| Age | Commit message (Collapse) | Author |
|---|
|
work reliably yet, though
|
|
overlay video in one case
|
|
- restrict S_FBUF ioctl to CAP_SYS_ADMIN or CAP_SYS_RAWIO
- don't refuse setting S_FBUF when overlay is active
(changes become visible with the next update)
|
|
the device lock
- this fixes the "in_atomic():1, irqs_disabled():1" warnings in the syslog
|
|
interrupted by a streaming capture any more. So:
- revamp video_begin()/video_end(), saa7146_start_preview()/saa7146_stop_preview
- clean up a lot of mess around the usage of these functions
- allow overlay to be suspended by streaming capture again, if ressources are available
|
|
- deny activation of streaming capture if overlay is active, apps must disable overlay first
- fix resource handling for overlay
- don't try to stop streaming capture when it's not running
|
|
|
|
|
|
|
|
(bug catched by new resource management)
|
|
not work
- make sure to disable the right video dma at capture stop
|
|
severe memory damage can occur...
|
|
|
|
- fix problems caused by changes in ioctl definitions
|
|
- deny usage of planar formats for video overlay
- more intelligent checks in VIDIOC_STREAMON and VIDIOC_STREAMOFF
- fix read() capture while overlay is running
|
|
- add resource management functions
- lock various stuff against each other, mostly vbi capture vs. planar capture
formats, which use the same video dma
- remove old commented out code
- add new "flag" variable to saa7146_formats struct, store the information
if a capture format is planar or packed
- check for return codes at various places
|
|
cleared the various capture queue and has shut down capturing
(Thanks Gerd for explaining the pitfalls here!)
|
|
- more video/vbi open/release fixes
|
|
capture code on device closure
- change minimal picture size to 48x32 just like other drivers
|
|
sanity checks of the arguments (if a bad pagetable is created,
unpredictable thinks can happen. trust me... ;-)
- follow this change for the several uses
|
|
which were commented out anyway
- fix one typo
|
|
- janitorial fixes (mostly return code handling)
- spelling fixes in Kconfig files
- line breaks to re-formatting
|
|
|
|
V4L2_FIELD_ALTERNATE vs. V4L2_FIELD_INTERLACED captures
|
|
- fix pgtable_build_single, it should work for all kinds of
buffers (system memory (kernel/user) and gfx-memory)
- add/change some debug messages
|
|
|
|
to Adrian Bunk <bunk@fs.tum.de> for reporting this)
[V4L] - set debug verbosity to 0 for hexium_gemini and hexium_orion drivers
[V4L] - make hexium_gemini and hexium_orion drivers depdend on i2c module
[V4L] - make saa7146 module depend on hexium_gemini and hexium_orion drivers
[V4L] - let the saa7146 i2c bus report itself as I2C_ADAP_CLASS_TV_ANALOG
|
|
of a per-extension information. This is needed for the DVB-C cards,
which seem to have swapped field order and need to propagate some
video4linux specific stuff for the analog module to work.
|
|
Kannst Du das bitte ändern und die EXPORT_SYMBOL*()-Gesichten in ein saa7146_ksyms.c oder wie auch immer kapseln?
"saa7146_core.c" has all dependencies for the "saa7146_core.o" module
(saa7146_core.c and saa7146_i2c.c)
The dependencies for the "saa7146_vv.o" module have been moved from
the source files to "saa7146_vv_ksyms.c", the Makefiles habe been
changed accordingly.
Ok?
|
|
|
|
to different frames were sometimes mixed)
- first try to get alternating field capture running
- add/change some debug messages
|
|
there are no overlay informations.
|
|
- change saa7146_vv.h the way described earlier
- follow these changes in budget-av and hexium driver
|
|
Alan Cox reformatted the code to follow his coding style when
he fixed the merge bugs.
