epson1355fb.c: fix error handling code

Fix error handling code

Signed-off-by: Roland Stigge <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

cirrusfb: convert to generic boolean

Convert to generic boolean.

Signed-off-by: Richard Knutsson <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

VIDEO: add spaces on either side of the case "..." operator

Following the programming advice laid down in the gcc manual, make
sure the case "..." operator has spaces on either side.

Signed-off-by: Robert P. J. Day <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

skeletonfb: various corrections

This is mainly correction of types, typos and missing characters in the
skeletonfb.c file found while trying to prepare a new fb driver.

[adaplas]
Additions on power management and fixes as per akpm

Signed-off-by: Krzysztof Helt <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

nvidiafb: access CRT registers safely

Use Read/WriteCrtc() to access CRTC registers in nv_i2c.c.  These are safer
  because it uses the correct CRTC base (0x3bx or 0x3dx).

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

nvidiafb: ensure that CRTC registers are accessible

- Ensure that CRTC registers are accessible by unlocking them on set_par(),
  otherwise i2c reading will fail.

- The function nvidia_vga_protect(), does not protect the VGA registers, but
  turns off the screen.  Rename it to nvidia_screen_off().

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rivafb: fixed reversed DDC ports

Fixed reversed DDC port so assignments are congruent with nvidiafb's and X's
nv driver.

Signed-off-by: Antonino Daplas <[email protected]>
Cc: Jean Delvare <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rivafb/nvidiafb: Various cleanups

Various cleanups to rivafb/nvidiafb's I2C code:
* Drop useless par->bus.
* Refactor I2C bus deletion code.
* Drop useless variable initialization.
* Remove unneeded include of <linux/i2c-id.h>.
* Simplify +1/-1.
* Add __devinit tags where possible.

[adaplas]
The varible initialization are not useless. However, rivafb must
check if i2c bus are created properly before reading the EDID
block.

Signed-off-by: Jean Delvare <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rivafb: handle I2C bus creation failure

I2C bus creation may fail, let rivafb handle that case properly.

Signed-off-by: Jean Delvare <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rivafb/nvidiafb: Enable hardware monitoring

Let the hardware monitoring drivers probe the second rivafb/nvidiafb I2C bus
for devices.

Signed-off-by: Jean Delvare <khali@linux-fr>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: don't show logo if driver or fbcon are modular

It was always intended for the logo to be drawn only if both fbcon and the
driver that is mapped to it are both compiled statically.  Currently, if fbcon
is loaded prior to the driver, the logo is not shown.  Reverse the order, and
the code may attempt to draw the logo which is __initdata.  By accident, this
bug is rarely seen because this method of loading the modules is not common
and secondly, a code in fb_prepare_logo() that checks the height of the logo
(now a random value) rarely succeeds.

Fix by drawing the logo only if both fbcon and the driver are statically
compiled.

Signed-off-by: Antonino Daplas <[email protected]>
Cc: Jean Delvare <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

s3fb: fix PCI must_checks

drivers/video/s3fb.c:1078: warning: ignoring return value of 'pci_enable_device', declared with attribute warn_unused_result

Signed-off-by: Randy Dunlap <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

vt: expose system-wide UTF-8 default setting via sysfs

Create a variable, default_utf8, that defines the system-wide default UTF-8
setting.  This variable can be altered via sysfs. If the variable is properly
set, this should mimimize breakage of UTF-8 encoded consoles when doing a
reset or echo -e '\033c' and of newly opened/allocated consoles.

This is based from patches by Jan Engelhardt and Paul LeoNerd Evans.

Signed-off-by: Antonino Daplas <[email protected]>
Cc: Jan Engelhardt <[email protected]>
Cc: Paul LeoNerd Evans <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbcon: don't draw cursor when it's disabled

When the cursor and echo are disabled on the current console, pressing a
key will cause a black rectangle to be painted in the cursor's position.
Fix this by not touching the framebuffer in fbcon_cursor() when the
cursor is off.

Signed-off-by: Michal Januszewski <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

SIS USB2VGA Warning fix

drivers/usb/misc/sisusbvga/sisusb_con.c:1436: warning: initialization from incompatible pointer type

Signed-off-by: WANG Cong <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Cc: Jan Engelhardt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

vt-add-color-support-to-the-underline-and-italic-attributes-fix

Cc: "Antonino A. Daplas" <[email protected]>
Cc: Jan Engelhardt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

vt: add color support to the "underline" and "italic" attributes

Add color support to the "underline" and "italic" attributes as in
OpenBSD/NetBSD-style (vt220) and xterm.

