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]
Date:   Sun, 20 Nov 2016 13:56:16 +0100
From:   Tobias Jakobi <tjakobi@...h.uni-bielefeld.de>
To:     Russell King - ARM Linux <linux@...linux.org.uk>,
        Tobias Jakobi <tjakobi@...h.uni-bielefeld.de>
Cc:     linux-samsung-soc <linux-samsung-soc@...r.kernel.org>,
        viro@...iv.linux.org.uk,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        nicolas.pitre@...aro.org
Subject: Re: commit 4dd1837d7589f468ed109556513f476e7a7f9121 breaks build

Hey Russell,

thanks for the quick reply and looking into this!

Added Nicolas Pitre to Cc since the ksym trim stuff came from him.


Russell King - ARM Linux wrote:
> On Sun, Nov 20, 2016 at 12:43:38PM +0100, Tobias Jakobi wrote:
>> Hello Russell,
>>
>> Russell King - ARM Linux wrote:
>>> None of the autobuilders are finding this problem, and they build a
>>> wide range of configurations.  Maybe it's something specific to how
>>> you're building the kernel - can you give step by step instructions
>>> how to reproduce this please?
>> Just using plain make here, no cross-compiler or anything involved. gcc
>> is version 4.9.4.
> 
> Hmm.  The commit you mention was merged during the merge window, so
> anyone building 4.9-rc1 or later should be seeing this problem, but
> they aren't.  I've built many kernels post -rc1, and I'm not seeing
> the problem with my 4.7.4 cross compiler.
> 
>> Build platform is a Exynos4412 based ODROID-X2.
>>
>>
>> The complete log from the 'build modules' step:
>>>   Building modules, stage 2.
>>>   Kernel: arch/arm/boot/Image is ready
>>>   MODPOST 52 modules
>>>   XZKERN  arch/arm/boot/compressed/piggy_data
>>>   SHIPPED arch/arm/boot/compressed/lib1funcs.S
>>>   SHIPPED arch/arm/boot/compressed/ashldi3.S
>>>   SHIPPED arch/arm/boot/compressed/bswapsdi2.S
>>>   AS      arch/arm/boot/compressed/lib1funcs.o
>>>   AS      arch/arm/boot/compressed/ashldi3.o
>>>   AS      arch/arm/boot/compressed/bswapsdi2.o
>>> ERROR: "_set_bit" [sound/usb/snd-usbmidi-lib.ko] undefined!
>>> ERROR: "_test_and_set_bit" [sound/usb/snd-usbmidi-lib.ko] undefined!
>>> ERROR: "_clear_bit" [sound/usb/snd-usbmidi-lib.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [sound/usb/snd-usb-audio.ko] undefined!
>>> ERROR: "_set_bit" [sound/usb/snd-usb-audio.ko] undefined!
>>> ERROR: "_test_and_set_bit" [sound/usb/snd-usb-audio.ko] undefined!
>>> ERROR: "_clear_bit" [sound/usb/snd-usb-audio.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [sound/core/seq/snd-seq.ko] undefined!
>>> ERROR: "_test_and_set_bit" [sound/core/seq/snd-seq.ko] undefined!
>>> ERROR: "_set_bit" [net/bridge/bridge.ko] undefined!
>>> ERROR: "_clear_bit" [net/bridge/bridge.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [net/bluetooth/hidp/hidp.ko] undefined!
>>> ERROR: "_set_bit" [net/bluetooth/hidp/hidp.ko] undefined!
>>> ERROR: "_clear_bit" [net/bluetooth/hidp/hidp.ko] undefined!
>>> ERROR: "_set_bit" [net/bluetooth/bnep/bnep.ko] undefined!
>>> ERROR: "_test_and_set_bit" [net/bluetooth/bnep/bnep.ko] undefined!
>>> ERROR: "_clear_bit" [net/bluetooth/bnep/bnep.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [fs/fuse/fuse.ko] undefined!
>>> ERROR: "_set_bit" [fs/fuse/fuse.ko] undefined!
>>> ERROR: "_test_and_set_bit" [fs/fuse/fuse.ko] undefined!
>>> ERROR: "_clear_bit" [fs/fuse/fuse.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [fs/cifs/cifs.ko] undefined!
>>> ERROR: "_set_bit" [fs/cifs/cifs.ko] undefined!
>>> ERROR: "_test_and_set_bit" [fs/cifs/cifs.ko] undefined!
>>> ERROR: "_clear_bit" [fs/cifs/cifs.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [fs/btrfs/btrfs.ko] undefined!
>>> ERROR: "_set_bit" [fs/btrfs/btrfs.ko] undefined!
>>> ERROR: "_test_and_set_bit" [fs/btrfs/btrfs.ko] undefined!
>>> ERROR: "_clear_bit" [fs/btrfs/btrfs.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [drivers/usb/storage/usb-storage.ko] undefined!
>>> ERROR: "_set_bit" [drivers/usb/storage/usb-storage.ko] undefined!
>>> ERROR: "_clear_bit" [drivers/usb/storage/usb-storage.ko] undefined!
>>> ERROR: "_set_bit" [drivers/usb/gadget/libcomposite.ko] undefined!
>>> ERROR: "_set_bit" [drivers/usb/gadget/function/usb_f_rndis.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [drivers/usb/gadget/function/u_ether.ko] undefined!
>>> ERROR: "_test_and_set_bit" [drivers/usb/gadget/function/u_ether.ko] undefined!
>>> ERROR: "_clear_bit" [drivers/usb/gadget/function/u_ether.ko] undefined!
>>> ERROR: "_set_bit" [drivers/usb/gadget/function/u_ether.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [drivers/media/platform/s5p-mfc/s5p-mfc.ko] undefined!
>>> ERROR: "_set_bit" [drivers/media/platform/s5p-mfc/s5p-mfc.ko] undefined!
>>> ERROR: "_test_and_set_bit" [drivers/media/platform/s5p-mfc/s5p-mfc.ko] undefined!
>>> ERROR: "_clear_bit" [drivers/media/platform/s5p-mfc/s5p-mfc.ko] undefined!
>>> ERROR: "_clear_bit" [drivers/md/dm-crypt.ko] undefined!
>>> ERROR: "_set_bit" [drivers/md/dm-crypt.ko] undefined!
>>> ERROR: "_set_bit" [drivers/gpu/arm/mali/maligpu.ko] undefined!
>>> ERROR: "_clear_bit" [drivers/gpu/arm/mali/maligpu.ko] undefined!
>>> ERROR: "_test_and_clear_bit" [drivers/bluetooth/btusb.ko] undefined!
>>> ERROR: "_test_and_set_bit" [drivers/bluetooth/btusb.ko] undefined!
>>> ERROR: "_set_bit" [drivers/bluetooth/btusb.ko] undefined!
>>> ERROR: "_clear_bit" [drivers/bluetooth/btusb.ko] undefined!
>>> ERROR: "_set_bit" [drivers/bluetooth/btintel.ko] undefined!
>>> ERROR: "_set_bit" [drivers/bluetooth/btbcm.ko] undefined!
>>> scripts/Makefile.modpost:91: recipe for target '__modpost' failed
>>> make[1]: *** [__modpost] Error 1
>>> Makefile:1191: recipe for target 'modules' failed
>>> make: *** [modules] Error 2
>>
>>
>> This is the config I'm using for the build:
>> https://github.com/tobiasjakobi/odroid-environment/blob/master/sourcecode/system/vanilla-4.9-debug.conf
> 
> You have CONFIG_TRIM_UNUSED_KSYMS enabled - I wonder if that's the
> issue.  I've just enabled that here, and building I see:
Thanks, I had this enabled on purpose since I don't need any out-of-tree
drivers for this board.

