After suspend/resume, camera light comes on.

[cjb]

In a normal boot, the camera light is off except for when the device has been opened by the camera activity. When we suspend and resume, the light comes on even though the device isn't open.

[cjb]

cafe_pci_resume() does a "cafe_ctlr_power_up(cam);" regardless of whether cam->users == 0, and the open path doesn't.

Here's a first shot at a patch that checks "users" before powering the camera back up, attached. Jon, could you take a look?

[kimquirk]

Would like to fix this on the B4 boards.

[cjb]

Closing this out, since the fix is working fine for us. It would be great to have the patch reviewed and sent upstream, though. Jon, could you take a look?

Some thoughts:

• the "if users > 0" logic could be moved into the powerup function itself, since the same test is done in both places the function's called from now.
• the mutex is probably unnecessary?

[corbet]

Replying to cjb:

Closing this out, since the fix is working fine for us. It would be great to have the patch reviewed and sent upstream, though. Jon, could you take a look?

Yes, I'll take a look and try it out. I have one more trip in my Period Of Insane Travel (a short hop to Tokyo) then I'll have more time for such things. That should still make it easily possible to get the patch into 2.6.23. When you have a moment, can you mail me a version with a signed-off-by on it?

[cjb]

Replying to corbet:

When you have a moment, can you mail me a version with a signed-off-by on it?

Done. Thanks!

[dilinger]

Replying to cjb:

Closing this out, since the fix is working fine for us. It would be great to have the patch reviewed and sent upstream, though. Jon, could you take a look? Some thoughts: * the "if users > 0" logic could be moved into the powerup function itself, since the same test is done in both places the function's called from now. * the mutex is probably unnecessary?

The power_down should really not be necessary; we should probably figure out why ctrl_init is turning the light on/enabling the camera. However, cjb's patch has been committed to olpc-2.6.

[jg]

Needs to to be reviewed and go upstream via John Corbet, but working in our builds.

[corbet]

[Better late than never...] Patch looks good to me; I'll be forwarding it upstream. Sorry for taking so long.