From: Josh Poimboeuf <jpoimboe@redhat.com>
To: Andy Lutomirski <luto@amacapital.net>
Cc: Jiri Kosina <jikos@kernel.org>, Ingo Molnar <mingo@redhat.com>,
X86 ML <x86@kernel.org>,
Heiko Carstens <heiko.carstens@de.ibm.com>,
"linux-s390@vger.kernel.org" <linux-s390@vger.kernel.org>,
live-patching@vger.kernel.org,
Michael Ellerman <mpe@ellerman.id.au>,
Chris J Arges <chris.j.arges@canonical.com>,
Jessica Yu <jeyu@redhat.com>,
linuxppc-dev@lists.ozlabs.org, Petr Mladek <pmladek@suse.com>,
Jiri Slaby <jslaby@suse.cz>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Vojtech Pavlik <vojtech@suse.com>,
Miroslav Benes <mbenes@suse.cz>,
Peter Zijlstra <peterz@infradead.org>
Subject: Re: [RFC PATCH v2 05/18] sched: add task flag for preempt IRQ tracking
Date: Wed, 22 Jun 2016 13:40:42 -0500 [thread overview]
Message-ID: <20160622184042.br7ov37pl6eydr3y@treble> (raw)
In-Reply-To: <CALCETrXR9iMO1veHREumqTOvQuLYYQFH_=VRhQ6oXusXi+NFNg@mail.gmail.com>
On Wed, Jun 22, 2016 at 11:26:21AM -0700, Andy Lutomirski wrote:
> On Wed, Jun 22, 2016 at 11:22 AM, Josh Poimboeuf <jpoimboe@redhat.com> wrote:
> > On Wed, Jun 22, 2016 at 10:59:23AM -0700, Andy Lutomirski wrote:
> >> > So I got a chance to look at this some more. I'm thinking that to make
> >> > this feature more consistently useful, we shouldn't only annotate
> >> > pt_regs frames for calls to handlers; other calls should be annotated as
> >> > well: preempt_schedule_irq, CALL_enter_from_user_mode,
> >> > prepare_exit_to_usermode, SWAPGS, TRACE_IRQS_OFF, DISABLE_INTERRUPTS,
> >> > etc. That way, the unwinder will always be able to find pt_regs from an
> >> > interrupt/exception, even if starting from one of these other calls.
> >> >
> >> > But then, things get ugly. You have to either setup and tear down the
> >> > frame for every possible call, or do a higher-level setup/teardown
> >> > across multiple calls, which invalidates several assumptions in the
> >> > entry code about the location of pt_regs on the stack.
> >> >
> >> > Also problematic is that several of the macros (like TRACE_IRQS_IRETQ)
> >> > make assumptions about the location of pt_regs. And they're used by
> >> > both syscall and interrupt code. So if we didn't create a frame pointer
> >> > header for syscalls, we'd basically need two versions of the macros: one
> >> > for irqs/exceptions and one for syscalls.
> >> >
> >> > So I think the cleanest way to handle this is to always allocate two
> >> > extra registers on the stack in ALLOC_PT_GPREGS_ON_STACK. Then all
> >> > entry code can assume that pt_regs is at a constant location, and all
> >> > the above problems go away. Another benefit is that we'd only need two
> >> > saves instead of three -- the pointer to pt_regs is no longer needed
> >> > since pt_regs is always immediately after the frame header.
> >> >
> >> > I worked up a patch to implement this -- see below. It writes the frame
> >> > pointer in all entry paths, including syscalls. This helps keep the
> >> > code simple.
> >> >
> >> > The downside is a small performance penalty: with getppid()-in-a-loop on
> >> > my laptop, the average syscall went from 52ns to 53ns, which is about a
> >> > 2% slowdown. But I doubt it would be measurable in a real-world
> >> > workload.
> >> >
> >> > It looks like about half the slowdown is due to the extra stack
> >> > allocation (which presumably adds a little d-cache pressure on the stack
> >> > memory) and the other half is due to the stack writes.
> >> >
> >> > I could remove the writes from the syscall path but it would only save
> >> > about half a ns, and it would make the code less robust. Plus it's nice
> >> > to have the consistency of having *all* pt_regs frames annotated.
> >>
> >> This is a bit messy, and I'm not really sure that the entry code
> >> should be have to operate under constraints like this. Also,
> >> convincing myself this works for NMI sounds unpleasant.
