From: Andrii Nakryiko <andrii@kernel.org>
To: linux-trace-kernel@vger.kernel.org, rostedt@goodmis.org,
mhiramat@kernel.org
Cc: x86@kernel.org, peterz@infradead.org, mingo@redhat.com,
tglx@linutronix.de, bpf@vger.kernel.org, rihams@fb.com,
linux-perf-users@vger.kernel.org,
Andrii Nakryiko <andrii@kernel.org>
Subject: [PATCH 0/4] Fix user stack traces captured from uprobes
Date: Wed, 8 May 2024 14:26:01 -0700 [thread overview]
Message-ID: <20240508212605.4012172-1-andrii@kernel.org> (raw)
This patch set reports two issues with captured stack traces.
First issue, fixed in patch #2, deals with fixing up uretprobe trampoline
addresses in captured stack trace. This issue happens when there are pending
return probes, for which kernel hijacks some of the return addresses on user
stacks. The code is matching those special uretprobe trampoline addresses with
the list of pending return probe instances and replaces them with actual
return addresses.
Second issue, which patch #3 is trying to fix with the help of heuristic, is
having to do with capturing user stack traces in entry uprobes. At the very
entrance to user function, frame pointer in rbp register is not yet setup, so
actual caller return address is still pointed to by rsp. Patch is using
a simple heuristic, looking for `push %rbp` instruction, to fetch this extra
direct caller return address, before proceeding to unwind the stack using rbp.
Consider this patch #3 an RFC, if there are better suggestions how this can be
solved, I'd be happy to hear that.
Patch #4 adds tests into BPF selftests, that validate that captured stack
traces at various points is what we expect to get. This patch, while being BPF
selftests, is isolated from any other BPF selftests changes and can go in
through non-BPF tree without the risk of merge conflicts.
Patches are based on latest linux-trace's probes/for-next branch.
Andrii Nakryiko (4):
uprobes: rename get_trampoline_vaddr() and make it global
perf,uprobes: fix user stack traces in the presence of pending
uretprobes
perf,x86: avoid missing caller address in stack traces captured in
uprobe
selftests/bpf: add test validating uprobe/uretprobe stack traces
arch/x86/events/core.c | 20 ++
include/linux/uprobes.h | 3 +
kernel/events/callchain.c | 42 +++-
kernel/events/uprobes.c | 17 +-
.../bpf/prog_tests/uretprobe_stack.c | 185 ++++++++++++++++++
.../selftests/bpf/progs/uretprobe_stack.c | 96 +++++++++
6 files changed, 359 insertions(+), 4 deletions(-)
create mode 100644 tools/testing/selftests/bpf/prog_tests/uretprobe_stack.c
create mode 100644 tools/testing/selftests/bpf/progs/uretprobe_stack.c
--
2.43.0
next reply other threads:[~2024-05-08 21:26 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-08 21:26 Andrii Nakryiko [this message]
2024-05-08 21:26 ` [PATCH 1/4] uprobes: rename get_trampoline_vaddr() and make it global Andrii Nakryiko
2024-05-08 21:26 ` [PATCH 2/4] perf,uprobes: fix user stack traces in the presence of pending uretprobes Andrii Nakryiko
2024-05-15 9:30 ` Peter Zijlstra
2024-05-15 14:32 ` Andrii Nakryiko
2024-05-20 15:20 ` Jiri Olsa
2024-05-20 23:56 ` Andrii Nakryiko
2024-05-08 21:26 ` [PATCH 3/4] perf,x86: avoid missing caller address in stack traces captured in uprobe Andrii Nakryiko
2024-05-20 15:20 ` Jiri Olsa
2024-05-08 21:26 ` [PATCH 4/4] selftests/bpf: add test validating uprobe/uretprobe stack traces Andrii Nakryiko
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240508212605.4012172-1-andrii@kernel.org \
--to=andrii@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=linux-perf-users@vger.kernel.org \
--cc=linux-trace-kernel@vger.kernel.org \
--cc=mhiramat@kernel.org \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=rihams@fb.com \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).