<!-- Some styling for better description lists --><style type='text/css'>dt { font-weight: bold;float: left;display:inline;margin-right: 1em} dd { display:block; margin-left: 2em}</style>

   ***: http_GK1wmSU has left
   larsc: <u>hverkuil</u>: do you understand what Dan Carpenter means with "shift wrap-around" in the adv7604 patch?
   hverkuil: https://stackoverflow.com/questions/7401888/why-doesnt-left-bit-shift-for-32-bit-integers-work-as-expected-when-used
   <br> However, I think that applies to signed ints, not to unsigned.
   <br> I think the current code works fine since page is unsigned.
   <br> No, it's wrong for unsigned as well
   <br> int main()
   <br> {
   <br>        unsigned a = 1;
   <br>        printf("%u %u %u\n", a &lt;&lt; 32, a &lt;&lt; 33, a &lt;&lt; 34);
   <br>        return 0;
   <br> }
   <br> $ ./x
   <br> 1 2 4
   <br> gcc gives me warnings too:
   <br> x.c: In function ‘main’:
   <br> x.c:6:25: warning: left shift count &gt;= width of type [-Wshift-count-overflow]
   <br>  printf("%u %u %u\n", a &lt;&lt; 32, a &lt;&lt; 33, a &lt;&lt; 34);
   <br>                         ^~
   <br> x.c:6:34: warning: left shift count &gt;= width of type [-Wshift-count-overflow]
   <br>  printf("%u %u %u\n", a &lt;&lt; 32, a &lt;&lt; 33, a &lt;&lt; 34);
   <br>                                  ^~
   <br> x.c:6:43: warning: left shift count &gt;= width of type [-Wshift-count-overflow]
   <br>  printf("%u %u %u\n", a &lt;&lt; 32, a &lt;&lt; 33, a &lt;&lt; 34);
   larsc: wow, did not expect that
   <br> thanks
   mchehab: hverkuil, larsc: I guess that, if you force the argument to be 64 bits e. g. with something like: a &lt;&lt; 32LL, it may work (although I didn't test)
   <br> I remember, however, that we did have a trouble in the past with things like that with v4l std (with is a typedef for u64)
   <br> on 32 bits, the shift still does the wrong thing
   <br> the BIT() macros has some special "magic" to solve those things
   hverkuil: with u64 you only postpone the problem until you hit a shift of 64 or more.
   <br> And I didn't see any special magic in the BIT macro.
   larsc: the issue seems to be that some archs just ignores the upper bits when the shift is done in hardware
   <br> that's like x &lt;&lt; (y &amp; 0x1f)
   mchehab: #define GENMASK_ULL(h, l) \
   <br> (((~0ULL) &lt;&lt; (l)) &amp; (~0ULL &gt;&gt; (BITS_PER_LONG_LONG - 1 - (h))))
   <br> for example, uses lots of typecasts to do the right thing
   <br> bbl
   hverkuil: <u>mchehab</u>: just for fun:
   <br> printf("%llx\n", GENMASK_ULL(65, 0));
   <br> fffffffffffffffc
   <br> :-)
   ***: http_GK1wmSU has left
   dersteffi: hey folks, I'm having trouble getting a video codec to work on an i.MX6 custom board. I posted on the list last monday: http://www.spinics.net/lists/linux-media/msg119119.html
   <br> can anybody help me? :)
   hverkuil: <u>dersteffi</u>: https://www.linuxtv.org/downloads/v4l-dvb-apis-new/v4l-drivers/imx.html
   dersteffi: I'll check it, thought I found everything documenting this. Thx for now!
   <br> <u>hverkuil</u>: I see this basicly recommends setting up the links via media-ctl. However, media-ctl always fails for me with 'Failed to enumerate /dev/media0 (-2)', which makes sense, there is no /dev/media* device in my tree. When I specify an existing /dev/video* device via -d to media-ctl, it fails with 'Failed to enumerate /dev/video5 (-25)'
   hverkuil: If you don't have a /dev/mediaX device, then something went very wrong. Is CONFIG_MEDIA_CONTROLLER set in your kernel config?
   dersteffi: it is.
   hverkuil: Check the /dev/media0 permissions
   <br> Perhaps they are wrong. (only the very latest systemd version 234 sets this correctly)
   <br> The udev rule should be:
   <br> SUBSYSTEM=="media", GROUP="video"
   dersteffi: hm, udev. my system currentliy runs on a buildroot image, busybox init, devtmpfs as device management
   hverkuil: If there is no udev then you probably have to make the node.
   <br> $ ls -l /dev/media0  
   <br> crw-rw---- 1 root video 242, 0 Aug  4 14:47 /dev/media0
   <br> ^^^^ the major/minor numbers to give to mknod
   dersteffi: ok, in paralell I'm building myimage with mdev.
   <br> was gonig to ask about major and minor :)
   <br> hm, now media-ctl fails with -6 (no such device or address).
   <br> # ls -l /dev/media0
   <br> crw-rw----    1 root     video     242,   0 Apr  1 15:50 /dev/media0
   <br> are you sure the major/minor numbers apply?
   hverkuil: Hmm, perhaps not. It's using alloc_chrdev_region(&amp;media_dev_t, 0, MEDIA_NUM_DEVICES, MEDIA_NAME);
   <br> I.e. requesting a major to be allocated.
   <br> You need udev in that case, I think.
   <br> (sorry about the confusion)
   dersteffi: no problem :)
   hverkuil: the major/minor is likely available from /sys somewhere.
   dersteffi: the build with mdev will take a few more minutes (udev is not available in buildroot, only eudev, which breaks my lvds panel for now). You also mentioned systemd will create the node?
   hverkuil: No, it's udev that produces it. But udev is part of systemd these days.
   dersteffi: hm, systemd 234 is available in buildroot, i can also try that. I will test both and report back for the record. Thanks for now, I've been stall on this for about a week and searching in the wrong corners.
   ***: prabhakarlad has left
   mchehab: <u>hverkuil</u>: well, 65 is too high for 64 bits :-)
   -: headless posts IMR driver v7 just right before his supper
   headless: 7th time the charm? :-)