On Wed, Feb 12, 2020 at 10:31:25AM -0300, Arnaldo Carvalho de Melo wrote: > Em Wed, Feb 12, 2020 at 12:13:46PM +0100, Jiri Olsa escreveu: > > On Tue, Feb 11, 2020 at 04:32:23PM -0300, Arnaldo Carvalho de Melo wrote: > > > Historically vmlinux was preferred because it contains function sizes, > > > but with all these out of the blue symbols, we need to prefer starting > > > with /proc/kallsyms and, as we do now, continue getting updates via > > > PERF_RECORD_KSYMBOL. > > > > Humm, but then trampolines don't generate that, right? Or does it? If it > > > doesn't, then we will know about just the trampolines in place when the > > > record/top session starts, reparsing /proc/kallsyms periodically seems > > > excessive? > > > I plan to extend the KSYMBOL interface to contain trampolines/dispatcher > > data, > > That seems like the sensible, without looking too much at all the > details, to do, yes. > > > plus we could do some inteligent fallback to /proc/kallsyms in case > > vmlinux won't have anything > > At this point what would be the good reason to prefer vmlinux instead of > going straight to using /proc/kallsyms? symbol (with sizes) and code for dwarf unwind, processor trace jirka > > We have support for taking a snapshot of it at 'perf top' start, i.e. > right at the point we need to resolve a kernel symbol, then we get > PERF_RECORD_KSYMBOL for things that gets in place after that. > > And as well we save it to the build-id cache so that later, at 'perf > report/script' time we can resolve kernel symbols, etc. > > vmlinux is just what is in there right before boot, after that, for > quite some time, _lots_ of stuff happens :-) > > - Arnaldo >
