6 years agocatch up to kernel 3.3 master
Bernie Thompson [Wed, 25 Apr 2012 19:29:37 +0000]
catch up to kernel 3.3

6 years agobuild without mode helpers
Bernie Thompson [Tue, 20 Mar 2012 23:26:47 +0000]
build without mode helpers

6 years agosync with minor kernel changes
Bernie Thompson [Thu, 01 Mar 2012 22:04:23 +0000]
sync with minor kernel changes

6 years agofree urbs in USB disconnect()
Bernie Thompson [Sat, 11 Feb 2012 01:20:50 +0000]
free urbs in USB disconnect()

6 years agoadvise on early kernels without unlink_framebuffer
Bernie Thompson [Fri, 10 Feb 2012 02:41:58 +0000]
advise on early kernels without unlink_framebuffer

6 years agobe tolerant of sysfs attribute create failures
Bernie Thompson [Fri, 10 Feb 2012 01:14:45 +0000]
be tolerant of sysfs attribute create failures

6 years agodouble-check fb pointer before use
Bernie Thompson [Fri, 10 Feb 2012 01:00:06 +0000]
double-check fb pointer before use

6 years ago_MODULE support no longer sufficient for kernel dependencies
Bernie Thompson [Thu, 09 Feb 2012 23:51:13 +0000]
_MODULE support no longer sufficient for kernel dependencies

6 years agomove fb init to workitem, to avoid slowing or having errors interfere with usb enumeration
Bernie Thompson [Mon, 06 Feb 2012 05:21:58 +0000]
move fb init to workitem, to avoid slowing or having errors interfere with usb enumeration

6 years agoEven more flexible Makefile
Bernie Thompson [Sun, 05 Feb 2012 19:38:01 +0000]
Even more flexible Makefile

6 years agoMore flexible Makefile
Bernie Thompson [Sun, 05 Feb 2012 18:37:22 +0000]
More flexible Makefile

6 years agocopy module to updates for kernel versions with udlfb already present
Bernie Thompson [Fri, 23 Dec 2011 22:55:22 +0000]
copy module to updates for kernel versions with udlfb already present

6 years agofix backward compat for older kernels without vzalloc, etc.
Bernie Thompson [Mon, 03 Oct 2011 02:58:11 +0000]
fix backward compat for older kernels without vzalloc, etc.

6 years agoEnable fb_defio and console by default
Bernie Thompson [Sun, 21 Aug 2011 20:44:26 +0000]
Enable fb_defio and console by default
Module options allow disabling for situations where they cause problems
(e.g console conflict with --shared-vt, fb_defio overhead)

6 years agoBy default, udlfb allocates a 2nd buffer to shadow what's across
Stuart Hopkins [Mon, 01 Aug 2011 05:07:53 +0000]
By default, udlfb allocates a 2nd buffer to shadow what's across
the bus on the USB device. It can operate without this shadow,
but then it cannot tell which pixels have changed, and must send all.
Saves host memory, but worsens the USB 2.0 bus bottleneck.
This option allows users in very low memory situations (e.g.
bifferboard) to optionally turn off this shadow framebuffer.
Signed-off-by: Bernie Thompson
[bernie@plugable.com: renamed and reworked option]
Signed-off-by: Stuart Hopkins

6 years agomore catch-up with 3.0 kernel
Bernie Thompson [Mon, 01 Aug 2011 00:28:32 +0000]
more catch-up with 3.0 kernel

6 years agoSearch DL config descriptor for vendor descriptor if not found
Andrew Kephart [Mon, 01 Aug 2011 00:06:45 +0000]
Search DL config descriptor for vendor descriptor if not found
externally
For at least one DisplayLink device, the vendor-specific information can
be found in the config descriptor instead of as a separate, device-level
descriptor. This patch searches the current interface (of the current
config descriptor) for the DL vendor-specific descriptor.
Signed-off-by: Andrew Kephart
Signed-off-by: Bernie Thompson