> >>
> >> Maybe we should go back to my idea of just listing the call sites in a table.
> >
> > So are you suggesting something like:
> >
> > .macro ENTRY_CALL func pt_regs_offset=0
> > call \func
> > 1: .pushsection .entry_calls, "a"
> > .long 1b - .
> > .long \pt_regs_offset
> > .popsection
> > .endm
> >
> > and then change every call in the entry code to ENTRY_CALL?
>
> Yes, exactly, modulo whether the section name is good. hpa is
> probably the authority on that.
Well, as you probably know, I don't really like peppering ENTRY_CALL
everywhere. :-/
Also I wonder how we could annotate the hypercalls, for example
DISABLE_INTERRUPTS actually wraps the call in a push/pop pair.
--
Josh
next prev parent reply other threads:[~2016-06-22 18:40 UTC|newest]
Thread overview: 118+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-04-28 20:44 [RFC PATCH v2 00/18] livepatch: hybrid consistency model Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 01/18] x86/asm/head: clean up initial stack variable Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 02/18] x86/asm/head: use a common function for starting CPUs Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 03/18] x86/asm/head: standardize the bottom of the stack for idle tasks Josh Poimboeuf
2016-04-29 18:46 ` Brian Gerst
2016-04-29 20:28 ` Josh Poimboeuf
2016-04-29 19:39 ` Andy Lutomirski
2016-04-29 20:50 ` Josh Poimboeuf
2016-04-29 21:38 ` Andy Lutomirski
2016-04-29 23:27 ` Josh Poimboeuf
2016-04-30 0:10 ` Andy Lutomirski
2016-04-28 20:44 ` [RFC PATCH v2 04/18] x86: move _stext marker before head code Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 05/18] sched: add task flag for preempt IRQ tracking Josh Poimboeuf
2016-04-29 18:06 ` Andy Lutomirski
2016-04-29 20:11 ` Josh Poimboeuf
2016-04-29 20:19 ` Andy Lutomirski
2016-04-29 20:27 ` Josh Poimboeuf
2016-04-29 20:32 ` Andy Lutomirski
2016-04-29 21:25 ` Josh Poimboeuf
2016-04-29 21:37 ` Andy Lutomirski
2016-04-29 22:11 ` Jiri Kosina
2016-04-29 22:57 ` Josh Poimboeuf
2016-04-30 0:09 ` Andy Lutomirski
2016-04-29 22:41 ` Josh Poimboeuf
2016-04-30 0:08 ` Andy Lutomirski
2016-05-02 13:52 ` Josh Poimboeuf
2016-05-02 15:52 ` Andy Lutomirski
2016-05-02 17:31 ` Josh Poimboeuf
2016-05-02 18:12 ` Andy Lutomirski
2016-05-02 18:34 ` Ingo Molnar
2016-05-02 19:44 ` Josh Poimboeuf
2016-05-02 19:54 ` Jiri Kosina
2016-05-02 20:00 ` Jiri Kosina
2016-05-03 0:39 ` Andy Lutomirski
2016-05-04 15:16 ` David Laight
2016-05-19 23:15 ` Josh Poimboeuf
2016-05-19 23:39 ` Andy Lutomirski
2016-05-20 14:05 ` Josh Poimboeuf
2016-05-20 15:41 ` Andy Lutomirski
2016-05-20 16:41 ` Josh Poimboeuf
2016-05-20 16:59 ` Andy Lutomirski
2016-05-20 17:49 ` Josh Poimboeuf
2016-05-23 23:02 ` Jiri Kosina
2016-05-24 1:42 ` Andy Lutomirski
2016-05-23 21:34 ` Andy Lutomirski
2016-05-24 2:28 ` Josh Poimboeuf
2016-05-24 3:52 ` Andy Lutomirski
2016-06-22 16:30 ` Josh Poimboeuf
2016-06-22 17:59 ` Andy Lutomirski
2016-06-22 18:22 ` Josh Poimboeuf
2016-06-22 18:26 ` Andy Lutomirski
2016-06-22 18:40 ` Josh Poimboeuf [this message]
2016-06-22 19:17 ` Andy Lutomirski
2016-06-23 16:19 ` Josh Poimboeuf
2016-06-23 16:35 ` Andy Lutomirski
2016-06-23 18:31 ` Josh Poimboeuf
2016-06-23 20:40 ` Josh Poimboeuf
2016-06-23 22:00 ` Andy Lutomirski
2016-06-23 0:09 ` Andy Lutomirski
2016-06-23 15:55 ` Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 06/18] x86: dump_trace() error handling Josh Poimboeuf
2016-04-29 13:45 ` Minfei Huang
