Add a scrollback buffers for each VGA console. The benefit is that
the scrollback history is not flushed when switching between consoles
but is persistent.
The buffers are allocated on demand when a new console is opened.

This breaks tools like clear_console that rely on flushing the
scrollback history by switching back and forth between consoles
which is why this feature is disabled by default.
Use the escape sequence \e[3J instead for flushing the buffer.
Signed-off-by: NManuel Schölling <manuel.schoelling@gmx.de>
Reviewed-by: NAndrey Utkin <andrey_utkin@fastmail.com>
Tested-by: NAndrey Utkin <andrey_utkin@fastmail.com>
Tested-by: NAdam Borowski <kilobyte@angband.pl>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This new callback is in preparation for persistent scrollback buffer
support for VGA consoles.
With a single scrollback buffer for all consoles, we could flush the
buffer just by invocating consw->con_switch(). But when each VGA console
has its own scrollback buffer, we need a new callback to tell the
video console driver which buffer to flush.
Signed-off-by: NManuel Schölling <manuel.schoelling@gmx.de>
Reviewed-by: NAndrey Utkin <andrey_utkin@fastmail.com>
Tested-by: NAndrey Utkin <andrey_utkin@fastmail.com>
Tested-by: NAdam Borowski <kilobyte@angband.pl>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This refactoring is in preparation for persistent scrollback
support for VGA console.
Signed-off-by: NManuel Schölling <manuel.schoelling@gmx.de>
Reviewed-by: NAndrey Utkin <andrey_utkin@fastmail.com>
Tested-by: NAndrey Utkin <andrey_utkin@fastmail.com>
Tested-by: NAdam Borowski <kilobyte@angband.pl>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

These macros:
* CAN_LOAD_EGA_FONTS
* CAN_LOAD_PALETTE
* TRIDENT_GLITCH
* VGA_CAN_DO_64KB
* SLOW_VGA
are either always set or always unset. They come from the linux 2.1
times. And given nobody switched them to some configurable options, I
assume nobody actually uses them.

So remove the macros and leave in place appropriate branches of the
conditional code.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: <linux-fbdev@vger.kernel.org>
Cc: <linux-doc@vger.kernel.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

These variables:
* vga_can_do_color
* vgacon_text_mode_force
* vga_font_is_default
* vga_hardscroll_enabled
* vga_hardscroll_user_enable
* vga_init_done
* vga_is_gfx
* vga_palette_blanked
* vga_512_chars
are used exclusively as a boolean value, so make them really a bool.

Remove also useless "? true : false".

__read_mostly annotations removed too as they obfuscate the code and I
doubt they improve anything measurable given the variables are used
from .con_scroll, .con_startup and such.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: <linux-fbdev@vger.kernel.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The code is mirrorred in scrolldelta implementations of both vgacon
and sisusb. Let's move the code to a separate helper where we will
perform a common cleanup and further changes.

While we are moving the code, make it linear and save one indentation
level. This is done by returning from the "!lines" then-branch
immediatelly. This allows flushing the else-branch 1 level to the
left, obviously.

Few more new lines and comments were added too.

And do not forget to export the helper function given sisusb can be
built as module.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Thomas Winischhofer <thomas@winischhofer.net>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: <linux-fbdev@vger.kernel.org>
Cc: <linux-usb@vger.kernel.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Scrolling helpers scrup and scrdown both accept 'top' and 'bottom' as
unsigned int. Number of lines 'nr' is accepted as int, but all callers
pass down unsigned too. So change the type of 'nr' to unsigned too.
Now, promote unsigned int from the helpers up to the con_scroll
hook which actually accepted all those as signed int.

Next, the 'dir' parameter can have only two values and we define
constants for that: SM_UP and SM_DOWN. Switch them to enum and do
proper type checking on 'dir' too.

