Quoting Luiz Augusto von Dentz (2024-08-15 19:34:41) > On Thu, Aug 15, 2024 at 1:26 PM Anton Khirnov <anton@xxxxxxxxxxx> wrote: > > > Also, the advertisement monitor interface already does almost exactly > > what I want, except for the fact it forces me to receive advertisements > > from devices I'm not interested in. > > Yeah, I was gonna say that, but this is a vendor extension, I don't follow, what exactly is a vendor extension? IIUC in-controller handling of advertisement monitor or_patterns is a Microsoft extension, but what I want is just a nontrivial device whitelist, which should be standard. > not really following about forcing you to do anything, or you are > saying you want to connect rather than parse something on the > advertisement? It is a little bit hard to give you any advice if you > are not really explaining what is the use case, I wanted to avoid flooding this list with too many unnecessary details, but seems I went too far :) Sorry. I'm writing a data collector for bluetooth weight scales, which I'd like to run in the background and work automatically without user interaction. Advertisement reports for a weight scale contain ServiceData that indicates whether new measurements are available - when the collector sees such an advertisement it should connect to the scale and retrieve the measurements. So what the collector needs from the bluetooth stack is to monitor advertisement reports for a given list of devices (the scales it previously registered with). The advertisement monitor API almost gives me this, except it always uses an unfiltered filter policy (as per the block right above the done: label in hci_update_accept_list_sync()), which means advertisements from all devices are processed by the kernel and sent to userspace. This is not a fundamental problem, but it seems wasteful as the controller should be able to filter just the devices I want. > also if you are involving a daemon then perhaps you are not really > using bluetoothd? Just to be clear, I am using bluetoothd (via bleak), but AFAIU the issue is in the kernel code. Thanks, -- Anton Khirnov