6 years agosyncronize with cosmetic changes in 3.0 kernel
Bernie Thompson [Sun, 31 Jul 2011 23:28:33 +0000]
syncronize with cosmetic changes in 3.0 kernel

6 years agoudlfb was binding its framebuffer to its parent device, rather than its interface. Found and fix suggested by Kay Sievers
Bernie Thompson [Sat, 25 Jun 2011 04:37:40 +0000]
udlfb was binding its framebuffer to its parent device, rather than its interface. Found and fix suggested by Kay Sievers

6 years agoThis patch fixes a problem where a DisplayLink device autoselects a
Bill Katsak [Wed, 22 Jun 2011 21:41:20 +0000]
This patch fixes a problem where a DisplayLink device autoselects a
suboptimal framebuffer resolution. The problem occured with a
Plugable UGA-2K-A connected to a Samsung EX2220X display.
In this situation, the driver always allocates a 1600x1200
framebuffer, even after it indicates in its output that 1920x1080
is a valid mode. If my interpretation of the code is correct, the
problem was in the block that selects the best mode. The existing
code unset the FB_MISC_1ST_DETAIL flag if ANY of the modes is
determined to be invalid (on my adapter, 1680x1680 is invalid).
This in turn causes the fb_find_best_display() function to
disregard the first mode (1920x1080). The comment for the line
that unsets the flag implies that this should happen if we have
removed the top/best mode...however, this else block is not
qualified, and happens if we remove any mode. I simply put in a
condition so that we only unset this flag if the first mode is
invalid.
Signed-Off-By: Bernie Thompson

7 years agoFix __user casing in dlfb_ops_ioctl and a missing copy_from_user
Dr. David Alan Gilbert [Sun, 22 May 2011 21:02:22 +0000]
Fix __user casing in dlfb_ops_ioctl and a missing copy_from_user
Signed-off-by: Dr. David Alan Gilbert
Signed-off-by: Bernie Thompson

7 years agoFix bad boolean. Thanks to Dr. David Alan Gilbert for report
Bernie Thompson [Sun, 22 May 2011 20:22:06 +0000]
Fix bad boolean. Thanks to Dr. David Alan Gilbert for report

7 years agoAdd more reliable and comprehensive support for DPMS/FB_BLANK modes
Bernie Thompson [Thu, 23 Dec 2010 06:53:37 +0000]
Add more reliable and comprehensive support for DPMS/FB_BLANK modes

7 years agoTighten udlfb's matching to include interface subclass and protocol
Bernie Thompson [Thu, 23 Dec 2010 03:38:14 +0000]
Tighten udlfb's matching to include interface subclass and protocol
Allows future DisplayLink chips to increment subclass or protocol to signal
that the chip or composite interface is using an incompatible protocol.
Assumes all udlfb compatible devices will have subclass and protocol of zero.
This believed to be true for all current (2010) USB 2.0 era chips and firmware.

7 years agoRestrict match to vendor-specific interfaces of composite DisplayLink devices
Andrew Kephart [Thu, 23 Dec 2010 03:05:45 +0000]
Restrict match to vendor-specific interfaces of composite DisplayLink devices
For composite devices with DisplayLink vid, udlfb is matching the non-DL
interfaces (audio, HID, etc.) as well as the primary video interface
(vendor-specific class). This patch incorporates the interface class
as part of the device id match.
Signed-off-by: Bernie Thompson

7 years agoRevert, so we can split device ID match into two commits
Bernie Thompson [Thu, 23 Dec 2010 02:09:37 +0000]
Revert, so we can split device ID match into two commits
This reverts commit 6e336bcfd2e84e836bff44175dbe509919e799ee.