Signed-off-by: Jan Engelhardt <[email protected]>
Acked-by: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

vt: allow for the palette to be exposed and changed via sysfs

Allow for the palette to be exposed and changed via sysfs.  A call to
/usr/bin/reset will slurp the new definitions in for the current console.

Already posted at http://lkml.org/lkml/2006/1/15/149

Signed-off-by: Jan Engelhardt <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

nvidiafb: Fix reversed DDC port

After I added some debugging printks I've found that code became a bit
confused because it believed that primary monitor is 1920x540, but later it
found in CRTC0's registers that panel size is 1920x1200 (Windows also agree
that 1920x1200 is primary monitor, and 1920x1080i secondary one).

When I applied attached patch then my monitor became as happy as it was
before I connected HDMI cable to secondary output.

Signed-off-by: Petr Vandrovec <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Cc: Dave Airlie <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

radeonfb: Add support for Radeon xpress 200m

Added support for radeon xpress 200m(rs480).  Note that the card doesn't
like dynclk turned on.

Signed-off-by: Johan Henriksson <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: correct image offsets when rotating logo

Correct the image offsets when rotating the logo.  Before image->dx and
image->dy were always zero, so nobody ever noticed.

Signed-off-by: Geert Uytterhoeven <[email protected]>
Acked-By: James Simmons <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: fb_do_show_logo() updates

fb_do_show_logo() updates:
  - Use width and height of the passed image instead of the global variable
    fb_logo
  - Explicitly pass the number of logos to draw

Signed-off-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Geoff Levand <[email protected]>
Acked-By: James Simmons <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: avoid vertical overflow when making space for the logo

fbcon_prepare_logo(): Avoid vertical overflow when making space for the logo

Signed-off-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Geoff Levand <[email protected]>
Acked-By: James Simmons <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

remove unused header file: drivers/video/riva/nv4ref.h

Signed-off-by: Robert P. J. Day <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

tgafb accelerated code

Add accelerated panning and accelerated color and mono image drawing.
Please apply.

Signed-off-by: James Simmons <[email protected]>
Acked-by: Maciej W. Rozycki <[email protected]> (tested, too!)
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

tgafb: TURBOchannel support

This is support for the TC variations of the TGA boards (properly known as
SFB+ or Smart Frame Buffer Plus boards).  The 8-plane SFB+ board uses the
Bt459 RAMDAC (unlike its PCI TGA counterpart, which uses the Bt485), so
bits have been added to support this chip as well.

Signed-off-by: Maciej W. Rozycki <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Cc: James Simmons <[email protected]>
Acked-by: Ralf Baechle <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/video/hecubafb.c: make 4 functions static

This patch makes the following needlessly global functions static:
- hcb_wait_for_ack()
- hcb_wait_for_ack_clear()
- apollo_send_data()
- apollo_send_command()

Signed-off-by: Adrian Bunk <[email protected]>
Cc: Jaya Kumar <[email protected]>
Cc: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: hecuba Framebuffer Driver

This patch implements support for the E-Ink/hecuba display device.  It uses
deferred IO support.

[[email protected]: linker section fixes]
Signed-off-by: Jaya Kumar <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fb: fsync() method for deferred I/O flush.

There are cases when we do not want to wait on the delay for automatically
updating the "real" framebuffer, this implements a simple ->fsync() hook
for explicitly flushing the deferred I/O work.  The ->page_mkwrite()
handler will rearm the work queue normally.

(akpm: nuke unneeded ifdefs, forward-delcare struct dentry)

Signed-off-by: Paul Mundt <[email protected]>
Cc: Jaya Kumar <[email protected]>
Acked-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

make fb_deferred_io_mkwrite() static

Signed-off-by: Adrian Bunk <[email protected]>
Cc: Jaya Kumar <[email protected]>
Cc: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: mm: Deferred IO support

This implements deferred IO support in fbdev.  Deferred IO is a way to delay
and repurpose IO.  This implementation is done using mm's page_mkwrite and
page_mkclean hooks in order to detect, delay and then rewrite IO.  This
functionality is used by hecubafb.

[adaplas]
This is useful for graphics hardware with no directly addressable/mappable
framebuffer. Implementing this will allow the "framebuffer" to be accesible
from user space via mmap().

