| 1 |
On Sun, 11 Feb 2007 07:47:29 -0500 |
| 2 |
Richard Freeman <rich@××××××××××××××.net> wrote: |
| 3 |
|
| 4 |
> -----BEGIN PGP SIGNED MESSAGE----- |
| 5 |
> Hash: SHA1 |
| 6 |
> |
| 7 |
> Hmm - after a reboot yesterday the order of my various /dev/video#'s |
| 8 |
> have changed. I'm guessing this was related to the recent baselayout |
| 9 |
> update (one of those reasons I always reboot within a few days after a |
| 10 |
> baselayout change - I want to make sure I can reboot at all lest the |
| 11 |
> system be down while I'm out of town). Or maybe it was due to this |
| 12 |
> monthly udev update cron job that seems to get triggered nowadays. |
| 13 |
> |
| 14 |
> Any ideas what might cause this to happen? |
| 15 |
> |
| 16 |
> It took a while to figure out what was happening - I run myth and all |
| 17 |
> it knew is that the device it was trying to access wasn't initializing |
| 18 |
> correctly. I figured the card had some issue, but it eventually |
| 19 |
> turned out that I was addressing the wrong card and doing it in the |
| 20 |
> wrong way. |
| 21 |
> |
| 22 |
> This seems to be one of those potential unix achilles-heels. Devices |
| 23 |
> just have those generic /dev/devicename mknods, but there isn't |
| 24 |
> anything that uniquely identifies a specific device. If these mknods |
| 25 |
> change order then everything gets confused. I guess a solution would |
| 26 |
> be to assign some kind of GUID to each device and use that to address |
| 27 |
> them - but that of course gets rid of the elegance of the |
| 28 |
> everything-is-a-file philosophy. Maybe create two links to the |
| 29 |
> device - one with a classic name, and another which is a GUID-based |
| 30 |
> filename, and software can use either one. |
| 31 |
|
| 32 |
You can usually use udev to rename the devices (or to provide |
| 33 |
alternative names). This is commonly done for network interfaces, for |
| 34 |
example. |
| 35 |
|
| 36 |
> I had a similar issue with a pair of USB serial ports I bought. Now, |
| 37 |
> this is probably not linux's fault - but the devices had NO uniquely |
| 38 |
> identifying info embedded in them as far as I could tell. So, I was |
| 39 |
> very nervous about them switching around their mknods after reboots, |
| 40 |
> after moving them around, etc. In the end I edited the udev |
| 41 |
> configuration to create a second mknod for each device that was |
| 42 |
> associated with the specific USB port they were plugged into (so much |
| 43 |
> for plug-and-play). My understanding is that windows has the same |
| 44 |
> problem with these sorts of devices - they work real great until you |
| 45 |
> have a bunch of them. |
| 46 |
> |
| 47 |
> Does anybody know if a generic solution exists to these sorts of |
| 48 |
> problems in linux, or how to mitigate these sorts of issues? With the |
| 49 |
> increased usage of USB I'd think that situations like this will only |
| 50 |
> come up more often... |
| 51 |
|
| 52 |
> -----BEGIN PGP SIGNATURE----- |
| 53 |
> Version: GnuPG v1.4.6 (GNU/Linux) |
| 54 |
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org |
| 55 |
> |
| 56 |
> iD8DBQFFzxBhG4/rWKZmVWkRAjDOAKDP4f+KaU/a+X7Cz79N9MGPLugX5QCgyUEp |
| 57 |
> 9lA35LPJOpIx0D8ty4XvVyk= |
| 58 |
> =tvvJ |
| 59 |
> -----END PGP SIGNATURE----- |
| 60 |
|
| 61 |
|
| 62 |
-- |
| 63 |
Kevin F. Quinn |
Archives