7 years agoTighten USB device ID matching to include class, subclass, and protocol
Andrew Kephart [Thu, 23 Dec 2010 01:46:18 +0000]
Tighten USB device ID matching to include class, subclass, and protocol
Solves two problems:
1) Composite DisplayLink devices with HID/audio/other functions will have
those functions matched against the appropriate driver
2) Future DisplayLink chips can increment subclass or protocol to signal
that a future chip or chip function is using an incompatible protocol
Believed compatible with all current USB 2.0 era chips and firmware
Signed-off-by: Bernie Thompson

7 years agoSwapped kmalloc paramters, allocating key structure for defio support.
Joe Perches [Thu, 23 Dec 2010 01:08:20 +0000]
Swapped kmalloc paramters, allocating key structure for defio support.
Amazing that we had this bug (only on defio path). Caught and fixed by Joe Perches by running tool during 2.6.37 cycle.
May explain and fix some of our hard-to-track-down defio problems!

7 years agoLinus flagged sysfs attributes that are world writable.
Greg Kroah-Hartman [Thu, 23 Dec 2010 00:59:54 +0000]
Linus flagged sysfs attributes that are world writable.
Was probably benign for reseting metrics, but better to take the change,
to sync with 2.6.37

7 years agoRestore missing ifdef to allow compile without fbdefio
Bernie Thompson [Mon, 04 Oct 2010 19:57:55 +0000]
Restore missing ifdef to allow compile without fbdefio

7 years agoSmall changes to sync with patches submitted to Linux kernel on August 18, 2010 (likely will be taken for 2.6.37)
Bernie Thompson [Thu, 19 Aug 2010 04:18:16 +0000]
Small changes to sync with patches submitted to Linux kernel on August 18, 2010 (likely will be taken for 2.6.37)

7 years agocheckpatch.pl cosmetic changes
Bernie Thompson [Tue, 17 Aug 2010 22:11:34 +0000]
checkpatch.pl cosmetic changes

7 years agoonly queue workitem to release urb sem when necessary (when using fb_defio) defio-2-6-35
Bernie Thompson [Tue, 17 Aug 2010 17:51:50 +0000]
only queue workitem to release urb sem when necessary (when using fb_defio)

7 years agofix null pointer deref
Bernie Thompson [Sun, 15 Aug 2010 23:56:43 +0000]
fix null pointer deref

7 years agoFixes to fb_defio support for 2.6.35 and later kernels
Bernie Thompson [Sun, 15 Aug 2010 23:11:48 +0000]
Fixes to fb_defio support for 2.6.35 and later kernels

7 years agocosmetic checkpatch.pl fixes
Bernie Thompson [Wed, 11 Aug 2010 17:38:39 +0000]
cosmetic checkpatch.pl fixes

7 years agoKeep defio enable/disable as dynamic, per framebuffer switch (rather than
Bernie Thompson [Wed, 11 Aug 2010 15:37:50 +0000]
Keep defio enable/disable as dynamic, per framebuffer switch (rather than
module global). Reverting prior change, before making more minor adjustments.
This reverts commit bbf46517cd9ee699c7e9f8a7f77ec64a1af534ee.

7 years agomove defio enablement to kernel module option, rather than more dynamic sysfs attribute
Bernie Thompson [Wed, 11 Aug 2010 04:42:36 +0000]
move defio enablement to kernel module option, rather than more dynamic sysfs attribute

7 years agoApply cosmetic changes from 2.6.35 kernel
Bernie Thompson [Mon, 09 Aug 2010 17:04:04 +0000]
Apply cosmetic changes from 2.6.35 kernel

7 years agoreally apply misplaced ifdef change
Bernie Thompson [Mon, 09 Aug 2010 17:03:41 +0000]
really apply misplaced ifdef change

7 years agoset appropriate permissions on kernel module parameter for sysfs
Bernie Thompson [Sun, 08 Aug 2010 22:18:51 +0000]
set appropriate permissions on kernel module parameter for sysfs

7 years agofix misplaced endif
Bernie Thompson [Sun, 08 Aug 2010 22:18:24 +0000]
fix misplaced endif

7 years agoremove metrics_misc sysfs attribute. Sysfs guidelines say only a single value per attribute.
Bernie Thompson [Sun, 08 Aug 2010 21:34:56 +0000]
remove metrics_misc sysfs attribute. Sysfs guidelines say only a single value per attribute.

