XEN-4.3.0 tools-bugs [FLASK] [SEABIOS]

XEN-4.3.0 tools-bugs [FLASK] [SEABIOS]

[Vadim A. Misbakh-Soloviov]

[-- Attachment #1.1: Type: text/plain, Size: 2270 bytes --]

Hi! I've found few bugs in Xen-4.3.0:

FLASK part:

1) first of all, I'm discouraged, that flask policy builds without
FLASK_ENABLE=y. I guess, it is because I've SELinux in dom0, but,
anyway, since I've not enabled FLASK myself, I suggest that it will not
be build.

2) Moreover, that flask policy builds despite of I don't enabled flask,
somewhy it installs as "/boot" due to
> POLICY_LOADPATH = $(DESTDIR)/boot
(I checked it by changing value to boot3, and that policy was renamed too)

I guess, nobody get that issue due to less of people build xen/xen-tools
in sandbox, while Gentoo, for example, have both build and install in
sandboxed environment (so, there is no system /boot dir), and policy get
"/boot" name.


bundled SeaBIOS part:

3)

make -C seabios-dir all
  Working around non-functional -fwhole-program
make[5]: Entering directory
`/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
make rule.
  Build default config
#
# configuration written to
/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote/.config
#
make[5]: Leaving directory
`/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
  Working around non-functional -fwhole-program
  Compiling IASL out/acpi-dsdt.hex
make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
make rule.
  Compiling IASL out/ssdt-proc.hex
  Build Kconfig config file


That jobserver issues makes buildsystem continue build process  at
installation phase.

4) if user tries to build xen-tools with UTF-8 locale, he gets:

Version: rel-1.7.1-3-g3a28511-20130727_172824-node1
Traceback (most recent call last):
  File "./tools/layoutrom.py", line 630, in <module>
    main()
  File "./tools/layoutrom.py", line 607, in main
    info16 = parseObjDump(infile16, '16')
  File "./tools/layoutrom.py", line 538, in parseObjDump
    relocsection = sectionmap[sectionname]
KeyError: '.text.asm.out/../src/smp.c.77'

(there is some more info: https://bugs.gentoo.org/show_bug.cgi?id=445986 )





Is is any fixes for that issues already?


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 901 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [Xen-users] XEN-4.3.0 tools-bugs [FLASK] [SEABIOS]

[Ian Campbell]

On Sat, 2013-07-27 at 18:21 +0400, Vadim A. Misbakh-Soloviov wrote:
> Hi! I've found few bugs in Xen-4.3.0:

Thanks. For future reference it would be really useful to send these as
individual mails.

We also discourage cross-posting. I've moved the thread to -devel.

> FLASK part:
> 
> 1) first of all, I'm discouraged, that flask policy builds without
> FLASK_ENABLE=y. I guess, it is because I've SELinux in dom0, but,
> anyway, since I've not enabled FLASK myself, I suggest that it will not
> be build.

Building the policy seems harmless even if FLASK is not enabled? If
nothing else it gets us better build testing coverage

> 2) Moreover, that flask policy builds despite of I don't enabled flask,
> somewhy it installs as "/boot" due to
> > POLICY_LOADPATH = $(DESTDIR)/boot
> (I checked it by changing value to boot3, and that policy was renamed too)
> 
> I guess, nobody get that issue due to less of people build xen/xen-tools
> in sandbox, while Gentoo, for example, have both build and install in
> sandboxed environment (so, there is no system /boot dir), and policy get
> "/boot" name.

Yes, it seems that the install rule should gain an "$(INSTALL_DIR)
$(DESTDIR)$(POLICY_LOADPATH)" at the start. COuld you send such a patch
please. See http://wiki.xen.org/wiki/Submitting_Xen_Patches for guidance
if you need it.

> 
> 
> bundled SeaBIOS part:
> 
> 3)
> 
> make -C seabios-dir all
>   Working around non-functional -fwhole-program
> make[5]: Entering directory
> `/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
> make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
> make rule.
>   Build default config
> #
> # configuration written to
> /var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote/.config
> #
> make[5]: Leaving directory
> `/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
>   Working around non-functional -fwhole-program
>   Compiling IASL out/acpi-dsdt.hex
> make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
> make rule.
>   Compiling IASL out/ssdt-proc.hex
>   Build Kconfig config file
> 
> 
> That jobserver issues makes buildsystem continue build process  at
> installation phase.

I'm not sure I understand. Are you saying that this issue means some
stuff doesn't get built at "make build" time but only gets built at
"make install" time? That would be very strange. I'm not sure it would
necessarily be related to this jobserver thing.

> 4) if user tries to build xen-tools with UTF-8 locale,

Which locale specifically? What is your $LANG?

