linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jiri Olsa <jolsa@redhat.com>
To: Milind Chabbi <chabbi.milind@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Namhyung Kim <namhyung@kernel.org>,
	linux-kernel@vger.kernel.org,
	Michael Kerrisk-manpages <mtk.manpages@gmail.com>,
	linux-man@vger.kernel.org, Michael Ellerman <mpe@ellerman.id.au>,
	Andi Kleen <ak@linux.intel.com>, Kan Liang <kan.liang@intel.com>,
	Hari Bathini <hbathini@linux.vnet.ibm.com>,
	Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>,
	Jin Yao <yao.jin@linux.intel.com>
Subject: Re: [PATCH] perf/core: fast breakpoint modification via _IOC_MODIFY_BREAKPOINT
Date: Thu, 9 Nov 2017 14:12:33 +0100	[thread overview]
Message-ID: <20171109131233.GA2942@krava> (raw)
In-Reply-To: <20171109074658.GC14419@krava>

On Thu, Nov 09, 2017 at 08:46:58AM +0100, Jiri Olsa wrote:

SNIP

> > Jirka,
> > 
> > I carefully looked at bp_cpuinfo[] and nr_slots[] data structures.
> > nr_slots[] is an array of length two (one slot of TYPE_INST and
> > another for TYPE_DATA).
> > The accounting "thinks" that there is one limit on the number of
> > instruction breakpoints and another limit on the number of data
> > breakpoints.
> > The assumption is clearly broken; for example, on x86 there exists a
> > limit on the *total* number of all breakpoints disregarding their kind
> > and the code has failed to capture this aspect.
> 
> there's the CONFIG_HAVE_MIXED_BREAKPOINTS_REGS that puts DATA and INST
> under one count on x86.. but that seems to be the enabled only for:
> 
> 	arch/sh/Kconfig:        select HAVE_MIXED_BREAKPOINTS_REGS
> 	arch/x86/Kconfig:       select HAVE_MIXED_BREAKPOINTS_REGS
> 
> > 
> > As such, modify_user_hw_breakpoint() makes no attempt to keep the
> > counts correct. Instead, it simply tries to change and install a new
> > breakpoint and fails if the hardware disallows.
> > This can lead to a situation where, say on x86, someone creates 4
> > TYPE_DATA breakpoints, then changes one of them to TYPE_INS via
> > modify_user_hw_breakpoint() and then releases the TYPE_INS breakpoint.
> > Since the accounting still thinks that there are four TYPE_DATA
> > breakpoints, it will disallow creating a new TYPE_DATA breakpoint,
> > although there is place for one TYPE_DATA breakpoint.
> > 
> > This convinces me that the problem and the solution are outside of
> > this current patch.
> > Do you agree?
> 
> I'll leave this decision to maintainer ;-) but seems better to fix
> the interface before we add any new dependent function calls

how about something like below (untested)

looks like there's no irq caller for modify_user_hw_breakpoint,
so we should be fine with locking nr_bp_mutex

jirka


---
diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c
index 3f8cb1e14588..f062b68399ea 100644
--- a/kernel/events/hw_breakpoint.c
+++ b/kernel/events/hw_breakpoint.c
@@ -448,6 +448,8 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	else
 		perf_event_disable(bp);
 
+	release_bp_slot(bp);
+
 	bp->attr.bp_addr = attr->bp_addr;
 	bp->attr.bp_type = attr->bp_type;
 	bp->attr.bp_len = attr->bp_len;
@@ -455,9 +457,9 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 	if (attr->disabled)
 		goto end;
 
-	err = validate_hw_breakpoint(bp);
+	err = reserve_bp_slot(bp);
 	if (!err)
-		perf_event_enable(bp);
+		err = validate_hw_breakpoint(bp);
 
 	if (err) {
 		bp->attr.bp_addr = old_addr;
@@ -469,6 +471,7 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
 		return err;
 	}
 
+	perf_event_enable(bp);
 end:
 	bp->attr.disabled = attr->disabled;
 

  reply	other threads:[~2017-11-09 13:12 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <CAMmz+Y=Py0dw63tuww+Oa4rWi_Hghhs3DHmNX=Tf1Yt_JH4O+Q@mail.gmail.com>
2017-11-06  9:23 ` [PATCH] perf/core: fast breakpoint modification via _IOC_MODIFY_BREAKPOINT Jiri Olsa
     [not found]   ` <CAMmz+YkB955Na6wOMmgqZX_TxqsBh86FiLi8EXmOrg1vwm-fGA@mail.gmail.com>
2017-11-08 14:15     ` Jiri Olsa
2017-11-08 15:02       ` Milind Chabbi
2017-11-08 15:12         ` Jiri Olsa
2017-11-08 15:51           ` Milind Chabbi
2017-11-08 15:57             ` Jiri Olsa
2017-11-08 16:59               ` Milind Chabbi
2017-11-09  7:52                 ` Jiri Olsa
2017-11-09 13:12                   ` Jiri Olsa [this message]
2017-11-09 18:59                     ` Milind Chabbi
2017-11-12 19:09                       ` Milind Chabbi
2017-11-13  7:46                         ` Jiri Olsa
2017-11-13  8:02                           ` Milind Chabbi
2017-11-26 19:31                             ` Jiri Olsa
2017-11-27  6:43                               ` [PATCH] perf/core: Enable the bp only if the .disable field is 0 Milind Chabbi
2017-11-27  6:50                                 ` Milind Chabbi
2017-11-27  9:25                                   ` Jiri Olsa
     [not found] <CAMmz+YnaoN3-7DN5WysQvhWNyGhM7_WDz5AQAnvP6FO_GMnMgw@mail.gmail.com>
2017-11-06 15:03 ` [PATCH] perf/core: fast breakpoint modification via _IOC_MODIFY_BREAKPOINT Arnaldo Carvalho de Melo
2017-11-06 22:09 Milind Chabbi
2017-11-06 23:16 ` Andi Kleen
2017-11-07  8:15   ` Peter Zijlstra
2017-11-07 17:09     ` Andi Kleen
2017-11-07  8:14 ` Peter Zijlstra
2017-11-07 15:43   ` Milind Chabbi
2017-11-07 17:24     ` Andi Kleen
2017-11-07 17:42       ` Milind Chabbi
2017-11-07 19:01         ` Peter Zijlstra
2017-11-07 19:31           ` Milind Chabbi
2017-11-08 13:35 ` kbuild test robot
2017-11-08 13:51 ` kbuild test robot

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=20171109131233.GA2942@krava \
    --to=jolsa@redhat.com \
    --cc=acme@kernel.org \
    --cc=ak@linux.intel.com \
    --cc=alexander.shishkin@linux.intel.com \
    --cc=chabbi.milind@gmail.com \
    --cc=hbathini@linux.vnet.ibm.com \
    --cc=kan.liang@intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-man@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=mpe@ellerman.id.au \
    --cc=mtk.manpages@gmail.com \
    --cc=namhyung@kernel.org \
    --cc=peterz@infradead.org \
    --cc=sukadev@linux.vnet.ibm.com \
    --cc=yao.jin@linux.intel.com \
    /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).