7 years agoswitch to kernel version checks for bad aperture logic workaround, to get it working on 2.6.35
Bernie Thompson [Sat, 07 Aug 2010 00:54:11 +0000]
switch to kernel version checks for bad aperture logic workaround, to get it working on 2.6.35

7 years agoMake udlfb to handle usb_alloc_coherent API change for all kernel versions
Bernie Thompson [Sat, 07 Aug 2010 00:47:55 +0000]
Make udlfb to handle usb_alloc_coherent API change for all kernel versions

7 years agofix break from api change sfs: add struct file* to bin_attr callbacks
Bernie Thompson [Fri, 06 Aug 2010 22:16:08 +0000]
fix break from api change sfs: add struct file* to bin_attr callbacks

7 years agofix size_t to int warning
Bernie Thompson [Fri, 06 Aug 2010 21:47:54 +0000]
fix size_t to int warning

7 years agofbdev's aperature overlap algorithm changed in 2.6.35. Fix udlfb to accomodate
Bernie Thompson [Fri, 06 Aug 2010 21:28:59 +0000]
fbdev's aperature overlap algorithm changed in 2.6.35. Fix udlfb to accomodate

7 years agoAdd boolean module option console to enable allowing fbcon to connect. fbcon aggressively consumes the first framebuffer it finds, which isn't usually what users want. So make option disabled by default.
Bernie Thompson [Fri, 06 Aug 2010 19:56:40 +0000]
Add boolean module option console to enable allowing fbcon to connect. fbcon aggressively consumes the first framebuffer it finds, which isn't usually what users want. So make option disabled by default.

7 years agoFix shutdown hang introduced with commit e67ccc746f
Bernie Thompson [Thu, 08 Jul 2010 19:01:15 +0000]
Fix shutdown hang introduced with commit e67ccc746f

7 years agoSupport writing default/backup edid via /sys/class/graphics/fb?/edid
Bernie Thompson [Thu, 08 Jul 2010 18:33:28 +0000]
Support writing default/backup edid via /sys/class/graphics/fb?/edid
Required making udlfb much more flexible to let edid be dynamic
and changable, and then also to support changing/reallocating the
framebuffer size based on the (new) max mode supported by the monitor.
Writing a backup edid blob to the 'edid' file will cause udlfb to
attempt to re-read the edid from hardware.
If we still don't have a good edid, it will use the backup edid instead.
The intent is to 1) enable user-supplied edid
2) enable udev rules to backup the last known good edid
for a device, and then provide that edid as a backup for all future starts

7 years agoSet result to sucess when default VESA mode is found, in case of no EDID
Bernie Thompson [Thu, 01 Jul 2010 17:54:36 +0000]
Set result to sucess when default VESA mode is found, in case of no EDID

7 years agodlfb_free_urb_list is now the responsibility of dlfb_free
Bernie Thompson [Thu, 01 Jul 2010 16:36:48 +0000]
dlfb_free_urb_list is now the responsibility of dlfb_free

7 years agoFix incorrect/circular refcounts. Urbs are covered under main udlfb refcount
Bernie Thompson [Thu, 01 Jul 2010 16:21:58 +0000]
Fix incorrect/circular refcounts. Urbs are covered under main udlfb refcount

7 years agoFix fb_ioctl dereference of null fb_info
Bernie Thompson [Sun, 27 Jun 2010 03:45:00 +0000]
Fix fb_ioctl dereference of null fb_info
Rethinking of how we release and free objects.
Teardown path now treats usb and framebuffer teardown more independently.
The driver can now enter a "virtual" state when usb disconnect happens
where the virtual framebuffer can live long enough (at least) to allow
clients to release cleanly.
This also lays the groundwork for hotplug reconnection of a USB display
with a particular serial # to a framebuffer that already exists for it.