>  he gets:
> 
> Version: rel-1.7.1-3-g3a28511-20130727_172824-node1
> Traceback (most recent call last):
>   File "./tools/layoutrom.py", line 630, in <module>
>     main()
>   File "./tools/layoutrom.py", line 607, in main
>     info16 = parseObjDump(infile16, '16')
>   File "./tools/layoutrom.py", line 538, in parseObjDump
>     relocsection = sectionmap[sectionname]
> KeyError: '.text.asm.out/../src/smp.c.77'
> 
> (there is some more info: https://bugs.gentoo.org/show_bug.cgi?id=445986 )

This looks like a SeaBIOS issue. Have you checked the upstream SeaBIOS
tree for a fix? The right way to go about fixing this would probably be
to report upstream, can you do that? (please feel free to CC me and
xen-devel). That patch is probably to use LANG=C at various points in
the SeaBIOS makefile.

Ian.

Re: [PATCH] XEN-4.3.0 tools-bugs [FLASK] [SEABIOS]

[Vadim A. Misbakh-Soloviov]

[-- Attachment #1.1: Type: text/plain, Size: 4584 bytes --]

29.07.2013 13:48, Ian Campbell пишет:
> On Sat, 2013-07-27 at 18:21 +0400, Vadim A. Misbakh-Soloviov wrote:
>> Hi! I've found few bugs in Xen-4.3.0:
> 
> Thanks. For future reference it would be really useful to send these as
> individual mails.
> 
> We also discourage cross-posting. I've moved the thread to -devel.
> 
>> FLASK part:
>>
>> 1) first of all, I'm discouraged, that flask policy builds without
>> FLASK_ENABLE=y. I guess, it is because I've SELinux in dom0, but,
>> anyway, since I've not enabled FLASK myself, I suggest that it will not
>> be build.
> 
> Building the policy seems harmless even if FLASK is not enabled? If
> nothing else it gets us better build testing coverage

To be honest, not that it is harmless to build policy. Just I think, it
is unneded, if FLASK is disabled. And, anyway, main issue is that it
installs as "/boot"

>> 2) Moreover, that flask policy builds despite of I don't enabled flask,
>> somewhy it installs as "/boot" due to
>>> POLICY_LOADPATH = $(DESTDIR)/boot
>> (I checked it by changing value to boot3, and that policy was renamed too)
>>
>> I guess, nobody get that issue due to less of people build xen/xen-tools
>> in sandbox, while Gentoo, for example, have both build and install in
>> sandboxed environment (so, there is no system /boot dir), and policy get
>> "/boot" name.
> 
> Yes, it seems that the install rule should gain an "$(INSTALL_DIR)
> $(DESTDIR)$(POLICY_LOADPATH)" at the start. COuld you send such a patch
> please. See http://wiki.xen.org/wiki/Submitting_Xen_Patches for guidance
> if you need it.

Ok, just sent patches.


> 
>>
>>
>> bundled SeaBIOS part:
>>
>> 3)
>>
>> make -C seabios-dir all
>>   Working around non-functional -fwhole-program
>> make[5]: Entering directory
>> `/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
>> make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
>> make rule.
>>   Build default config
>> #
>> # configuration written to
>> /var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote/.config
>> #
>> make[5]: Leaving directory
>> `/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
>>   Working around non-functional -fwhole-program
>>   Compiling IASL out/acpi-dsdt.hex
>> make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
>> make rule.
>>   Compiling IASL out/ssdt-proc.hex
>>   Build Kconfig config file
>>
>>
>> That jobserver issues makes buildsystem continue build process  at
>> installation phase.
> 
> I'm not sure I understand. Are you saying that this issue means some
> stuff doesn't get built at "make build" time but only gets built at
> "make install" time? That would be very strange. I'm not sure it would
> necessarily be related to this jobserver thing.

Yes, I mean that some qemu related part builds at 'make install' time,
and I guess, it is due to that jobserver thing.

>> 4) if user tries to build xen-tools with UTF-8 locale,
> 
> Which locale specifically? What is your $LANG?
> 

ru_RU.UTF-8, but as mentioned in bug on Gentoo bugzilla, all the same on
italian UTF-8 locale.

>>  he gets:
>>
>> Version: rel-1.7.1-3-g3a28511-20130727_172824-node1
>> Traceback (most recent call last):
>>   File "./tools/layoutrom.py", line 630, in <module>
>>     main()
>>   File "./tools/layoutrom.py", line 607, in main
>>     info16 = parseObjDump(infile16, '16')
>>   File "./tools/layoutrom.py", line 538, in parseObjDump
>>     relocsection = sectionmap[sectionname]
>> KeyError: '.text.asm.out/../src/smp.c.77'
>>
>> (there is some more info: https://bugs.gentoo.org/show_bug.cgi?id=445986 )
> 
> This looks like a SeaBIOS issue. Have you checked the upstream SeaBIOS
> tree for a fix?

Does it counts, that system-wide seabios package (1.7.2.1 and 1.7.2.2
versions), builds and installs fine?

