On Mon, 19 Mar 2007, Paul Walmsley wrote: > > Also your solution cosumes significantly more memory (adding 16 bits to > > every hid_blacklist entry, duplicating the information into list that in > > most cases will stay intact and just mirror the contents of > > hid_blacklist). Not that it's a big issue, but still worth noting. > Right now there are about 160 entries in the hid_blacklist, so adding the mask > costs about 320 bytes extra. If that's a concern, a bigger issue is the > list_head added to each hid_blacklist entry - two pointers per entry - so, > 1280 bytes on a 32-bit machine (not counting the list head itself)
Hi Paul, What is your opinion on an alternative aproach - the hid_blacklist[] could stay intact, and the linked list in your patch will be used in a slightly different way - it will be used to define *exceptions* (which will be possible to be tuned in runtime) from hid_blacklist, rather than duplicating the information which is already present there? > If you'd rather not to have the mask code in the patch, I'll drop it. It's > only in there to remove the runtime-immutable special-case code for Wacom and > Codemercs devices, so it's not really useful for my specific concern. My opinion is that two exceptions (wacom and codemercs) can be easily handled by few lines of code, and extending the hid_blacklist[] just because of these two users looks maybe a little bit over-generalized to me. Anyway, I think that the possibility to modify the hid_blakclist[] in runtime could be useful. If you agree with the above and would care to redo the pathces in such way, I would be inclined to merge them. Thanks, -- Jiri Kosina
