LWN: Comments on "A reworked BPF API" https://lwn.net/Articles/606089/ This is a special feed containing comments posted to the individual LWN article titled "A reworked BPF API". en-us Sat, 20 Sep 2025 08:18:31 +0000 Sat, 20 Sep 2025 08:18:31 +0000 https://www.rssboard.org/rss-specification lwn@lwn.net A reworked BPF API https://lwn.net/Articles/606787/ https://lwn.net/Articles/606787/ PaXTeam <div class="FormattedComment"> what is the meaning of a negative attr_len and insn_idx?<br> </div> Sun, 27 Jul 2014 19:34:48 +0000 A reworked BPF API https://lwn.net/Articles/606700/ https://lwn.net/Articles/606700/ adamgundy wasn't there a requirement that all new system calls should have a 'flags' argument? Fri, 25 Jul 2014 16:16:53 +0000 A reworked BPF API https://lwn.net/Articles/606601/ https://lwn.net/Articles/606601/ samroberts <div class="FormattedComment"> Isn't it more likely that the silence reflects a lack of usage?<br> </div> Thu, 24 Jul 2014 20:28:41 +0000 A reworked BPF API https://lwn.net/Articles/606387/ https://lwn.net/Articles/606387/ neilbrown <div class="FormattedComment"> That "fixup" mechanism seems vaguely reminiscent of something that "binder" does.<br> binder doesn't use fds, but it has similar numeric handles not unlike the "watch descriptors" that inotify uses.<br> <p> A "binder_buffer" (which holds a message) contains some data and then some "offsets".<br> The "offsets" are references into the data part of the buffer where "flat_binder_object"s are stored, which contain a type, some flags and an object identifier.<br> Based on the the type, some transformation is performed on the object identifier to convert between "remote" and "local" references. (see switch (fp-&gt;type) in binder_transaction()).<br> <p> This is much the same as converting between kernel-internal pointers and user-space fd numbers.<br> <p> </div> Thu, 24 Jul 2014 02:37:43 +0000