This patch now introduces these coding style changes, so
that we don't wipe out his changes with the next patchset.
|
|
- if the analog module is detected, the saa7113 is initialized and
some more v4l2 ioctls are available. you can use "xawtv" now to
switch between "dvb" and "analog" input. when you are one the
"analog" input, you can tune in analog channels with the cursor
keys via v4l2.
currently, this is a big hack -- tuning is not mutually exclusive,
so "szap" and v4l2 tuning can interfere with each other. the
demodulator address is hardcoded to 0x09.
Other changes:
- changed the saa7146 ioctl parameters, give out the data of the
current device open "fh" (=> "file handle"), not the pointer to the
device structure "dev". It is "dev = fh->dev".
Some stuff does not work:
- analog audio does not work. does the msp3400 need to be reprogrammed?
- one field is "missing", so the picture is very bad and
capturing does not work neither. this needs to be investigated by
looking at the programming the windows driver uses for the saa7146
|
|
to a separate "analog-2.4" directory. Main reasons:
- the drivers don't have anything to do with dvb 8-)
- the drivers are already in the kernel (besides the "Hexium" driver)
and don't change that much any more
- the analog drivers rely on the i2c subsystem, which was heavily modified
between 2.4 and 2.5. When making a new DVB patchset I have to work around
these differences every time, which is very annoying
|
|
use KBUILD_BASENAME for 2.4 instead
some changes to the i2c core to make it work with both 2.4 and late 2.5
kernels (not done yet -- only interesting for analog drivers)
- add MODULE_DEVICE_TABLE to all drivers, so that pci "hotplug" can detect
the devices automatically. Alan Cox says, that we should use this to load
the firmware (not done yet)
- fix the long standing "insmod/rmmod with multiple budget cards" bug -- it
was a double kfree() call...
|
|
makelinks: adjust the script for the upcoming kernel patches
driver.mxb: prevent the tuner module from probing unnecessary i2c addresses
getlinks: put symlinks for the header files to the right place(TM)
videodev.c: add a verbose error message just for the case that a user
has compiled the v4l module into the kernel
media/Kconfig: the dpc driver does not the tuner module
saa7146_hlp.c: honour the swapped field order for av7110 based cards
in capture mode, too
saa7146_video.c: add some verbose messages for the v4l2 controls
mxb.c: hopefully fix the detection and initialization of the sound arena module
|
|
- sync Kconfig and Makefiles with 2.5.65
- fix various things in saa7146 because of latest v4l changes in 2.5.65
- makelinks: removed the compat module from the 2.5.x build process
- makelinks: removed the linkage of ttusb and bt8xx -- enable this again for
the next patches, when ttusb is more stable and the bt8xx related patches
have been applied by Gerd Knorr
- dvb: removed the compat module from the 2.5.x build process, added #ifdefs
around the inclusion of compat.h. this was necessary because having compat
code in the newest 2.5.x kernel is awkward.
- dvb: move around of dvb_usercopy again. because I removed the compat module
from the 2.5.x build process, dvb_usercopy had to be moved to another file
- dvb: change MOD_*_USE_COUNT to module_get/module_put in various files,
but I'm not sure if I did this right. please review these changes.
- dvb: removed the crc32 code from dvb_demux.c, because it's in compat.c, too.
- dvb: removed some cruft from dvd_i2c.c, which is present in compat.h
- all files: apply small fixes to reflect the new location of some header
files
|
|
"dvb-kernel" tree into shape for direct inclusion into the
2.5.x kernel.
|
|
"nvrec"... (red/blue were swapped)
|
|
- V4L2_PIX_FMT_YUV422P: YUV 4:2:2 planar (Y-Cb-Cr)
- V4L2_PIX_FMT_YVU420: YVU 4:2:0 planar (Y-Cb-Cr)
- V4L2_PIX_FMT_YUV420: YUV 4:2:0 planar (Y-Cb-Cr)
Planar modes don't suffer from the saa7146 byte swap bug, so we
can deliver the formats needed by various video encoders, like "mp1e"
or "nvrec".
