From: Steven Rostedt <rostedt@goodmis.org>
To: Vince Weaver <vincent.weaver@maine.edu>
Cc: "H. Peter Anvin" <hpa@zytor.com>,
Peter Zijlstra <peterz@infradead.org>,
Linux Kernel <linux-kernel@vger.kernel.org>,
Ingo Molnar <mingo@redhat.com>
Subject: Re: perf_fuzzer compiled for x32 causes reboot
Date: Thu, 27 Feb 2014 21:57:26 -0500 [thread overview]
Message-ID: <20140227215726.7018c861@gandalf.local.home> (raw)
In-Reply-To: <alpine.DEB.2.10.1402272032140.6271@vincent-weaver-1.um.maine.edu>
On Thu, 27 Feb 2014 20:34:34 -0500 (EST)
Vince Weaver <vincent.weaver@maine.edu> wrote:
> > I would actually suggest we do the equivalent on i386 as well.
> >
> > Vince, could you try this patch as an experiment?
>
> OK with your patch applied it does not segfault.
>
Vince, Great! Can you remove Peter's patch, and try this one. It
removes the crud to save the cr2 from entry_64.S and makes both i386
and x86_64 do the same thing in regards to cr2 handling.
-- Steve
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index 1e96c36..937cb8d 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -1854,29 +1854,11 @@ end_repeat_nmi:
call save_paranoid
DEFAULT_FRAME 0
- /*
- * Save off the CR2 register. If we take a page fault in the NMI then
- * it could corrupt the CR2 value. If the NMI preempts a page fault
- * handler before it was able to read the CR2 register, and then the
- * NMI itself takes a page fault, the page fault that was preempted
- * will read the information from the NMI page fault and not the
- * origin fault. Save it off and restore it if it changes.
- * Use the r12 callee-saved register.
- */
- movq %cr2, %r12
-
/* paranoidentry do_nmi, 0; without TRACE_IRQS_OFF */
movq %rsp,%rdi
movq $-1,%rsi
call do_nmi
- /* Did the NMI take a page fault? Restore cr2 if it did */
- movq %cr2, %rcx
- cmpq %rcx, %r12
- je 1f
- movq %r12, %cr2
-1:
-
testl %ebx,%ebx /* swapgs needed? */
jnz nmi_restore
nmi_swapgs:
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
index 6fcb49c..f1a6294 100644
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -443,7 +443,6 @@ enum nmi_states {
NMI_LATCHED,
};
static DEFINE_PER_CPU(enum nmi_states, nmi_state);
-static DEFINE_PER_CPU(unsigned long, nmi_cr2);
#define nmi_nesting_preprocess(regs) \
do { \
@@ -452,14 +451,11 @@ static DEFINE_PER_CPU(unsigned long, nmi_cr2);
return; \
} \
this_cpu_write(nmi_state, NMI_EXECUTING); \
- this_cpu_write(nmi_cr2, read_cr2()); \
} while (0); \
nmi_restart:
#define nmi_nesting_postprocess() \
do { \
- if (unlikely(this_cpu_read(nmi_cr2) != read_cr2())) \
- write_cr2(this_cpu_read(nmi_cr2)); \
if (this_cpu_dec_return(nmi_state)) \
goto nmi_restart; \
} while (0)
@@ -512,8 +508,21 @@ static inline void nmi_nesting_postprocess(void)
dotraplinkage notrace __kprobes void
do_nmi(struct pt_regs *regs, long error_code)
{
+ unsigned long cr2;
+
nmi_nesting_preprocess(regs);
+ /*
+ * Save off the CR2 register. If we take a page fault in the NMI then
+ * it could corrupt the CR2 value. If the NMI preempts a page fault
+ * handler before it was able to read the CR2 register, and then the
+ * NMI itself takes a page fault, the page fault that was preempted
+ * will read the information from the NMI page fault and not the
+ * origin fault. Save it off and restore it if it changes.
