On Tue, Jul 02, 2024 at 08:35:08PM -0700, Andrii Nakryiko wrote: > On Tue, Jul 2, 2024 at 6:11 PM Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote: > > On Tue, Jul 02, 2024 at 05:06:14PM -0700, Andrii Nakryiko wrote: > > > In general, even with false positives, I think it's overwhelmingly > > > better to get correct entry stack trace 99.9% of the time, and in the > > > rest 0.01% cases it's fine having one extra bogus entry (but the rest > > > should still be correct), which should be easy for humans to recognize > > > and filter out, if necessary. > > > > Agreed, this is a definite improvement overall. > > Cool, I'll incorporate that into v3 and send it soon. > > > > > BTW, soon there will be support for sframes instead of frame pointers, > > at which point these checks should only be done for the frame pointer > > case. > > Nice, this is one of the reasons I've been thinking about asynchronous > stack trace capture in BPF (see [0] from recent LSF/MM). > [0] https://docs.google.com/presentation/d/1k10-HtK7pP5CMMa86dDCdLW55fHOut4co3Zs5akk0t4 I don't seem to have permission to open it. > Few questions, while we are at it. Does it mean that > perf_callchain_user() will support working from sleepable context and > will wait for data to be paged in? Is anyone already working on this? > Any pointers? I had a prototype here: https://lkml.kernel.org/lkml/cover.1699487758.git.jpoimboe@xxxxxxxxxx Hopefully I can get started on v2 soon. -- Josh