Signed-off-by: Jaya Kumar <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

atyfb: kill dead code

Coverity Bug 68, kill dead code.

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

neofb: fill transp->msb_right with the correct value

Fill transp->msb_right with the correct value.  fbcon ignores this, but user
apps may not.

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: fix obvious bug in show_pan()

show_pan will display the value of the xoffset twice, instead of the xoffset
and yoffset.  Fix.

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: ignore VESA modes if framebuffer is disabled

If the option vga=<VESA graphics mode> is added to the boot parameter, it will
activate graphics mode, but without any framebuffer support, the user is left
with an unusable display.

Change the behavior such that the user is instead prompted for another mode
(ala vga=ask).

NOTE: People can always use vbetool to set a graphics mode if this is really
desired, but the number of people doing this approaches zero.

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

nvidiafb: bring back generic ddc reading

Make nvidiafb use fb_ddc_read().  This patch was submitted before but was
reverted due to problems in a non-x86 platform.  This includes a fix for that
where ddc reading is bypassed if there is no DDC bus (duh).

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

intelfb: fix ring space calculation

If there is less than RING_MIN_FREE available in the ring buffer,
dinfo->ring_space is set to a big value forcing wait_ring to return.

Fix by making ring space = 0 if ring space < RING_MIN_FREE.

Signed-off-by: Dave Airlie <[email protected]>
Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: add Ultrasharp UXGA to broken monitor database

This particular monitor does not have a limits block and has only one set of
monitor timings.  Fix by adding a limits block to the EDID and extend the
horizontal sync frequency range to 30 kHz and 75 Khz.

Signed-off-by: Antonino Daplas <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

make drivers/video/display/display-sysfs.c:display_class static

Signed-off-by: Adrian Bunk <[email protected]>
Cc: James Simmons <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Cc: Greg KH <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: display class

Add the new display class.  This is meant to unite the various solutions to
display units ie acpi output device, auxdisplay and the defunct lcd class
in the backlight directory.

Signed-off-by: James Simmons <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Cc: Greg KH <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

change rivafb_remove to __devexit

Change rivafb_remove to __deviexit to fix MODPOST warnings:

WARNING: drivers/video/riva/rivafb.o - Section mismatch: reference to
.exit.text:rivafb_remove from .data.rel.local after 'rivafb_driver' (at offset 0x28)

Signed-off-by: Prarit Bhargava <[email protected]>
Acked-by: James Simmons <[email protected]>
Cc: "Antonino A. Daplas" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

aty128fb: fix blanking

I have a problem with blanking. The soundcard uses speakers of the monitor.
Sound is muted when the screen blanks due to a bug in aty128fb.c.

Here is a fragment of linux/fb.h
/* VESA Blanking Levels */
#define VESA_NO_BLANKING        0
#define VESA_VSYNC_SUSPEND      1
#define VESA_HSYNC_SUSPEND      2
#define VESA_POWERDOWN          3

enum {
        /* screen: unblanked, hsync: on,  vsync: on */
        FB_BLANK_UNBLANK       = VESA_NO_BLANKING,

        /* screen: blanked,   hsync: on,  vsync: on */
        FB_BLANK_NORMAL        = VESA_NO_BLANKING + 1,

        /* screen: blanked,   hsync: on,  vsync: off */
        FB_BLANK_VSYNC_SUSPEND = VESA_VSYNC_SUSPEND + 1,

        /* screen: blanked,   hsync: off, vsync: on */
        FB_BLANK_HSYNC_SUSPEND = VESA_HSYNC_SUSPEND + 1,

        /* screen: blanked,   hsync: off, vsync: off */
        FB_BLANK_POWERDOWN     = VESA_POWERDOWN + 1
};

So FB_BLANK_NORMAL is 1, FB_BLANK_VSYNC_SUSPEND is 2,
FB_BLANK_HSYNC_SUSPEND is 3, FB_BLANK_POWERDOWN is 4.
And now:
blank = FB_BLANK_NORMAL (1)
blank & FB_BLANK_HSYNC_SUSPEND (1 & 3) is true,
so normal blank caused hsync suspend and sound is muted.

