Note: Yeah, I'm abusing the latest Debian
flamewar meme for some attention mongering. Sue me.
A few days ago, I have tested my cheap, crappy webcam I bought more than a year ago for the first time. Using the latest SN9C10x driver (which is already included in recent 2.6.1x kernels) worked fine, i.e., the USB webcam was recognized.
I tried running sonic-snap (site is currently down, try the Google Cache) in order to get snapshot images and/or videos off the webcam, which initially didn't work. But I soon found out what the cause of the problem was and created a trivial patch which fixed the problem for me.
Then, after I made myself look like an idiot by reporting a bug against a non-existing sonic-snap Debian package, I finally sent the patch to the upstream author. I'll probably ITP the package, though, as I might be using the webcam more often...
I also tested the webcam with Gnomemeeting for some videoconferencing fun, which didn't work at first either. After some stupid guessing and googling, I finally found out that you need to apt-get install libpt-plugins-v4l2, which is not installed by default (why?). Other than that (and apart from the really, really crappy image quality of the webcam), it worked really nice...
Here's the exact
lsusb output for Google to parse, other people might be searching for this info:
Bus 003 Device 003: ID 0c45:602a Microdia Meade ETX-105EC Camera
I didn't notice this for months. I was only able to use ca. 900 MB of my 1024 MB of RAM. It only happened accidentally that I read through my logs and noticed this:
Dec 26 19:33:37 localhost kernel: Warning only 896MB will be used.
Dec 26 19:33:37 localhost kernel: Use a HIGHMEM enabled kernel.
Dec 26 19:33:37 localhost kernel: 896MB LOWMEM available.
After setting the "High Memory Support" kernel option to 4GB and a quick kernel recompile, I can now use my full RAM. Nice, eh?
Update 2005-12-31: Thanks a lot to all the people pointing out (via comments in the blog and emails) that there's a better way to do this!
Some interesting Free Software releases:
Lots of stuff to try out...
Sometimes funny things happen. I spent several hours yesterday, trying to figure out why my laptop is responding so darned slow. I suspected it had something to do with the hard drive and I found out quite quickly that (U)DMA was disabled, hence the CPU had to do all the work. OK, no problem, I'll just do a
hdparm -c1 -d1 /dev/hda and everything will be fine. Or so I thought.
What I got was this:
setting using_dma to 1 (on)
HDIO_SET_DMA failed: Operation not permitted
using_dma = 0 (off)
Which means DMA could not be enabled. I noticed an error message in the output of
dmesg which seemed related:
ide0: Speed warnings UDMA 3/4/5 is not functional. Some people had the same problems because they were missing the correct option in the kernel (mine is
CONFIG_BLK_DEV_PIIX), but that was not my problem. After a few hours of googling and 6 or 7 kernel recompiles, I gave up and went to bed.
Now to the funny part: Today, John Choffee posted a comment about bashpodder in my blog. Curious as I am, I also visited his blog and in his "LinkFeed" box a tiny entry caught my attention: [PATCH] ich6m-pciid-piix.patch. Now guess what this patch (for Linux 2.6) does. It adds support for my specific type of IDE/(S)ATA controller, the "
Intel Corporation 82801FBM (ICH6M)". Patch, recompile kernel, reboot,
hdparm -c1 -d1 /dev/hda, bingo!
Here's the output of
hdparm -tT /dev/hda:
Before the patch:
Timing cached reads: 2468 MB in 2.00 seconds = 1232.95 MB/sec
Timing buffered disk reads: 8 MB in 3.84 seconds = 2.08 MB/sec
After the patch:
Timing cached reads: 2624 MB in 2.00 seconds = 1312.20 MB/sec
Timing buffered disk reads: 88 MB in 3.00 seconds = 29.33 MB/sec
Thanks John, you're my personal hero today.
Update 2006-03-01: The URL for the patch is broken. This one works.