+ * Use the r12 callee-saved register.
+ */
+ cr2 = read_cr2();
+
nmi_enter();
inc_irq_stat(__nmi_count);
@@ -523,6 +532,10 @@ do_nmi(struct pt_regs *regs, long error_code)
nmi_exit();
+ /* Reads are cheaper than writes */
+ if (unlikely(cr2 != read_cr2()))
+ write_cr2(cr2);
+
/* On i386, may loop back to preprocess */
nmi_nesting_postprocess();
}
next prev parent reply other threads:[~2014-02-28 2:57 UTC|newest]
Thread overview: 115+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-02-21 20:25 perf_fuzzer causes reboot Vince Weaver
2014-02-21 22:13 ` perf_fuzzer compiled for x32 " Vince Weaver
2014-02-21 22:34 ` Vince Weaver
2014-02-22 4:50 ` Vince Weaver
2014-02-22 5:03 ` H. Peter Anvin
2014-02-22 6:26 ` H. Peter Anvin
2014-02-23 5:18 ` Vince Weaver
2014-02-23 5:24 ` H. Peter Anvin
2014-02-23 6:07 ` H. Peter Anvin
2014-02-23 14:05 ` Vince Weaver
2014-02-24 3:02 ` Vince Weaver
2014-02-24 5:22 ` H. Peter Anvin
2014-02-24 15:35 ` Vince Weaver
2014-02-24 16:34 ` Vince Weaver
2014-02-24 16:47 ` H. Peter Anvin
2014-02-24 17:10 ` Vince Weaver
2014-02-24 17:25 ` Peter Zijlstra
2014-02-24 17:32 ` Vince Weaver
2014-02-24 17:40 ` H. Peter Anvin
2014-02-24 18:00 ` Vince Weaver
2014-02-24 18:07 ` Vince Weaver
2014-02-24 18:34 ` H. Peter Anvin
2014-02-24 19:13 ` Steven Rostedt
2014-02-24 19:15 ` H. Peter Anvin
2014-02-24 19:30 ` Peter Zijlstra
2014-02-24 19:32 ` Steven Rostedt
2014-02-25 3:49 ` H. Peter Anvin
2014-02-25 14:07 ` Vince Weaver
2014-02-25 14:34 ` H. Peter Anvin
2014-02-25 14:43 ` Steven Rostedt
2014-02-25 15:33 ` Vince Weaver
2014-02-26 15:06 ` Vince Weaver
2014-02-27 22:06 ` Vince Weaver
2014-02-27 22:31 ` Steven Rostedt
2014-02-27 22:52 ` H. Peter Anvin
2014-02-27 23:30 ` Steven Rostedt
2014-02-27 23:46 ` H. Peter Anvin
2014-02-28 1:00 ` Vince Weaver
2014-02-28 20:34 ` Paul E. McKenney
2014-02-28 20:47 ` Steven Rostedt
2014-02-28 20:54 ` Peter Zijlstra
2014-02-28 21:17 ` Paul E. McKenney
2014-02-28 21:27 ` Peter Zijlstra
2014-02-28 21:51 ` Paul E. McKenney
2014-02-28 21:55 ` Peter Zijlstra
2014-02-28 22:05 ` Steven Rostedt
2014-02-28 22:23 ` Paul E. McKenney
2014-02-28 1:34 ` Vince Weaver
2014-02-28 2:17 ` H. Peter Anvin
2014-02-28 2:57 ` Steven Rostedt [this message]
2014-02-28 11:11 ` Peter Zijlstra
2014-02-28 13:37 ` Steven Rostedt
2014-02-28 14:15 ` Vince Weaver
2014-02-28 14:23 ` Steven Rostedt
2014-02-28 15:07 ` Vince Weaver
2014-02-28 15:13 ` H. Peter Anvin
2014-02-28 15:40 ` Peter Zijlstra
2014-02-28 16:15 ` H. Peter Anvin
2014-02-28 16:29 ` Steven Rostedt