2016-04-29 14:00 ` Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 07/18] stacktrace/x86: function for detecting reliable stack traces Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 08/18] livepatch: temporary stubs for klp_patch_pending() and klp_patch_task() Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 09/18] livepatch/x86: add TIF_PATCH_PENDING thread flag Josh Poimboeuf
2016-04-29 18:08 ` Andy Lutomirski
2016-04-29 20:18 ` Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 10/18] livepatch/powerpc: " Josh Poimboeuf
2016-05-03 9:07 ` Petr Mladek
2016-05-03 12:06 ` Miroslav Benes
2016-04-28 20:44 ` [RFC PATCH v2 11/18] livepatch/s390: reorganize TIF thread flag bits Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 12/18] livepatch/s390: add TIF_PATCH_PENDING thread flag Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 13/18] livepatch: separate enabled and patched states Josh Poimboeuf
2016-05-03 9:30 ` Petr Mladek
2016-05-03 13:48 ` Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 14/18] livepatch: remove unnecessary object loaded check Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 15/18] livepatch: move patching functions into patch.c Josh Poimboeuf
2016-05-03 9:39 ` Petr Mladek
2016-04-28 20:44 ` [RFC PATCH v2 16/18] livepatch: store function sizes Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 17/18] livepatch: change to a per-task consistency model Josh Poimboeuf
2016-05-04 8:42 ` Petr Mladek
2016-05-04 15:51 ` Josh Poimboeuf
2016-05-05 9:41 ` Miroslav Benes
2016-05-05 13:06 ` Petr Mladek
2016-05-04 12:39 ` barriers: was: " Petr Mladek
2016-05-04 13:53 ` Peter Zijlstra
2016-05-04 16:51 ` Josh Poimboeuf
2016-05-04 14:12 ` Petr Mladek
2016-05-04 17:25 ` Josh Poimboeuf
2016-05-05 11:21 ` Petr Mladek
2016-05-09 15:42 ` Miroslav Benes
2016-05-04 17:02 ` Josh Poimboeuf
2016-05-05 10:21 ` Petr Mladek
2016-05-04 14:48 ` klp_task_patch: " Petr Mladek
2016-05-04 14:56 ` Jiri Kosina
2016-05-04 17:57 ` Josh Poimboeuf
2016-05-05 11:57 ` Petr Mladek
2016-05-06 12:38 ` Josh Poimboeuf
2016-05-09 12:23 ` Petr Mladek
2016-05-16 18:12 ` Josh Poimboeuf
2016-05-18 13:12 ` Petr Mladek
2016-05-06 11:33 ` Petr Mladek
2016-05-06 12:44 ` Josh Poimboeuf
2016-05-09 9:41 ` Miroslav Benes
2016-05-16 17:27 ` Josh Poimboeuf
2016-05-10 11:39 ` Miroslav Benes
2016-05-17 22:53 ` Jessica Yu
2016-05-18 8:16 ` Jiri Kosina
2016-05-18 16:51 ` Josh Poimboeuf
2016-05-18 20:22 ` Jiri Kosina
2016-05-23 9:42 ` David Laight
2016-05-23 18:44 ` Jiri Kosina
2016-05-24 15:06 ` David Laight
2016-05-24 22:45 ` Jiri Kosina
2016-06-06 13:54 ` [RFC PATCH v2 17/18] " Petr Mladek
2016-06-06 14:29 ` Josh Poimboeuf
2016-04-28 20:44 ` [RFC PATCH v2 18/18] livepatch: add /proc/<pid>/patch_state Josh Poimboeuf
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=20160622184042.br7ov37pl6eydr3y@treble \
--to=jpoimboe@redhat.com \
--cc=chris.j.arges@canonical.com \
--cc=heiko.carstens@de.ibm.com \
--cc=jeyu@redhat.com \
--cc=jikos@kernel.org \
--cc=jslaby@suse.cz \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-s390@vger.kernel.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=live-patching@vger.kernel.org \
--cc=luto@amacapital.net \
--cc=mbenes@suse.cz \
--cc=mingo@redhat.com \
--cc=mpe@ellerman.id.au \
--cc=peterz@infradead.org \
--cc=pmladek@suse.com \
--cc=vojtech@suse.com \
--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 NNTP newsgroup(s).