I'll check if disabling TRIM_UNUSED_KSYMS also does the trick for me.


With best wishes,
Tobias


>   EXPORTS lib/lib-ksyms.o
>   LD      lib/built-in.o
>   CHK     include/generated/autoksyms.h
>   KSYMS   symbols: before=0, after=2413, changed=2413
>   UPD     include/generated/autoksyms.h
> 
> and then the build seems to restart from the very beginning, which seems
> rather fishy - it doubles the build time of the kernel.  The second
> build doesn't even hit any ccache entries from the first build.  Then we
> get through the rebuild, and I get:
> 
>   EXPORTS lib/lib-ksyms.o
>   LD      lib/built-in.o
>   CHK     include/generated/autoksyms.h
>   LD      vmlinux.o
> ...
>   LD      vmlinux
>   SORTEX  vmlinux
>   SYSMAP  System.map
>   Building modules, stage 2.
>   OBJCOPY arch/arm/boot/Image
>   Kernel: arch/arm/boot/Image is ready
>   AS      arch/arm/boot/compressed/head.o
>   LZO     arch/arm/boot/compressed/piggy_data
>   MODPOST 465 modules
> ERROR:   CC      arch/arm/boot/compressed/misc.o
> "_clear_bit" [net/sched/sch_teql.ko] undefined!
> ERROR: "_set_bit" [net/sched/sch_teql.ko] undefined!
> ERROR: "_set_bit" [net/sched/cls_u32.ko] undefined!
> ERROR: "_test_and_set_bit" [net/netfilter/nfnetlink_acct.ko] undefined!
> ERROR: "_clear_bit" [net/netfilter/nfnetlink_acct.ko] undefined!
> ERROR: "_test_and_clear_bit" [net/mac80211/mac80211.ko] undefined!
> ERROR: "_set_bit" [net/mac80211/mac80211.ko] undefined!
> ERROR: "_test_and_set_bit" [net/mac80211/mac80211.ko] undefined!
> ERROR: "_clear_bit" [net/mac80211/mac80211.ko] undefined!
> 
> So yes, it's caused by this ksyms trimming.
> 
> Now, the thing is that include/generated/autoksyms.h contains:
> 
> #define __KSYM__change_bit 1
> #define __KSYM__clear_bit 1
> #define __KSYM__set_bit 1
> #define __KSYM__test_and_change_bit 1
> #define __KSYM__test_and_clear_bit 1
> #define __KSYM__test_and_set_bit 1
> 
> amongst many other stuff - presumably this is a list of ksyms which
> should be included, rather than a list of what should be excluded.
> 
> I'm afraid that I never saw the patches which changed the way the ksyms
> stuff worked for the assembly files, and these patches have caused
> several problems for ARM.  I'm not aware of what the changes were, or
> how the damn exports now work, so I'm currently at a loss what to
> suggest at the moment, other than to disable CONFIG_TRIM_UNUSED_KSYMS in
> your configuration until someone can work out wtf is going on.
> 
> I guess I'm going to have to work out how this stuff now works - don't
> expect a solution quickly, especially as building with this option
> enabled is soo time consuming (like, half an hour per build.)
> 

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