> The right way to go about fixing this would probably be
> to report upstream, can you do that? (please feel free to CC me and
> xen-devel).

Sure, I can, but I doubt, it give some result, since system-wide seabois
(from code.coreboot.org) installs just fine.

> That patch is probably to use LANG=C at various points in
> the SeaBIOS makefile.

Actually, it is already temporary workaround, but I doubt it is good
idea to do such things widely, so I think, it needs real fix of issue,
instead of hining it by locale change.

--

Best regards,
mva


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 901 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH] XEN-4.3.0 tools-bugs [FLASK] [SEABIOS]

[Ian Campbell]

On Mon, 2013-07-29 at 21:11 +0400, Vadim A. Misbakh-Soloviov wrote:
> > 
> > Yes, it seems that the install rule should gain an "$(INSTALL_DIR)
> > $(DESTDIR)$(POLICY_LOADPATH)" at the start. COuld you send such a patch
> > please. See http://wiki.xen.org/wiki/Submitting_Xen_Patches for guidance
> > if you need it.
> 
> Ok, just sent patches.

Thanks.

> >> bundled SeaBIOS part:
> >>
> >> 3)
> >>
> >> make -C seabios-dir all
> >>   Working around non-functional -fwhole-program
> >> make[5]: Entering directory
> >> `/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
> >> make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
> >> make rule.
> >>   Build default config
> >> #
> >> # configuration written to
> >> /var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote/.config
> >> #
> >> make[5]: Leaving directory
> >> `/var/tmp/portage/app-emulation/xen-tools-4.3.0/work/xen-4.3.0/tools/firmware/seabios-dir-remote'
> >>   Working around non-functional -fwhole-program
> >>   Compiling IASL out/acpi-dsdt.hex
> >> make[6]: warning: jobserver unavailable: using -j1.  Add `+' to parent
> >> make rule.
> >>   Compiling IASL out/ssdt-proc.hex
> >>   Build Kconfig config file
> >>
> >>
> >> That jobserver issues makes buildsystem continue build process  at
> >> installation phase.
> > 
> > I'm not sure I understand. Are you saying that this issue means some
> > stuff doesn't get built at "make build" time but only gets built at
> > "make install" time? That would be very strange. I'm not sure it would
> > necessarily be related to this jobserver thing.
> 
> Yes, I mean that some qemu related part builds at 'make install' time,
> and I guess, it is due to that jobserver thing.

All the lack of jobserver should do is linearise a parallel make (i..e
fallback to -j1), I don't think it should effect the dependency rules,
unless it is something really subtle. TBH I've no idea what might be
going on here.

> 
> >> 4) if user tries to build xen-tools with UTF-8 locale,
> > 
> > Which locale specifically? What is your $LANG?
> > 
> 
> ru_RU.UTF-8, but as mentioned in bug on Gentoo bugzilla, all the same on
> italian UTF-8 locale.
> 
> >>  he gets:
> >>
> >> Version: rel-1.7.1-3-g3a28511-20130727_172824-node1
> >> Traceback (most recent call last):
> >>   File "./tools/layoutrom.py", line 630, in <module>
> >>     main()
> >>   File "./tools/layoutrom.py", line 607, in main
> >>     info16 = parseObjDump(infile16, '16')
> >>   File "./tools/layoutrom.py", line 538, in parseObjDump
> >>     relocsection = sectionmap[sectionname]
> >> KeyError: '.text.asm.out/../src/smp.c.77'
> >>
> >> (there is some more info: https://bugs.gentoo.org/show_bug.cgi?id=445986 )
> > 
> > This looks like a SeaBIOS issue. Have you checked the upstream SeaBIOS
> > tree for a fix?
> 
> Does it counts, that system-wide seabios package (1.7.2.1 and 1.7.2.2
> versions), builds and installs fine?

Xen 4.3.0 contains 1.7.1. I can't see anything in that interval which
would fix this, but perhaps you can spot something? It also doesn't seem
like the sort of thing which would be .config specific.

> > The right way to go about fixing this would probably be
> > to report upstream, can you do that? (please feel free to CC me and
> > xen-devel).
> 
> Sure, I can, but I doubt, it give some result, since system-wide seabois
> (from code.coreboot.org) installs just fine.

Do you mean via the gentoo ebuild?
http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/sys-firmware/seabios/seabios-1.7.2.1.ebuild shows:

        src_compile() {
        	if ! use binary ; then
        		LANG=C emake out/bios.bin
        	fi
        }
        
In which case this seems like it is probably an upstream issue which
gentoo has worked around.

> 
> > That patch is probably to use LANG=C at various points in
> > the SeaBIOS makefile.
> 
> Actually, it is already temporary workaround, but I doubt it is good
> idea to do such things widely, so I think, it needs real fix of issue,
> instead of hining it by locale change.
> 
> --
> 
> Best regards,
> mva
>