2014-02-28 19:33 ` [PATCH] x86: Rename copy_from_user_nmi() to copy_from_user_trace() Steven Rostedt
2014-02-28 20:46 ` Peter Zijlstra
2014-02-28 20:51 ` Steven Rostedt
2014-02-28 20:58 ` Peter Zijlstra
2014-02-28 21:01 ` Steven Rostedt
2014-02-28 21:17 ` Peter Zijlstra
2014-02-28 20:56 ` perf_fuzzer compiled for x32 causes reboot Peter Zijlstra
2014-02-28 21:06 ` Steven Rostedt
2014-03-01 9:16 ` Ingo Molnar
2014-03-01 9:50 ` Borislav Petkov
2014-03-01 16:50 ` H. Peter Anvin
2014-03-04 23:05 ` Borislav Petkov
2014-03-03 9:16 ` Peter Zijlstra
2014-02-28 20:55 ` Peter Zijlstra
2014-02-28 15:20 ` Steven Rostedt
2014-02-28 15:30 ` Steven Rostedt
2014-02-28 20:38 ` H. Peter Anvin
2014-02-28 20:46 ` Steven Rostedt
2014-02-28 21:18 ` Vince Weaver
2014-02-28 21:30 ` Steven Rostedt
2014-02-28 23:34 ` Vince Weaver
2014-03-01 0:43 ` H. Peter Anvin
2014-03-01 3:36 ` Steven Rostedt
2014-03-01 16:24 ` Andi Kleen
2014-03-02 15:34 ` Vince Weaver
2014-03-02 16:02 ` Vince Weaver
2014-02-28 9:39 ` Peter Zijlstra
2014-02-24 17:40 ` Peter Zijlstra
2014-02-24 17:41 ` Vince Weaver
2014-02-24 17:42 ` H. Peter Anvin
2014-02-24 17:52 ` H. Peter Anvin
2014-02-28 15:33 [PATCH] x86 trace: Fix page fault tracing bug Jiri Olsa
2014-02-28 15:47 ` Peter Zijlstra
2014-02-28 16:05 ` [PATCHv2] " Jiri Olsa
2014-02-28 16:11 ` H. Peter Anvin
2014-02-28 16:23 ` Steven Rostedt
2014-02-28 16:15 ` Steven Rostedt
2014-03-05 0:03 ` [tip:x86/urgent] x86, trace: Fix CR2 corruption when tracing page faults tip-bot for Jiri Olsa
2014-03-05 11:14 ` Peter Zijlstra
2014-03-05 12:20 ` Steven Rostedt
2014-03-05 12:25 ` Peter Zijlstra
2014-03-05 12:33 ` Steven Rostedt
2014-03-05 12:54 ` Peter Zijlstra
2014-03-05 13:02 ` Peter Zijlstra
2014-03-05 13:07 ` Peter Zijlstra
2014-03-05 12:36 ` Peter Zijlstra
2014-03-05 13:00 ` Steven Rostedt
2014-03-05 13:08 ` Peter Zijlstra
2014-03-05 21:37 ` H. Peter Anvin
2014-03-06 8:40 ` Peter Zijlstra
2014-03-06 11:02 ` Steven Rostedt
2014-03-06 14:53 ` [PATCH] x86: Further robustify CR2 handling vs tracing Peter Zijlstra
2014-03-07 23:07 ` [tip:x86/urgent] x86, trace: " tip-bot for Peter Zijlstra
2014-02-28 15:47 ` [PATCH] x86 trace: Fix page fault tracing bug Jiri Olsa
2014-02-28 16:00 ` Steven Rostedt
2014-02-28 16:01 ` Steven Rostedt
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=20140227215726.7018c861@gandalf.local.home \
--to=rostedt@goodmis.org \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=vincent.weaver@maine.edu \
/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).