linux-kernel - Re: [Patch -tip 0/4] Creation of the initcall tracer

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <48DA0A09.5080709@gmail.com>
Date:	Wed, 24 Sep 2008 10:36:09 +0100
From:	Frédéric Weisbecker <fweisbec@...il.com>
To:	Ingo Molnar <mingo@...e.hu>
CC:	linux-kernel@...r.kernel.org, Steven Rostedt <rostedt@...dmis.org>,
	Steven Noonan <steven@...inklabs.net>,
	Arjan van de Ven <arjan@...radead.org>,
	Thomas Gleixner <tglx@...utronix.de>,
	"H. Peter Anvin" <hpa@...or.com>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>
Subject: Re: [Patch -tip 0/4] Creation of the initcall tracer

Ingo Molnar a écrit :
> * Ingo Molnar <mingo@...e.hu> wrote:
> 
>> i've integrated tip/tracing/fastboot into tip/master and have started 
>> testing it. It passed the basic tests already so i've just pushed out 
>> the new tip/master. Please double-check that i have not messed up the 
>> rename or the integration somewhere.
> 
> FYI, -tip testing found a boot hang with CONFIG_BOOT_TRACER=y, on a 
> 32-bit T60 laptop. I've attached the config that provokes the hang. (you 
> need initcall_debug on the boot commandline)
> 
> The hang always happens after a specific initcall, shortly after tracing 
> is enabled. It just hangs indefinitely there, the last line displayed 
> was:
> 
>  [   0.004017] initcall ehci_hcd_init+0x0/0x80 returned 0 after 228 msecs
> 
> it should have printed this next:
> 
> [    0.808024] calling  ohci_hcd_mod_init+0x0/0x80 @ 10
> 
> but that line is never printed.
> 
> i've bisected the hang down to this specific commit:
> 
>  | 35df98dfe2b2503436a8cfda301657810df50a3a is first bad commit
>  | commit 35df98dfe2b2503436a8cfda301657810df50a3a
>  | Author: Frédéric Weisbecker <fweisbec@...il.com>
>  | Date:   Tue Sep 23 11:38:18 2008 +0100
>  |
>  |     tracing/ftrace: launch boot tracing after pre-smp initcalls
> 
> An idea: maybe the boot tracer is interacting with the ftrace self-tests 
> somehow? CONFIG_FTRACE_SELFTEST=y is enabled.
> 
> another thing i noticed is that the sysprof tracer was initialized and 
> self-tested shortly before the hang occured. Might be unrelated.
> 
> the bisection log is:
> 
>  # bad:  [6522eab5] Merge branch 'x86/uv'
>  # good: [699b81f5] Merge branch 'x86/iommu'
>  # bad:  [8007aae9] timers: fix build error in !oneshot case
>  # good: [7f29b23d] Merge branch 'core/rcu'
>  # bad:  [a54f17e3] manual merge of tracing/fastboot
>  # good: [bcbbe946] tracing/ftrace: make tracing suitable to run the b
>  # good: [d7095e5a] Merge branch 'fastboot'
>  # bad:  [35df98df] tracing/ftrace: launch boot tracing after pre-smp
>  # good: [42aaa6be] tracing/ftrace: give an entry on the config for bo
> 
> 	Ingo
> 
Ingo Molnar a écrit :
> * Ingo Molnar <mingo@...e.hu> wrote:
> 
>> i've integrated tip/tracing/fastboot into tip/master and have started 
>> testing it. It passed the basic tests already so i've just pushed out 
>> the new tip/master. Please double-check that i have not messed up the 
>> rename or the integration somewhere.
> 
> FYI, -tip testing found a boot hang with CONFIG_BOOT_TRACER=y, on a 
> 32-bit T60 laptop. I've attached the config that provokes the hang. (you 
> need initcall_debug on the boot commandline)
> 
> The hang always happens after a specific initcall, shortly after tracing 
> is enabled. It just hangs indefinitely there, the last line displayed 
> was:
> 
>  [   0.004017] initcall ehci_hcd_init+0x0/0x80 returned 0 after 228 msecs
> 
> it should have printed this next:
> 
> [    0.808024] calling  ohci_hcd_mod_init+0x0/0x80 @ 10
> 
> but that line is never printed.
> 
> i've bisected the hang down to this specific commit:
> 
>  | 35df98dfe2b2503436a8cfda301657810df50a3a is first bad commit
>  | commit 35df98dfe2b2503436a8cfda301657810df50a3a
>  | Author: Frédéric Weisbecker <fweisbec@...il.com>
>  | Date:   Tue Sep 23 11:38:18 2008 +0100
>  |
>  |     tracing/ftrace: launch boot tracing after pre-smp initcalls
> 
> An idea: maybe the boot tracer is interacting with the ftrace self-tests 
> somehow? CONFIG_FTRACE_SELFTEST=y is enabled.
> 
> another thing i noticed is that the sysprof tracer was initialized and 
> self-tested shortly before the hang occured. Might be unrelated.
> 
> the bisection log is:
> 
>  # bad:  [6522eab5] Merge branch 'x86/uv'
>  # good: [699b81f5] Merge branch 'x86/iommu'
>  # bad:  [8007aae9] timers: fix build error in !oneshot case
>  # good: [7f29b23d] Merge branch 'core/rcu'
>  # bad:  [a54f17e3] manual merge of tracing/fastboot
>  # good: [bcbbe946] tracing/ftrace: make tracing suitable to run the b
>  # good: [d7095e5a] Merge branch 'fastboot'
>  # bad:  [35df98df] tracing/ftrace: launch boot tracing after pre-smp
>  # good: [42aaa6be] tracing/ftrace: give an entry on the config for bo
> 
> 	Ingo
> 

Hehe! I was preparing a patch because I saw last night that the self-tests were reseting and touching the ring-buffer during the initcall tracing.
And now I can see the results of this problem....

Following is a patch that should fix it.
Thanks for this report!
---
[Patch -tip] Disable tracers self-tests when boot tracer is selected

The tracing engine resets the ring buffer and the tracers touch it too during self-tests. These self-tests happen during tracers registering
and work against boot tracing which is logging initcalls.
We have to disable tracing self-tests if the boot-tracer is selected.

Signed-off-by: Frederic Weisbecker <fweisbec@...il.com>
Reported-by: Ingo Molnar <mingo@...e.hu>
---

View attachment "selftests.diff" of type "text/plain" (944 bytes)