Finally, document the behaviour of the hook.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Thomas Winischhofer <thomas@winischhofer.net>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Cc: <linux-fbdev@vger.kernel.org>
Cc: <linux-usb@vger.kernel.org>
Cc: <linux-parisc@vger.kernel.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Namely convert:
* IS_FG -> con_is_fg
* DO_UPDATE -> con_should_update
* CON_IS_VISIBLE -> con_is_visible

DO_UPDATE was a weird name for a yes/no answer, so the new name is
con_should_update.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Thomas Winischhofer <thomas@winischhofer.net>
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: linux-usb@vger.kernel.org
Cc: linux-fbdev@vger.kernel.org
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

It is never called since commit 81732c3b (tty vt: Fix line
garbage in virtual console on command line edition) in 3.7. So remove
all the callbacks.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Thomas Winischhofer <thomas@winischhofer.net>
Cc: linux-usb@vger.kernel.org
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-fbdev@vger.kernel.org
Cc: linux-parisc@vger.kernel.org
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

* allow NULL consw->con_set_palette (some consoles define an empty
  hook)
* => remove empty hooks now
* return value of consw->con_set_palette is never checked => make the
  function void
* document consw->con_set_palette a bit
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Thomas Winischhofer <thomas@winischhofer.net>
Cc: linux-usb@vger.kernel.org
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-fbdev@vger.kernel.org
Cc: linux-parisc@vger.kernel.org
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

* allow NULL consw->con_scrolldelta (some consoles define an empty
  hook)
* => remove empty hooks now
* return value of consw->con_scrolldelta is never checked => make the
  function void
* document consw->con_scrolldelta a bit
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Cc: Thomas Winischhofer <thomas@winischhofer.net>
Cc: linux-usb@vger.kernel.org
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-fbdev@vger.kernel.org
Cc: linux-parisc@vger.kernel.org
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This means all ->con_set_palette have to have the second parameter
const too now.
Signed-off-by: NJiri Slaby <jslaby@suse.cz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Resolve shadow warnings that appear in W=2 builds by renaming
the "state" global to "vgastate".
Signed-off-by: NMark Rustad <mark.d.rustad@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@ti.com>

Touching the VGA resources on an IVB EFI machine causes hard hangs when
we then kick out the efifb. Ouch.

Apparently this also prevents unclaimed register errors on hsw and
hard machine hangs on my i855gm when trying to unbind fbcon.

Also, we want this to make I915_FBDEV=n safe.

v2: Rebase and pimp commit message.

v3: We also need to unregister the vga console, otherwise the unbind
of the fb console before module unload might resurrect it again.

v4: Ignore errors when the vga console is already unregistered - this
can happen when e.g. reloading i915.ko.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67813
Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: linux-fbdev@vger.kernel.org
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Reviewed-by: NDavid Herrmann <dh.herrmann@gmail.com>
Acked-by: NTomi Valkeinen <tomi.valkeinen@ti.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The vc_data.vc_uni_pagedir filed is currently long int, supposedly to
be served generically.  This, however, leads to lots of cast to
pointer, and rather it worsens the readability significantly.

Actually, we have now only a single uni_pagedir map implementation,
and this won't change likely.  So, it'd be much more simple and
error-prone to just use the exact pointer for struct uni_pagedir
instead of long.

Ditto for vc_uni_pagedir_loc.  It's a pointer to the uni_pagedir, thus
it can be changed similarly to the exact type.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The vgacon driver prepares a two element array of uni_pagedir_loc and
uses the second item as its own reference counter for sharing the
uni_pagedir.  And the code assumes blindly that the second item is
available if the assigned vc_uni_pagedir isn't the standard one, which
might be wrong (although currently it's so).

This patch fixes that wrong assumption, and gives a slight cleanup
along with it: namely, instead of array, just give the uni_pagedir_loc
and a separate refcount variable.  It makes the code a bit more
understandable at first glance.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This fixes the build error.
Signed-off-by: NDave Airlie <airlied@redhat.com>

Booting a 64-vcpu KVM guest, with CONFIG_PREEMPT_VOLUNTARY,
can result in a soft lockup:

BUG: soft lockup - CPU#41 stuck for 67s! [setfont:1505]
 RIP: 0010:[<ffffffff812c48da>]
[<ffffffff812c48da>] vgacon_do_font_op.clone.0+0x1ba/0x550

This is due to the 8192 (cmapsz) IO operations taking longer than expected
due to lock contention in QEMU.

Add conditional resched points in between writes allowing other tasks to
execute.
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

When we switch from 256->512 byte font rendering mode, it means the
current contents of the screen is being reinterpreted. The bit that holds
the high bit of the 9-bit font, may have been previously set, and thus
the new font misrenders.

