> -----Original Message----- > From: James Courtier-Dutton [mailto:[EMAIL PROTECTED]] > Sent: Thursday, November 21, 2002 1:26 AM > To: Scott Parkerson > Cc: [EMAIL PROTECTED] > Subject: Re: [Alsa-devel] Oops? Linux 2.4.19 + snd-usb-audio > (ALSA 0.9.0-rc5) > > I think that will happen with most PCMCIA and USB devices in > linux.
That may be true. However, it does not happen with the stock 2.4.19 USB audio drivers. An unexpected hot removal does not cause an Oops to occur. The application will usually get a write() error, and that's it. This is much more desirable to have happen than an Oops, especially in a production environment where a user might "accidentally" pull the plug. > I think people will have to wait until kernel 2.5.x or above > for better > plug-n-unplug support. I disagree. I think the device drivers can be made to handle a case where the device's mapped addresses have "disappeared" out from under them. This kind of stuff is not necessarily operating system dependent (although there may be some OS-dependent ways to check to see if the hardware is "OK" before attempting to post a write or read. You can cause the same crap to happen in Windows unless your driver is carefully written. > Maybe I am wrong, and it is just a co-incidence that all the > devices I > have had bugs in the drivers. I don't believe you are wrong. I just don't believe that it's a priority for some developers to make sure that their drivers fail gracefully under this condition. I mean, it's Linux -- isn't the audience "people with a clue"? (sarcasm mine). Plus, most drivers are written for devices that can't be hot plugged under most conditions (i.e. PCI adapters). As for how all this relates to ALSA -- I don't know if it's up to ALSA to check to see if a device is still there or not. It might be up to the individual driver to handle it. Regards, Scott
smime.p7s
Description: application/pkcs7-signature