7 years agoMake some info/warn/err messages more appropriate level
Bernie Thompson [Tue, 22 Jun 2010 16:36:51 +0000]
Make some info/warn/err messages more appropriate level

7 years agolog adapter vendor, product,and serial info
Bernie Thompson [Mon, 21 Jun 2010 04:35:17 +0000]
log adapter vendor, product,and serial info

7 years agoFix error in endian logic in main rendering path
Christian Melki [Sun, 20 Jun 2010 18:41:14 +0000]
Fix error in endian logic in main rendering path
Signed-off-by: Bernie Thompson

7 years agoinclude slab.h as future location of kmalloc, etc.
Bernie Thompson [Wed, 09 Jun 2010 03:10:35 +0000]
include slab.h as future location of kmalloc, etc.

7 years agoreplace usb_buffer_alloc with usb_alloc_coherent
Bernie Thompson [Wed, 26 May 2010 23:09:41 +0000]
replace usb_buffer_alloc with usb_alloc_coherent

7 years agoFix mismatch pointer warning on 64bit
Bernie Thompson [Wed, 26 May 2010 22:35:07 +0000]
Fix mismatch pointer warning on 64bit

8 years agoPrevious commit (6 days ago) introduced bug incorrectly freeing pointer that had been incremented. Causes serious intermittent kernel corruption. Fixed to free original alloc
Bernie Thompson [Wed, 12 May 2010 22:08:34 +0000]
Previous commit (6 days ago) introduced bug incorrectly freeing pointer that had been incremented. Causes serious intermittent kernel corruption. Fixed to free original alloc

8 years agoAdd mode limits for DL-125/DL-165 on higher res monitors. vendor-desc
Bernie Thompson [Mon, 03 May 2010 15:51:07 +0000]
Add mode limits for DL-125/DL-165 on higher res monitors.
Solves problem of black screen when connecting to monitors with
a higher preferred/native resolution then the chip supports.
Uses code already in place to fall back to best match between chip and monitor.
However - won't yet work with DL X Server, which uses EDID, not mode list.
(Will see successful green screen, but X will be visible but messed up)
Done by implementing support for DisplayLink's vendor extentions to firmware
descriptors. We read the max_area characteristic to learn what is the last
mode the chip can support, before screen is blanked out.
Thanks to DisplayLink for supplying info on this.

8 years agoresolve conflicts with edid reliability merge
Bernie Thompson [Thu, 15 Apr 2010 22:40:59 +0000]
resolve conflicts with edid reliability merge

8 years agoMerge branch 'dpms'
Bernie Thompson [Thu, 15 Apr 2010 22:22:49 +0000]
Merge branch 'dpms'

8 years agoreturn function not implemented error where possible non-mmap
Bernie Thompson [Thu, 15 Apr 2010 22:21:51 +0000]
return function not implemented error where possible

8 years agorestore character device access functionality (fb_sys_ read and write)
Bernie Thompson [Tue, 13 Apr 2010 00:14:44 +0000]
restore character device access functionality (fb_sys_ read and write)

8 years agoadd support for dpms power off dpms
Bernie Thompson [Sun, 11 Apr 2010 15:58:19 +0000]
add support for dpms power off

8 years agofix warning on newer kernels - dont use stack buffer for usb_control_msg edid-reliability
Bernie Thompson [Sat, 20 Mar 2010 22:11:26 +0000]
fix warning on newer kernels - dont use stack buffer for usb_control_msg

8 years agoset aperture for firmware framebuffer aperture overlap checks. Needed on kernels > 2.6.31 when a firmware framebuffer (efi, vesa, open) has been registered before udlfb
Bernie Thompson [Wed, 17 Mar 2010 19:55:00 +0000]
set aperture for firmware framebuffer aperture overlap checks. Needed on kernels > 2.6.31 when a firmware framebuffer (efi, vesa, open) has been registered before udlfb

8 years agodisable defio in udlfb until issues are resolved
Bernie Thompson [Wed, 17 Mar 2010 19:45:18 +0000]
disable defio in udlfb until issues are resolved