The problem case we see is grub2 writes spaces with the bit set, so it
ends up with data like 0x820, which gets reinterpreted into 0x120 char
which the font translates into G with a circumflex. This flashes up on
screen at boot and is quite ugly.

A current side effect of this patch though is that any rendering on the
screen changes color to a slightly darker color, but at least the screen
no longer corrupts.

v2: as suggested by hpa, always clear the attribute space, whether we
are are going to or from 512 chars.

Cc: stable@vger.kernel.org
Signed-off-by: NDave Airlie <airlied@redhat.com>

The vga_lock lock can be taken in atomic context and therefore
cannot be preempted on -rt - annotate it.

In mainline this change documents the low level nature of
the lock - otherwise there's no functional difference. Lockdep
and Sparse checking will work as usual.
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Now vgacon_scrollback_startup() uses slab, not bootmem,
the comment above it is obsolete, so does __init_refok.
Signed-off-by: NWANG Cong <amwang@redhat.com>
Signed-off-by: NPaul Mundt <lethal@linux-sh.org>

It seems there is a small problem of VGA palette corruption on EFI
machine.  When the kernel initializes the architecture, it checks if the
machine is a EFI machine and assumes that a VGA console can exist.

When it initializes the console in vgacon_startup it checks if it can
really use the VGA console.  I think this is where a check is missing.
Currently, the function can fail if a VESA boot mode is detected but not if
a EFI boot mode was used.

Thus vgacon_startup() doesn't fail and initialize the video card for a real
VGA mode.  This function changes the first 16entries of the VGA palette.

When the efifb driver kicks in, the palette is not restored to default
ramp value, thus the 16 first entry remain in a modified state.  The
following patch prevent this corruption.
Signed-off-by: NYannick Heneault <yheneaul@matrox.com>
Cc: Dave Airlie <airlied@linux.ie>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The big kernel lock has been removed from all these files at some point,
leaving only the #include.

Remove this too as a cleanup.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

vgacon_do_font_op releases and reacquires the BTM while holding
console_sem. This violates the rule that BTM has to be the
outer lock whenever we hold both.

There does not seem to be any reason to give up the BTM here,
so just stop doing that.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

As a preparation for replacing the big kernel lock
in the TTY layer, wrap all the callers in new
macros tty_lock, tty_lock_nested and tty_unlock.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

warning: symbol 'vgacon_text_mode_force' was not declared. Should it be static?
Signed-off-by: NThiago Farina <tfransosi@gmail.com>
Acked-by: NMatthew Garrett <mjg@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Pass the vga cursor state to the vt layer, ensuring that we don't hide
the cursor when the bootloader has deliberately disabled it.
Signed-off-by: NMatthew Garrett <mjg@redhat.com>
LKML-Reference: <1258142222-16092-3-git-send-email-mjg@redhat.com>
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>

Add support for setting a global default for whether or not a visible
cursor should be enabled when creating VCs. The default will be to do so,
unless overridden by the user at boot time or by a driver.
Signed-off-by: NMatthew Garrett <mjg@redhat.com>
LKML-Reference: <1258143251-5818-1-git-send-email-mjg@redhat.com>
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>

fbcon makes the (reasonable) assumption that it only needs to program the
hardware once, when fbcon_init() is called for the foreground console.

This doesn't always play well with vgacon because vgacon_deinit() is only
doing its job when the last console it owns is closed (when switching from
vgacon to fbcon, that's usually *after* fbcon_init() has set the new
mode).

Depending on the hardware this can cause the wrong framebuffer location to
be scanned out (e.g.  reproduced on nv05 with the nouveau framebuffer
driver).
Signed-off-by: NFrancisco Jerez <currojerez@riseup.net>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

fix the following 'make includecheck' warning:

  drivers/video/console/vgacon.c: linux/slab.h is included more than once.
Signed-off-by: NJaswinder Singh Rajput <jaswinderrajput@gmail.com>
Cc: Martin Mares <mj@ucw.cz>
Cc: mchehab@infradead.org
Cc: Sam Ravnborg <sam@ravnborg.org>
LKML-Reference: <1247067624.4382.88.camel@ht.satnam>

Slab is initialized before the console subsystem so use the slab allocator in
vgacon_scrollback_startup().
Signed-off-by: NPekka Enberg <penberg@cs.helsinki.fi>

Uwe Geuder noted that he gets random bitmaps on a text console if he tried
to type extended characters (like the e acute).  For him everything above
unicode 0xa0 was corrupted.