Cc: James Simmons <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, copyright and version changes

- add copyright
- move version one number upper and use MODULE_VERSION

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, probe cleanup

- add fail paths
- merge 3 similar initializations into one (Z, Ze, Y)

[[email protected]: build fix]
Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, dynamic ports

and save thus approx. 160k of .bss

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, fix tty device unregister

put_tty_driver after unregistering devices

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, conditions cleanup

- 0 is not NULL
- use unlikely to hit the icache in isr more likely
- remove or comment empty if/else paths

[[email protected]: build fix]
Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove locking macros

and use locally stored card structure if possible

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove unused timestamps

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, timers cleanup

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, fix blockmove

tty has no longer flip buffers accessible externally. Fix it by moving the
code to the tty_*flip* helpers.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove sleep_on

convert to wait_* and completion

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, make info->card a pointer

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, get rid of phys addresses

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, simplify variables initialization

- do not init static variables to 0
- simplify cy_init_card -- use memset(0) and do not zero each element
  separately, also reorder init, so that same entries are inited at one
  place

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, mark cyy_init_card as __devinit, not __init

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, printk cleanups

- add printk KERN_ levels to each printk
- substitute printk with dev_* when device struct is available

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, irq is int

don't fetch it to uchar

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove useless fileds from cyclades_card

pde, ctl_phys and base_phys are useless -- they are never used.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, unexport struct cyclades_card

Do not export internal card data to userspace. cytune doesn't use this
anyway.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, depends on PCI or ISA

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, switch to pci probing

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, use IS_CYC_Z macro

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: isicom, use pr_debug

isicom, use pr_debug

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: mxser, schedule for removal

mxser, schedule for removal

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, allow DEBUG_SHIRQ

Test if base addr is non-null in ISR to prove the card has been correctly
initialized.  This is needed for DEBUG_SHIRQ for example.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, clear interrupts before releasing

Without this patch, the driver sometimes causes "IRQXX: Nobody cares".  Fix it
by turning off irqs when releasing.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, tty_register_device separately for each device

Do not register all tty devices at the init time, delay it for the time until
some device is found.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, cy_init error handling

- do not panic if tty_register_driver fails
- handle fail paths properly

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove some global vars

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, init card struct immediately

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, move card entries init into function

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, create cy_pci_probe

Move probing code to separate function for easy pci probing conversion.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, init Ze immediately

There will be no other choice after introducing pci probing anyway.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, use pci_iomap/unmap

fork remove code for pci -- move it to separate, new, function and don't care
about pci in the former.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, create cy_init_Ze

Move Ze init code into new cy_init_Ze, because we will need it in another
place and it will make the code totally unreadable.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

x86, serial: convert legacy COM ports to platform devices

Make x86 COM ports into platform devices and don't probe for them
if we have PNP.

This prevents double discovery, where a device was found both by
the legacy probe and by 8250_pnp, e.g.,

    serial8250: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A
    00:02: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A

This also means IRDA devices without a UART PNP ID will no longer be
claimed by the serial driver, which might require changes in IRDA
drivers and administration.

In addition to this patch, you may need to configure a setserial init
script, e.g., /etc/init.d/setserial, so it doesn't poke legacy UART
stuff back in.  On Debian, "dpkg-reconfigure setserial" with the "kernel"
option does this.

To force the old legacy probe behavior even when we have PNPBIOS or
ACPI, load the new legacy_serial module (or build 8250 static) with
the "legacy_serial.force" option.

[[email protected]: fix makefiles]
Signed-off-by: Bjorn Helgaas <[email protected]>
Cc: Keith Owens <[email protected]>
Cc: Len Brown <[email protected]>
Cc: Adam Belay <[email protected]>
Cc: Matthieu CASTET <[email protected]>
Cc: Jean Tourrilhes <[email protected]>
Cc: Matthew Garrett <[email protected]>
Cc: Ville Syrjala <[email protected]>
Cc: Russell King <[email protected]>
Cc: Samuel Ortiz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

smsc-ircc2: add PNP support

Claim devices using PNP, unless the user explicitly specified device
addresses.  This can be disabled with the "smsc-ircc2.nopnp" option.

This removes the need for probing legacy addresses and helps untangle IR
devices from serial8250 devices.

Sometimes the SMC device is at a legacy COM port address but does not use the
legacy COM IRQ.  In this case, claiming the device using PNP rather than 8250
legacy probe means we can automatically use the correct IRQ rather than
forcing the user to use "setserial" to set the IRQ manually.

