Re: 2.6.26-rc5-mm1

[Mariusz Kozlowski <m.kozlowski@tuxland.pl>]

Hello,

> Andrew Morton <akpm@linux-foundation.org> wrote on 10.06.2008 07:01:49:
> > On Tue, 10 Jun 2008 06:57:02 +0200 Mariusz Kozlowski <m.
> > kozlowski@tuxland.pl> wrote:
> > 
> > > Witam, 
> > > 
> > > > On Mon, 9 Jun 2008 21:14:54 +0200
> > > > Mariusz Kozlowski <m.kozlowski@tuxland.pl> wrote:
> > > > 
> > > > > Hello Balbir,
> > > > > 
> > > > > > Andrew Morton wrote:
> > > > > > > Temporarily at
> > > > > > > 
> > > > > > >   http://userweb.kernel.org/~akpm/2.6.26-rc5-mm1/
> > > > > > > 
> > > > > > 
> > > > > > I've hit a segfault, the last few lines on my console are
> > > > > > 
> > > > > > 
> > > > > > Testing -fstack-protector-all feature
> > > > > > registered taskstats version 1
> > > > > > debug: unmapping init memory ffffffff80c03000..ffffffff80dd8000
> > > > > > init[1]: segfault at 7fff701fe880 ip 7fff701fee5e sp 
> > 7fff7006e6d0 error 7
> > > > > > 
> > > > > > With absolutely no stack trace. I'll dig deeper.
> > > > > 
> > > > > Hey, I see something similar and I actually have a stack 
> > trace. Here it goes:
> > > > > 
> > > > > bash[498] segfault at ffffffff80868b58 ip ffffffffff600412 sp 
> > 7fffa3d010f0 error 7
> > > > > init[1] segfault at ffffffff80868b58 ip ffffffffff600412 sp 
> > 7fff9e97f640 error 7
> > > > > init[1] segfault at ffffffff80868b58 ip ffffffffff600412 sp 
> > 7fff9e97eed0 error 7
> > > > > Kernel panic - not syncing: Attemted to kill init!
> > > > > Pid 1, comm: init Not tainted 2.6.26-rc5-mm1 #1
> > > > > 
> > > > > Call Trace:
> > > > > [<ffffffff80254632>] panic+0xe2/0x260
> > > > > [<ffffffff802fa8ba>] ? __slab_free+0x10a/0x630
> > > > > [<ffffffff80265a8e>] ? __sigqueue_free+0x5e/0x70
> > > > > [<ffffffff802851eb>] ? trace_hardirqs_off+0x1b/0x30
> > > > > [<ffffffff802851eb>] ? trace_hardirqs_off+0x1b/0x30
> > > > > [<ffffffff80259b54>] do_exit+0xb84/0xc30
> > > > > [<ffffffff80259c5a>] do_group_exit+0x5a/0x110
> > > > > [<ffffffff8026a3b5>] get_signal_to_deliver+0x2c5/0x620
> > > > > [<ffffffff8020bb3b>] do_notify_resume+0x11b/0xd10
> > > > > [<ffffffff8028da5b>] ? trace_hardirqs_on+0x1b/0x30
> > > > > [<ffffffff805cd0f3>] ? _spin_unlock_irqrestore+0x93/0x130
> > > > > [<ffffffff8026865c>] ? force_sig_info+0x10c/0x130
> > > > > [<ffffffff8022fb9c>] ? force_sig_info_fault+0x2c/0x40
> > > > > [<ffffffff802dd7dd>] ? print_vma_addr+0x10d/0x1d0
> > > > > [<ffffffff805cbb67>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> > > > > [<ffffffff8028d8da>] ? trace_hardirqs_on_caller+0x15a/0x2c0
> > > > > [<ffffffff8020d4c9>] retint_signal+0x46/0x8d
> > > > > 
> > > > > This was copied manually so typos are possible.
> > > > > 
> > > > 
> > > > Thanks.  Could someone send a config please?  Or a bisection result 
> ;)
> > > 
> > > In my case it turns out to be gcov patches - in which I'm interested
> > > in to see (and play with) the tests coverage.
> > > 
> > > #
> > > # gcov
> > > #
> > > kernel-call-constructors.patch
> > > kernel-introduce-gcc_version_lower-macro.patch
> > > seq_file-add-function-to-write-binary-data.patch
> > > GOOD
> > > gcov-add-gcov-profiling-infrastructure.patch
> > > GOOD
> > > gcov-create-links-to-gcda-files-in-build-directory.patch
> > > gcov-architecture-specific-compile-flag-adjustments.patch
> > > BAD
> > > 
> > > I can not bisect between the last two due to build error. Config 
> > is attached.
> > > 
> > 
> > (cc Peter)

Sorry for the delay. Unfortunately I don't have as much time for this
as I'd like to.

> Thanks for the report. These look like the "known architecture problems"
> that I've hinted at in the gcov announcement post (I'm assuming this is
> x86_64 as I've seem similar reports in the past).
> 
> Possible reasons:
> 
> 1) initrd overwrites kernel: When kernel and initrd are loaded to static
> addresses, the oversized gcov kernel may overlap with the initrd load
> address. Solution: move initrd loading address.

Not using initrd.

> 2) out-of-memory: Kernel plus profiling code may just not fit into a
> minimal memory configuration any more. Solution: add memory.

2G is ok I assume.

> 3) write-protection of kernel code: gcc keeps profiling code and data
> close together in the .text section. Solution: any mechanism that
> protects .text against writes should be disabled when running a
> profiled kernel.

Not using it.

> 4) as of yet undiscovered incompatibilities between arch-dependent code
> and gcc's -fprofile-arcs option. Examples would be:
> 
>  * code which is run before memory access preparations were made
>  * hard coded section sizes
>  * relative address displacements which are out of range
> 
> Unfortunately I neither have access to a machine nor the skill to debug
> 4) myself, so if 1)-3) can be ruled out,

Yes they can be ruled out.

> I'd like to ask for more help 
> on this one:
> 
> First off, someone needs to track down the offending file(s). This is
> done by putting a line containing "GCOV := n" in all Makefiles below
> arch/x86_64 (or go one step further back and set
> CONFIG_GCOV_PROFILE_ALL=n). If my assumption is correct, then the
> kernel should boot fine afterwards. In that case, remove the lines
> again one-by-one, while compiling and booting after each change. If the
> problem can be narrowed down to a single Makefile, replace the single
> "GCOV := n" line with multiple "GCOV_file.o := n" lines, one for each
> generated object file. Then again, same approach as before: remove
> those lines, compile and boot until it breaks. Finally post your
> results.

After a few hours and tons of reboots I narrowed it down to
arch/x86/kernel/Makefile:

a) works
	obj-y                   += tsc_$(BITS).o io_delay.o rtc.o
	GCOV_tsc_$(BITS).o      := n
	#GCOV_io_delay.o        := n
	#GCOV_rtc.o     := n

b) doesn't work
	obj-y                   += tsc_$(BITS).o io_delay.o rtc.o
	#GCOV_tsc_$(BITS).o     := n
	#GCOV_io_delay.o        := n
	#GCOV_rtc.o     := n

and that points to arch/x86/kernel/tsc_64.c

> At this point we would need someone with x86_64 arch skills to look at
> the file and find out why this code is broken with "-fprofile-arcs"
> enabled (on s390 we discovered at least one actual code bug this way,
> so the analysis might just be of general use). Alternatively we can
> just keep these files from being profiled.


	Mariusz

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>