After some digging there seems to be a little culprit in vgacon since the
beginning of ages (well git).  The function vgacon_font_get will store the
number of characters correctly in font->charcount but then calls to
vgacon_do_font_op(..., 0, 0).  Which means only the lower 256 characters
are actually stored to the fontdata.  The rest is left untouched.  So the
next time that saved data is used, the garbled font appears.  This happens
on every switch between text consoles.

Addresses https://bugs.launchpad.net/ubuntu/+source/linux/+bug/355057Signed-off-by: NStefan Bader <stefan.bader@canonical.com>
Tested-by: NUwe Geuder <ubuntuLp-ugeuder@sneakemail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

- (better, more, bigger ...) then -> (...) than
Signed-off-by: NFrederik Schwarzer <schwarzerf@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

Add mode setting support to the DRM layer.

This is a fairly big chunk of work that allows DRM drivers to provide
full output control and configuration capabilities to userspace.  It was
motivated by several factors:
  - the fb layer's APIs aren't suited for anything but simple
    configurations
  - coordination between the fb layer, DRM layer, and various userspace
    drivers is poor to non-existent (radeonfb excepted)
  - user level mode setting drivers makes displaying panic & oops
    messages more difficult
  - suspend/resume of graphics state is possible in many more
    configurations with kernel level support

This commit just adds the core DRM part of the mode setting APIs.
Driver specific commits using these new structure and APIs will follow.

Co-authors: Jesse Barnes <jbarnes@virtuousgeek.org>, Jakob Bornecrantz <jakob@tungstengraphics.com>
Contributors: Alan Hourihane <alanh@tungstengraphics.com>, Maarten Maathuis <madman2003@gmail.com>
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NEric Anholt <eric@anholt.net>
Signed-off-by: NDave Airlie <airlied@redhat.com>

There's no point in checking diff == c->vc_rows, because it can be true
only when count == 0, but we already checked that.  Additionally move
variables used only in one block to this block.
Signed-off-by: NMarcin Slusarz <marcin.slusarz@gmail.com>
Cc: Antonino Daplas <adaplas@gmail.com>
Acked-by: NKrzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Join multiple scr_memcpyw into 1-3 calls (usually 2).  (benchmarked
average speedup: 1%)
Signed-off-by: NMarcin Slusarz <marcin.slusarz@gmail.com>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Cc: Antonino Daplas <adaplas@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This reverts commit c9e587ab, and the
subsequent commits that fixed it up:

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

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

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

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

Alexander sent out a similar patch.
Requested-by: NAlan Cox <alan@lxorguk.ukuu.org.uk>
Tested-by: NJan Engelhardt <jengelh@medozas.de>
Cc: Alexander V. Lukyanov <lav@netis.ru>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

A command that causes a line feed while a background color is active,
such as

	perl -e 'print "x" x 60, "\e[44m", "x" x 40, "\e[0m\n"'
and
	perl -e 'print "x" x 40, "\e[44m\n", "x" x 40, "\e[0m\n"'

causes the line that was started as a result of the line feed to be completely
filled with the currently active background color instead of the default
color.

When scrolling, part of the current screen is memcpy'd/memmove'd to the new
region, and the new line(s) that will appear as a result are cleared using
memset.  However, the lines are cleared with vc->vc_video_erase_char, causing
them to be colored with the currently active background color.  This is
different from X11 terminal emulators which always paint the new lines with
the default background color (e.g.  `xterm -bg black`).

The clear operation (\e[1J and \e[2J) also use vc_video_erase_char, so a new
vc->vc_scrl_erase_char is introduced with contains the erase character used
for scrolling, which is built from vc->vc_def_color instead of vc->vc_color.
Signed-off-by: NJan Engelhardt <jengelh@computergmbh.de>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NMarcin Slusarz <marcin.slusarz@gmail.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Stop using magic macros for screen_info structure members.
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>