linux-kernel - Re: linux-next: build failure after merge of the ecryptfs tree

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:   Tue, 14 May 2019 13:22:44 +1200
From:   Michael Schmitz <schmitzmic@...il.com>
To:     Stephen Rothwell <sfr@...b.auug.org.au>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>
Cc:     Tyler Hicks <tyhicks@...onical.com>,
        Linux Next Mailing List <linux-next@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        "David S. Miller" <davem@...emloft.net>
Subject: Re: linux-next: build failure after merge of the ecryptfs tree

Stephen,

I wasn't aware of the other asix module when submitting the phy driver. 
The phy module gets autoloaded based on the PHY ID, so there's no reason 
why it couldn't be renamed.

May I suggest ax88796b for the new module name?

Cheers,

     Michael



On 14/05/19 12:56 PM, Stephen Rothwell wrote:
> Hi all,
>
> [excessive quoting for new CC's]
>
> On Tue, 14 May 2019 09:40:53 +0900 Masahiro Yamada <yamada.masahiro@...ionext.com> wrote:
>> On Tue, May 14, 2019 at 9:16 AM Stephen Rothwell <sfr@...b.auug.org.au> wrote:
>>> I don't know why this suddenly appeared after mergeing the ecryptfs tree
>>> since nothin has changed in that tree for some time (and nothing in that
>>> tree seems relevant).
>>>
>>> After merging the ecryptfs tree, today's linux-next build (x86_64
>>> allmodconfig) failed like this:
>>>
>>> scripts/Makefile.modpost:112: target '.tmp_versions/asix.mod' doesn't match the target pattern
>>> scripts/Makefile.modpost:113: warning: overriding recipe for target '.tmp_versions/asix.mod'
>>> scripts/Makefile.modpost:100: warning: ignoring old recipe for target '.tmp_versions/asix.mod'
>>> scripts/Makefile.modpost:127: target '.tmp_versions/asix.mod' doesn't match the target pattern
>>> scripts/Makefile.modpost:128: warning: overriding recipe for target '.tmp_versions/asix.mod'
>>> scripts/Makefile.modpost:113: warning: ignoring old recipe for target '.tmp_versions/asix.mod'
>>> make[2]: Circular .tmp_versions/asix.mod <- __modpost dependency dropped.
>>> Binary file .tmp_versions/asix.mod matches: No such file or directory
>>> make[2]: *** [scripts/Makefile.modpost:91: __modpost] Error 1
>>> make[1]: *** [Makefile:1290: modules] Error 2
>>>
>>> The only clue I can see is that asix.o gets built in two separate
>>> directories (drivers/net/{phy,usb}).
>> Module name should be unique.
>>
>> If both are compiled as a module,
>> they have the same module names:
>>
>> drivers/net/phy/asix.ko
>> drivers/net/usb/asix.ko
>>
>> If you see .tmp_version directory,
>> you will see asix.mod
>>
>> Perhaps, one overwrote the other,
>> or it already got broken somehow.
> So, the latter of these drivers (drivers/net/phy/asix.c) was added in
> v4.18-rc1 by commit
>
>    31dd83b96641 ("net-next: phy: new Asix Electronics PHY driver")
>
> If we can't have 2 modules with the same base name, is it too late to
> change its name?
>
> I am sort of suprised that noone else has hit this in the past year.
>
>>> I have the following files in the object directory:
>>>
>>> ./.tmp_versions/asix.mod
>>> ./drivers/net/usb/asix.ko
>>> ./drivers/net/usb/asix.mod.o
>>> ./drivers/net/usb/asix.o
>>> ./drivers/net/usb/asix_common.o
>>> ./drivers/net/usb/asix_devices.o
>>> ./drivers/net/usb/.asix.ko.cmd
>>> ./drivers/net/usb/.asix.mod.o.cmd
>>> ./drivers/net/usb/.asix.o.cmd
>>> ./drivers/net/usb/asix.mod.c
>>> ./drivers/net/usb/.asix_common.o.cmd
>>> ./drivers/net/usb/.asix_devices.o.cmd
>>> ./drivers/net/phy/asix.ko
>>> ./drivers/net/phy/asix.o
>>> ./drivers/net/phy/.asix.ko.cmd
>>> ./drivers/net/phy/.asix.mod.o.cmd
>>> ./drivers/net/phy/asix.mod.o
>>> ./drivers/net/phy/asix.mod.c
>>> ./drivers/net/phy/.asix.o.cmd
>>>
>>> ./.tmp_versions/asix.mod
>>>
>>> Looks like this:
>>>
>>> ------------------------------------------
>>> drivers/net/phy/asix.ko
>>> drivers/net/phy/asix.o
>>>
>>>
>>> ------------------------------------------
>>>
>>> What you can't see above are the 256 NUL bytes at the end of the file
>>> (followed by a NL).
>>>
>>> This is from a -j 80 build.  Surely there is a race condition here if the
>>> file in .tmp_versions is only named for the base name of the module and
>>> we have 2 modules with the same base name.
>>>
>>> I removed that file and redid the build and it succeeded.
>>>
>>> Mind you, I have no itdea why this file was begin rebuilt, the merge
>>> only touched these files:
>>>
>>> fs/ecryptfs/crypto.c
>>> fs/ecryptfs/keystore.c
>>>
>>> Puzzled ... :-(