If the PNP claim doesn't work, make sure you don't have a setserial init
script, e.g., /etc/init.d/setserial, configured to poke in legacy COM port
resources for the IRDA device.  That causes the serial driver to claim
resources needed by this driver.

Based on this patch by Ville Syrjälä:
    http://www.hpl.hp.com/personal/Jean_Tourrilhes/IrDA/ir260_smsc_pnp.diff

Signed-off-by: Bjorn Helgaas <[email protected]>
Cc: Keith Owens <[email protected]>
Cc: Len Brown <[email protected]>
Cc: Adam Belay <[email protected]>
Cc: Matthieu CASTET <[email protected]>
Cc: Jean Tourrilhes <[email protected]>
Cc: Matthew Garrett <[email protected]>
Cc: Ville Syrjala <[email protected]>
Cc: Russell King <[email protected]>
Acked-by: Samuel Ortiz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

smsc-ircc2: tidy up module parameter checking

To determine whether the user specified a module parameter, use some #defines
instead of checking for bare magic numbers.

Signed-off-by: Bjorn Helgaas <[email protected]>
Cc: Keith Owens <[email protected]>
Cc: Len Brown <[email protected]>
Cc: Adam Belay <[email protected]>
Cc: Matthieu CASTET <[email protected]>
Cc: Jean Tourrilhes <[email protected]>
Cc: Matthew Garrett <[email protected]>
Cc: Ville Syrjala <[email protected]>
Cc: Russell King <[email protected]>
Cc: Samuel Ortiz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

PNP: workaround HP BIOS defect that leaves SMCF010 device partly enabled

Some HP/Compaq firmware reports via ACPI that the SMCF010 IR device is
enabled, but in fact, it leaves the device partly disabled.

HP nw8240 BIOS 68DTV Ver.  F.0F, released 9/15/2005 is one BIOS that has this
problem.

Signed-off-by: Bjorn Helgaas <[email protected]>
Cc: Keith Owens <[email protected]>
Cc: Len Brown <[email protected]>
Cc: Adam Belay <[email protected]>
Cc: Matthieu CASTET <[email protected]>
Cc: Jean Tourrilhes <[email protected]>
Cc: Matthew Garrett <[email protected]>
Cc: Ville Syrjala <[email protected]>
Cc: Russell King <[email protected]>
Cc: Samuel Ortiz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

PNP: notice whether we have PNP devices (PNPBIOS or PNPACPI)

This series converts i386 and x86_64 legacy serial ports to be platform
devices and prevents probing for them if we have PNP.

This prevents double discovery, where a device was found both by the legacy
probe and by 8250_pnp.

This also prevents the serial driver from claiming IRDA devices (unless they
have a UART PNP ID).  The serial legacy probe sometimes assumed the wrong IRQ,
so the user had to use "setserial" to fix it.

Removing the need for setserial to make IRDA devices work seems good, but it
does break some things.  In particular, you may need to keep setserial from
poking legacy UART stuff back in by doing something like "dpkg-reconfigure
setserial" with the "kernel" option.  Otherwise, the setserial-discovered
"UART" will claim resources and prevent the IRDA driver from loading.

This patch:

If we can discover devices using PNP, we can skip some legacy probes.  This
flag ("pnp_platform_devices") indicates that PNPBIOS or PNPACPI is enabled and
should tell us about builtin devices.

Signed-off-by: Bjorn Helgaas <[email protected]>
Cc: Keith Owens <[email protected]>
Cc: Len Brown <[email protected]>
Cc: Adam Belay <[email protected]>
Cc: Matthieu CASTET <[email protected]>
Cc: Jean Tourrilhes <[email protected]>
Cc: Matthew Garrett <[email protected]>
Cc: Ville Syrjala <[email protected]>
Cc: Russell King <[email protected]>
Cc: Samuel Ortiz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove useless casts

cyclades, remove useless casts

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove volatiles

cyclades, remove volatiles

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, timer cleanup

cyclades, timer cleanup

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, cy_readX/writeX cleanup

cyclades, cy_readX/writeX cleanup

- cy_readX are placeholders for readX, remove it
- move cy_writeX macros into do {} while(0) to be safe

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Char: cyclades, remove PAUSE

cyclades, remove PAUSE

PAUSE expands to do {} while (0), it's useless.

Signed-off-by: Jiri Slaby <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Add IRQF_IRQPOLL flag on arm

Add IRQF_IRQPOLL for each timer interrupt.