8 years agochange sysfs edid_show to just return current edid no matter what, without requery
Bernie Thompson [Fri, 12 Mar 2010 20:58:46 +0000]
change sysfs edid_show to just return current edid no matter what, without requery

8 years agoon bad edid read, try several times before failing
Bernie Thompson [Fri, 12 Mar 2010 18:13:31 +0000]
on bad edid read, try several times before failing

8 years agoturn off defio, note why, and clarify related warnings
Bernie Thompson [Fri, 12 Mar 2010 16:31:20 +0000]
turn off defio, note why, and clarify related warnings

8 years agolog edid read errors more precisely. Don't overwrite good edid with bad
Bernie Thompson [Fri, 12 Mar 2010 16:23:49 +0000]
log edid read errors more precisely. Don't overwrite good edid with bad

8 years agoIt doesn't add value to re-query edid with ioctl asks for it, and may actually conflict with other activity with the device. Remove.
Bernie Thompson [Thu, 11 Mar 2010 17:38:05 +0000]
It doesn't add value to re-query edid with ioctl asks for it, and may actually conflict with other activity with the device. Remove.

8 years agomatch patches submitted for 2.6.34
Bernie Thompson [Fri, 19 Feb 2010 15:58:07 +0000]
match patches submitted for 2.6.34

8 years agoRemove unused alternative compression
Bernie Thompson [Fri, 12 Feb 2010 17:24:18 +0000]
Remove unused alternative compression
The chip's RLX mode just slightly better than using a combination of RAW
and RL commands. It's there in the version history for any future tests.

8 years agomake everything checkpatch.pl clean
Bernie Thompson [Fri, 12 Feb 2010 05:30:21 +0000]
make everything checkpatch.pl clean

8 years agostab at fixing 64-bit warnings
Bernie Thompson [Fri, 12 Feb 2010 04:31:39 +0000]
stab at fixing 64-bit warnings

8 years agomissing endif and cleanup
Bernie Thompson [Fri, 12 Feb 2010 04:01:12 +0000]
missing endif and cleanup

8 years agoAdd comment summarizing pros/cons of rlx
Bernie Thompson [Fri, 12 Feb 2010 04:00:37 +0000]
Add comment summarizing pros/cons of rlx

8 years agoMove compress function together with render_hline that uses it
Bernie Thompson [Fri, 12 Feb 2010 03:59:31 +0000]
Move compress function together with render_hline that uses it

8 years agoFix oops when shadow/backbuffer is not allocated
Bernie Thompson [Fri, 12 Feb 2010 03:58:53 +0000]
Fix oops when shadow/backbuffer is not allocated

8 years agoAdd alternate rle/raw rendering path similar to henrik's original
Bernie Thompson [Fri, 12 Feb 2010 01:09:59 +0000]
Add alternate rle/raw rendering path similar to henrik's original
Performance tests (x11perf, gtkperf, glxgears, video) appear to show
that RLX a wash. Since it's slightly more complex, it's tempting to
go back to the simpler rle-or-raw encoding.
rle/raw enabled by default.

8 years agoAdd best-effort prefetches for trim front & back, and urb cmd buffer
Bernie Thompson [Thu, 11 Feb 2010 00:16:28 +0000]
Add best-effort prefetches for trim front & back, and urb cmd buffer

8 years agoSome performance improvements to the rlx compress 16 function
Bernie Thompson [Wed, 10 Feb 2010 23:35:27 +0000]
Some performance improvements to the rlx compress 16 function
Mostly eliminating variables from inner loop, also adding prediction and
prefetch. More opportunities still left (e.g. prefetch urb buffer for write)

8 years agoGive user-mode clients the ability to avoid any defio overhead entirely
Bernie Thompson [Sat, 06 Feb 2010 18:12:52 +0000]
Give user-mode clients the ability to avoid any defio overhead entirely
By setting use_defio in sysfs to "0" before opening the framebuffer,
will cause udlfb to not initialize defio at all, using it's own mmap.
This avoids extra page faults if they're not needed.