Added some notes to the TODO file.
|
|
- removed the patches, they are not needed any more. you can now
use a vanilla 2.5.55 kernel
- created a new backport for video-buf and videodev from the 2.5.55
sources
- necessary fixes to the saa7146 driver due to changes in 2.5.55
- the core budget functions are now provided by a module called
"budget-core.o". (otherwise static compilation is not possible...)
- removed #ifdef MODULE crap from all files
Misc.
- "xawtv" (using xv extension) resulted in wrong colors. the problem
is within "xawtv". Gerd Knorr told me, that it'll be fixed in the
next release.
I tested this version with 2.4.20, 2.5.55 (modules) and 2.5.55
(static).
|
|
- removed unused variables
- other code clenaups
- corrected H_PIXELS_PAL to 720 (not used for DVB-cards)
|
|
startup" bug.
|
|
a previoulsy running overlay window should be re-enabled.
|
|
(tested on 2.4.18 and 2.5.53 -- the budget driver is
untested, Holger will test it the next week)
saa7146:
- completely separated the saa7146 driver into a core-driver (saa7146.o) and
a video/vbi part (saa7146_vv.o). this allows you to build the budget driver
without av7110 and videodev bloat (see below)
The following thinks have been approved by Holger Waechtler:
av7110 (now ttpci):
- for 2.5.x: removed the av7110 directory and put everything into ttpci.
- removed all budget card support from this driver
- compile and build fixes for the saa7146 changes
ttpci-budget (new):
- for 2.5.x: created a new ttpci-budget directory, added a new driver for
the budget cards
mxb:
- compile and build fixes for the saa7146 changes
Changed the build and compile files for both 2.5.x and build-2.4 of course...
|
|
videodev.o dependency,
so we can later split up the av7110 driver into a av7110 and budget card driver.
I tested this on 2.4.18-3 (original RedHat 7.3 kernel) and 2.5.53.
Changes in detail:
saa7146:
- fixed the unitialised timer warnings
- split saa7146.o into saa7146.o and saa7146_vv.o:
all videodev (ie. v4l) related stuff is now encapsulated in saa7146_vv.o.
the extension (ie. mxb or dvb-ttpci) now has to make sure to register the v4l
devices via helper functions provided by saa7146_vv
mxb:
- changes to make the driver compile with the new saa7146 and saa7146_vv parts
- fix the vbi bypass issue for MXBs with saa7146, rev. 1 card
av7110:
- changes to make the driver compile with the new saa7146 and saa7146_vv parts
- some changes related to the upcoming av7110 <=> budget splitup, but they are
commented out
- changed the saa7146 PAL values to experimental values that look good to me, but
kept the old values as a comment. where are the original values from? there is
a big black border on most channels on the left side which annoyed me...
- changed from video/generic_usercopy() to dvb_usercopy() and put this into the
source code, not into compat.c
- moved the crc_32* stuff from compat.c to the source file where it is used
- now compat.[ch] was removed completely, all references were deleted from
all other files.
other:
- updated the README and TODO files
- changed "makelinks" to remove alps_bsru6 stuff from kernel when making the symlinks
- fixed the build files in "build-2.4"
- fixed all other files needed for 2.5.x kernel build
|
|
(ie. the av7110) handle the odd/even fields differently. I added a new
field to the extension, which holds special flags of the extension and the
device. I declared the MXB way of handling fields as "normal", so the
av7110 extension sets SAA7146_EXT_SWAP_ODD_EVEN here.
@Holger: I added some superfluous "struct saa7146_dev *dev" variables
back to some functions. These cause ugly warnings at compile time, I admit,
but they are necessary if you want to use the debug macros to trace a
saa7146 register for example. Please leave them in for now -- I need this
when I switch between old-DVB / old-MXB <=> new-DVB / new-MXB drivers and
try to hack out the differences.
|
|
READMEs accordingly, compile fixes for the mxb driver.
|