Signed-off-by: Bernhard Walle <[email protected]>
Cc: Russell King <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Alan Cox <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Add IRQF_IRQPOLL flag on parisc

Add IRQF_IRQPOLL to the timer interrupt on parisc.

Signed-off-by: Bernhard Walle <[email protected]>
Cc: Kyle McMartin <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Grant Grundler <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Add IRQF_IRQPOLL flag on sh

Add IRQF_IRQPOLL on each timer interrupt on SH2.

Signed-off-by: Bernhard Walle <[email protected]>
Cc: Paul Mundt <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Alan Cox <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Add IRQF_IRQPOLL flag on IA64

Add IRQF_IRQPOLL for the timer interrupt on IA64.

Signed-off-by: Bernhard Walle <[email protected]>
Cc: "Luck, Tony" <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Alan Cox <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Add IRQF_IRQPOLL flag on i386

Add IRQF_IRQPOLL to timer interrupts on i386.

Signed-off-by: Bernhard Walle <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Alan Cox <[email protected]>
Cc: James Bottomley <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Add IRQF_IRQPOLL flag on x86_64

Add IRQF_IRQPOLL for the timer interrupt on x86_64.

Signed-off-by: Bernhard Walle <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Alan Cox <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Add IRQF_IRQPOLL flag (common code)

irqpoll is broken on some architectures that don't use the IRQ 0 for the timer
interrupt like IA64.  This patch adds a IRQF_IRQPOLL flag.

Each architecture is handled in a separate pach.  As I left the irq == 0 as
condition, this should not break existing architectures that use timer_irq ==
0 and that I did't address with that patch (because I don't know).

This patch:

This patch adds a IRQF_IRQPOLL flag that the interrupt registration code could
use for the interrupt it wants to use for IRQ polling.

Because this must not be the timer interrupt, an additional flag was added
instead of re-using the IRQF_TIMER constant.  Until all architectures will
have an IRQF_IRQPOLL interrupt, irq == 0 will stay as alternative as it should
not break anything.

Also, note_interrupt() is called on CPU-specific interrupts to be used as
interrupt source for IRQ polling.

Signed-off-by: Bernhard Walle <[email protected]>
Cc: Russell King <[email protected]>
Cc: Kyle McMartin <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Cc: Grant Grundler <[email protected]>
Cc: Paul Mundt <[email protected]>
Cc: "Luck, Tony" <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Alan Cox <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

udf: possible null pointer dereference while load_partition

sb_read may return NULL, let's explicitly check it.

Signed-off-by: Dmitriy Monakhov <[email protected]>
Cc: Jan Kara <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

udf: support files larger than 1G

Make UDF work correctly for files larger than 1GB.  As no extent can be
longer than (1<<30)-blocksize bytes, we have to create several extents if a
big hole is being created.  As a side-effect, we now don't discard
preallocated blocks when creating a hole.

Signed-off-by: Jan Kara <[email protected]>
Acked-by: Christoph Hellwig <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

udf: add assertions

Add a few assertions into udf_discard_prealloc() to check that the file is
sane (mostly helps debugging further patches ;).

Signed-off-by: Jan Kara <[email protected]>
Acked-by: Christoph Hellwig <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

udf: use get_bh()

Make UDF use get_bh() instead of directly accessing b_count and use
brelse() instead of udf_release_data() which does just brelse()...

Signed-off-by: Jan Kara <[email protected]>
Acked-by: Christoph Hellwig <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

UDF: introduce struct extent_position

Introduce a structure extent_position to store a position of an extent and
the corresponding buffer_head in one place.

Signed-off-by: Jan Kara <[email protected]>
Acked-by: Christoph Hellwig <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

udf: use sector_t and loff_t for file offsets

Use sector_t and loff_t for file offsets in UDF filesystem.  Otherwise an
overflow may occur for long files.  Also make inode_bmap() return offset in
the extent in number of blocks instead of number of bytes - for most
callers this is more convenient.

Signed-off-by: Jan Kara <[email protected]>
Acked-by: Christoph Hellwig <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

nfs: fix congestion control: use atomic_longs

Change the atomic_t in struct nfs_server to atomic_long_t in anticipation
of machines that can handle 8+TB of (4K) pages under writeback.

However I suspect other things in NFS will start going *bang* by then.

Signed-off-by: Peter Zijlstra <[email protected]>
Cc: Trond Myklebust <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>