8 years agoAdd use_defio sysfs option for controlling defio on/off. On by default.
Bernie Thompson [Sat, 06 Feb 2010 16:36:43 +0000]
Add use_defio sysfs option for controlling defio on/off. On by default.
With defio off, udlfb will stop processing defio pagefaults on framebuffer.
Will be set off by clients who more efficiently report damage via ioctls.
Examples
> cat /sys/class/graphics/fb0
1
> su -
> echo 0 > /sys/class/graphics/fb0

8 years agoMake things easier on people building module on older kernels
Bernie Thompson [Sat, 06 Feb 2010 00:41:57 +0000]
Make things easier on people building module on older kernels
driver can survive without FB_DEFIO and FB_SYS dependencies, but without
DEFIO standard fbdev apps (that use mmap without damage) won't work. And
without FB_SYS, fbcon won't work.

8 years agoMake framebuffer console actually work as expected
Bernie Thompson [Mon, 01 Feb 2010 19:39:14 +0000]
Make framebuffer console actually work as expected
udlfb logic was wrong on notifying damage if defio off
But, regardless, confirmed that touching the framebuffer
in kernel (from blit, fill, copy) isn't triggering faults
so handle_damage call is needed every time anyway.

8 years agoMake info flags reflect reality
Bernie Thompson [Mon, 01 Feb 2010 05:07:37 +0000]
Make info flags reflect reality

8 years agoFix problem where callers during teardown could get stuck waiting
Bernie Thompson [Fri, 29 Jan 2010 13:37:10 +0000]
Fix problem where callers during teardown could get stuck waiting
All grabs of the urb list counting semaphore now timeout. Downside
is pixels may be lost if timeout during normal processing. Fortunately
this is rare.

8 years agomove std_channel to be near and specific to probe reorganize
Bernie Thompson [Sat, 23 Jan 2010 01:52:07 +0000]
move std_channel to be near and specific to probe

8 years agoStructural cleanups - function names and groupings
Bernie Thompson [Sat, 23 Jan 2010 01:48:02 +0000]
Structural cleanups - function names and groupings
Header file just for public/client/hw definitions
dlfb_ops_ consistent prefix for fbdev callbacks
dlfb_usb_ consistent prefix for usb callbacks
A little cleaner grouping of functions

8 years agoSwitch last user of usb_bulk_msg to udlfb pre-alloced urb list
Bernie Thompson [Wed, 20 Jan 2010 23:30:43 +0000]
Switch last user of usb_bulk_msg to udlfb pre-alloced urb list

8 years agoGet framebuffer blanking interface working
Bernie Thompson [Wed, 13 Jan 2010 23:37:38 +0000]
Get framebuffer blanking interface working
E.g. echo 1 > /sys/class/graphics/fb0/blank
as su blanks the screen, turning of h and v sync
echo 0 restores the screen.
Two problems were solved:
usb_bulk_msg was silently failing (calling wrong?)
sending commands after blank automatically wakes DL chip

8 years agoAdded support for /sys/class/graphics/fb*/edid sysfs attribute
Bernie Thompson [Mon, 11 Jan 2010 18:12:39 +0000]
Added support for /sys/class/graphics/fb*/edid sysfs attribute
Reading this file re-queries EDID from the DisplayLink device.
And returns the 128-byte EDID block to the applicatoin.
Block can be parsed by "parse-edid < edid" (sudo apt-get install read-edid)
If monitor disconnects from DisplayLink device, re-reading this file once
the monitor is reconnected can reset state (but won't reset mode).

8 years agoMerge branch 'defio'
Bernie Thompson [Tue, 05 Jan 2010 21:19:23 +0000]
Merge branch 'defio'

8 years agoclarify build dependencies
Bernie Thompson [Tue, 05 Jan 2010 21:19:07 +0000]
clarify build dependencies