> > No, it's not an optimization for my particular setup, but more > > generally for CPU load. I thought that it might be better not to run > > advertisement code too frequently. But I guess that lower values > > should also be OK, e.g. 100 ms or 50 ms. There is one "issue" with > > that, though.... Now, the advertisement packet will be sent one > > interval after the advertisement was enabled. If that's indeed an > > issue, it can be fixed by calling the callback function in the moment > > when the timer is armed. > > I'd keep the initial logic of sending the advertisement immediately I've fixed that in the v2 patch. > Btw, make sure you run the testers to confirm you are not breaking any > kernel testers. I've run the test-runner to verify everything (--auto). The result seems to be no worse than without my patch. What do I mean by that? The case is that I was not able to run the test-runner on the vanilla master branch with 100% success rate. There is always some failure with my "setup". However, when checking only the tools/mgmt-tester test, the success rate seems to be higher than without it (but it might be just a fluke). I run it like this: "tools/test-runner -k ~/linux/arch/x86/boot/bzImage -- tools/mgmt-tester". Anyway, I can see one test always failing - "Read Exp Feature - Success (Index None)" - it happens with and without my patch. Also, I've verified that the advertising broadcast code was executed (by adding some prints to that function). To conclude that, I'm afraid that I don't know how to run the test suite properly... So, my verification might miss something. I would advise you to verify that patch with your setup as well (if you can). Regards, Arek
