[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20181024141618.GB15106@kernel.org>
Date: Wed, 24 Oct 2018 11:16:18 -0300
From: Arnaldo Carvalho de Melo <acme@...nel.org>
To: Adrian Hunter <adrian.hunter@...el.com>
Cc: Jiri Olsa <jolsa@...hat.com>, Andi Kleen <ak@...ux.intel.com>,
David Miller <davem@...emloft.net>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH V2 17/19] perf scripts python: exported-sql-viewer.py:
Add All branches report
Em Wed, Oct 24, 2018 at 04:56:03PM +0300, Adrian Hunter escreveu:
> However I found kernel disassembly doesn't work, I think due to:
> commit edeb0c90df3581b821a764052d185df985f8b8dc
> perf tools: Stop fallbacking to kallsyms for vdso symbols lookup
> As I mentioned here:
> https://lore.kernel.org/lkml/4c682937-3cee-6974-0970-68610e13ad37@intel.com/T/#u
> So I will look at a fix for that also.
Ok.
But then that function receives both the addr and the cpumode, so its
callers should take into account the mixed ip/addr cpumodes, i.e. if we
get a sample with an ip in kernel and an addr in userspace, all we know
is that the ip is in the kernel, so we can either check the x86-specific
(well, specific to archs that share the address space for kernel and
userspace) machine__kernel_ip(addr) and pass a kernel cpu mode to
thread__find_addr(), otherwise ask for thread__find(addr, cpumode=user),
no?
Since this is x86 specific stuff, using machine__kernel_ip(addr) to
figure out the addr cpumode seems the way to go, right?
- Arnaldo
Powered by blists - more mailing lists