.fd = fd;
pfd.events = POLLIN | POLLOUT;
if (poll(&pfd, 1, timeo) < 0) {
perror("poll()");
return 0;
}
if (pfd.revents & POLLIN)
fprintf(stdout, "\tPOLLIN\n");
if (pfd.revents & POLLOUT)
fprintf(stdout, "\tPOLLOUT\n");
if (pfd.revents & POLLERR)
fprintf(stdout, "\tPOLLERR\n");
if (pfd.revents & POLLHUP)
fprintf(stdout, "\tPOLLHUP\n");
if (pfd.revents & POLLRDHUP)
fprintf(stdout, "\tPOLLRDHUP\n");

return pfd.revents;
}

static int epoll_test_tty(void *data) {
int epfd, ifd = fileno(stdin), res;
struct epoll_event evt;

if (check_events(ifd, 0) != POLLOUT) {
fprintf(stderr, "Something is cooking on STDIN (%d)\n", ifd);
return 1;
}
epfd = xepoll_create(1);
fprintf(stdout, "Created epoll fd (%d)\n", epfd);
memset(&evt, 0, sizeof(evt));
evt.events = EPOLLIN;
xepoll_ctl(epfd, EPOLL_CTL_ADD, ifd, &evt);
if (check_events(epfd, 0) & POLLIN) {
res = epoll_wait(epfd, &evt, 1, 0);
if (res == 0) {
fprintf(stderr, "Epoll fd (%d) is ready when it shouldn't!\n",
epfd);
return 2;
}
}

return 0;
}

static int epoll_wakeup_chain(void *data) {
struct epoll_test_cfg *tcfg = data;
int i, res, epfd, bfd, nfd, pfds[2];
pid_t pid;
struct epoll_event evt;

memset(&evt, 0, sizeof(evt));
evt.events = EPOLLIN;

epfd = bfd = xepoll_create(1);

for (i = 0; i < tcfg->size; i++) {
nfd = xepoll_create(1);
xepoll_ctl(bfd, EPOLL_CTL_ADD, nfd, &evt);
bfd = nfd;
}
xpipe(pfds);
if (tcfg->flags & EPOLL_TF_LOOP)
{
xepoll_ctl(bfd, EPOLL_CTL_ADD, epfd, &evt);
/*
 * If we're testing for loop, we want that the wakeup
 * triggered by the write to the pipe done in the child
 * process, triggers a fake event. So we add the pipe
 * read size with EPOLLOUT events. This will trigger
 * an addition to the ready-list, but no real events
 * will be there. The the epoll kernel code will proceed
 * in calling f_op->poll() of the epfd, triggering the
 * loop we want to test.
 */
evt.events = EPOLLOUT;
}
xepoll_ctl(bfd, EPOLL_CTL_ADD, pfds[0], &evt);

/*
 * The pipe write must come after the poll(2) call inside
 * check_events(). This tests the nested wakeup code in
 * fs/eventpoll.c:ep_poll_safewake()
 * By having the check_events() (hence poll(2)) happens first,
 * we have poll wait queue filled up, and the write(2) in the
 * child will trigger the wakeup chain.
 */
if ((pid = xfork()) == 0) {
sleep(1);
write(pfds[1], "w", 1);
exit(0);
}

res = check_events(epfd, 2000) & POLLIN;

if (waitpid(pid, NULL, 0) != pid) {
perror("waitpid");
return -1;
}

return res;
}

static int epoll_poll_chain(void *data) {
struct epoll_test_cfg *tcfg = data;
int i, res, epfd, bfd, nfd, pfds[2];
pid_t pid;
struct epoll_event evt;

memset(&evt, 0, sizeof(evt));
evt.events = EPOLLIN;

epfd = bfd = xepoll_create(1);

for (i = 0; i < tcfg->size; i++) {
nfd = xepoll_create(1);
xepoll_ctl(bfd, EPOLL_CTL_ADD, nfd, &evt);
bfd = nfd;
}
xpipe(pfds);
if (tcfg->flags & EPOLL_TF_LOOP)
{
xepoll_ctl(bfd, EPOLL_CTL_ADD, epfd, &evt);
/*
 * If we're testing for loop, we want that the wakeup
 * triggered by the write to the pipe done in the child
 * process, triggers a fake event. So we add the pipe
 * read size with EPOLLOUT events. This will trigger
 * an addition to the ready-list, but no real events
 * will be there. The the epoll kernel code will proceed
 * in calling f_op->poll() of the epfd, triggering the
 * loop we want to test.
 */
evt.events = EPOLLOUT;
}
xepoll_ctl(bfd, EPOLL_CTL_ADD, pfds[0], &evt);

/*
 * The pipe write mush come before the poll(2) call inside
 * check_events(). This tests the nested f_op->poll calls code in
 * fs/eventpoll.c:ep_eventpoll_poll()
 * By having the pipe write(2) happen first, we make the kernel
 * epoll code to load the ready lists, and the following poll(2)
 * done inside check_events() will test nested poll code in
 * ep_eventpoll_poll().
 */
if ((pid = xfork()) == 0) {
write(pfds[1], "w", 1);
exit(0);
}
sleep(1);
res = check_events(epfd, 1000) & POLLIN;

if (waitpid(pid, NULL, 0) != pid) {
perror("waitpid");
return -1;
}

return res;
}

int main(int ac, char **av) {
int error;
struct epoll_test_cfg tcfg;

fprintf(stdout, "\n********** Testing TTY events\n");
error = run_forked_proc(epoll_test_tty, NULL);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing short wakeup chain\n");
error = run_forked_proc(epoll_wakeup_chain, &tcfg);
fprintf(stdout, error == POLLIN ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = EPOLL_MAX_CHAIN;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing long wakeup chain (HOLD ON)\n");
error = run_forked_proc(epoll_wakeup_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing short poll chain\n");
error = run_forked_proc(epoll_poll_chain, &tcfg);
fprintf(stdout, error == POLLIN ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = EPOLL_MAX_CHAIN;
tcfg.flags = 0;
fprintf(stdout, "\n********** Testing long poll chain (HOLD ON)\n");
error = run_forked_proc(epoll_poll_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = EPOLL_TF_LOOP;
fprintf(stdout, "\n********** Testing loopy wakeup chain (HOLD ON)\n");
error = run_forked_proc(epoll_wakeup_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

tcfg.size = 3;
tcfg.flags = EPOLL_TF_LOOP;
fprintf(stdout, "\n********** Testing loopy poll chain (HOLD ON)\n");
error = run_forked_proc(epoll_poll_chain, &tcfg);
fprintf(stdout, error == 0 ?
"********** OK\n": "********** FAIL (%d)\n", error);

return 0;
}

Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>