Re: sizeof - kernel modules

[drago01]

On Mon, Jun 15, 2015 at 7:38 PM, poma <pomidorabelisima(a)gmail.com&gt; wrote:
>  # du -hs /lib/modules/4.0.5-300.fc22.i686/ | awk '{print $1}'
>  42M
>
>  kernel-modules-4.0.5-300.fc22.i686.rpm  ...  17M
>
>  http://dl.fedoraproject.org/pub/fedora/linux/updates/testing/22/i386/k/
>  ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~
>
>  # du -hs /lib/modules/4.1.0-0.rc7.git0.1.fc23.x86_64/ | awk '{print $1}'
>  53M
>
>  kernel-modules-4.1.0-0.rc7.git0.1.fc23.x86_64.rpm  ...  18M
>
>  https://kojipkgs.fedoraproject.org/packages/kernel/4.1.0/0.rc7.git0.1.fc2...
>  ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~ ~~ ~ ~~ ~~~
>
>  # du -hs /lib/modules/4.1.0-0.rc7.git1.1.fc23.x86_64/ | awk '{print $1}'
>  439M
>
>  kernel-modules-4.1.0-0.rc7.git1.1.fc23.x86_64.rpm  ...  218M
>
>  http://dl.fedoraproject.org/pub/fedora/linux/development/rawhide/x86_64/o...
>  ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~ ~~ ~ ~~ ~~~ ~~~~
>
>  # du -hs /lib/modules/4.1.0-0.rc8.git0.1.fc23.x86_64/ | awk '{print $1}'
>  438M
>
>  kernel-modules-4.1.0-0.rc8.git0.1.fc23.x86_64.rpm  ...  218M
>
>  https://kojipkgs.fedoraproject.org/work/tasks/9420/10059420/
>  ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~
>
>
>  218 ÷ 18 = 12.111111111
>  439 ÷ 53 = 8.283018868
>
> >= 4.1.0-0.rc7.git1.1 cca. 10 x sizeof 
Those are debug builds.

sizeof - kernel modules

[Josh Boyer]

On Mon, Jun 15, 2015 at 1:45 PM, drago01 <drago01(a)gmail.com&gt; wrote:
>  On Mon, Jun 15, 2015 at 7:38 PM, poma
> <pomidorabelisima(a)gmail.com&gt; wrote:
> >
> > # du -hs /lib/modules/4.0.5-300.fc22.i686/ | awk '{print $1}'
> > 42M
> >
> > kernel-modules-4.0.5-300.fc22.i686.rpm  ...  17M
> >
> > http://dl.fedoraproject.org/pub/fedora/linux/updates/testing/22/i386/k/
> > ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~
> >
> > # du -hs /lib/modules/4.1.0-0.rc7.git0.1.fc23.x86_64/ | awk '{print $1}'
> > 53M
> >
> > kernel-modules-4.1.0-0.rc7.git0.1.fc23.x86_64.rpm  ...  18M
> >
> > https://kojipkgs.fedoraproject.org/packages/kernel/4.1.0/0.rc7.git0.1.fc2...
> > ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~ ~~ ~ ~~ ~~~
> >
> > # du -hs /lib/modules/4.1.0-0.rc7.git1.1.fc23.x86_64/ | awk '{print $1}'
> > 439M
> >
> > kernel-modules-4.1.0-0.rc7.git1.1.fc23.x86_64.rpm  ...  218M
> >
> >
> http://dl.fedoraproject.org/pub/fedora/linux/development/rawhide/x86_64/o...
> > ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~ ~~ ~ ~~ ~~~
> ~~~~
> >
> > # du -hs /lib/modules/4.1.0-0.rc8.git0.1.fc23.x86_64/ | awk '{print $1}'
> > 438M
> >
> > kernel-modules-4.1.0-0.rc8.git0.1.fc23.x86_64.rpm  ...  218M
> >
> > https://kojipkgs.fedoraproject.org/work/tasks/9420/10059420/
> > ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~
> >
> >
> > 218 ÷ 18 = 12.111111111
> > 439 ÷ 53 = 8.283018868
> >
> >>= 4.1.0-0.rc7.git1.1 cca. 10 x sizeof
>
>  Those are debug builds. 
They aren't actually.

It looks like something in the buildroot stopped stripping the kernel
modules between rc7.git0 and rc7.git1.  Both elf-utils and binutils
changed, so maybe something in there.  The same SRPM build on F22 has
the smaller size.

I haven't looked into it more than that.  If someone wants to poke at
this and figure out what broke, that would be excellent.  From a
kernel package perspective, there isn't anything that we did that
should cause this change in behavior that I'm aware of.

josh

sizeof - kernel modules

[Josh Stone]

CC elfutils-devel, this appears to be a regression.  I'm top-posting my
own findings, but I'll leave the kernel report below...

With elfutils-0.162, it appears that "strip" doesn't reduce the file
size when operating in-place (without -o).  This isn't particular to
funny kernel ELF -- I reproduced it on a local stap binary.

My full stap is 31MB.  With strip -o, I get a new 2.5MB file.  With
strip in-place, stap remains 31MB even though the ELF section data does
indicate all the DWARF sections are gone.

I bisected it to commit 77482c4bf63a.
https://git.fedorahosted.org/cgit/elfutils.git/commit/?id=77482c4bf63a921...

   if (elf->parent == NULL
       && (elf->maximum_size == ~((size_t) 0)
 	  || (size_t) size > elf->maximum_size)
-      && unlikely (ftruncate (elf->fildes, size) != 0))
+      && unlikely (posix_fallocate (elf->fildes, 0, size) != 0))

In elf_begin() w/ ELF_C_RDWR, we get maximum_size == ~0, but now this
posix_fallocate can only increase the file size.  The later part of
write_file() to reduce the size is only called for maximum_size != ~0.


On 06/15/2015 11:32 AM, Josh Boyer wrote:
>  On Mon, Jun 15, 2015 at 1:45 PM, drago01 <drago01(a)gmail.com&gt;
> wrote:
> > On Mon, Jun 15, 2015 at 7:38 PM, poma <pomidorabelisima(a)gmail.com&gt; wrote:
> >>
> >> # du -hs /lib/modules/4.0.5-300.fc22.i686/ | awk '{print $1}'
> >> 42M
> >>
> >> kernel-modules-4.0.5-300.fc22.i686.rpm  ...  17M
> >>
> >> http://dl.fedoraproject.org/pub/fedora/linux/updates/testing/22/i386/k/
> >> ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~
> >>
> >> # du -hs /lib/modules/4.1.0-0.rc7.git0.1.fc23.x86_64/ | awk '{print $1}'
> >> 53M
> >>
> >> kernel-modules-4.1.0-0.rc7.git0.1.fc23.x86_64.rpm  ...  18M
> >>
> >>
> https://kojipkgs.fedoraproject.org/packages/kernel/4.1.0/0.rc7.git0.1.fc2...
> >> ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~ ~~ ~ ~~
> ~~~
> >>
> >> # du -hs /lib/modules/4.1.0-0.rc7.git1.1.fc23.x86_64/ | awk '{print $1}'
> >> 439M
> >>
> >> kernel-modules-4.1.0-0.rc7.git1.1.fc23.x86_64.rpm  ...  218M
> >>
> >>
> http://dl.fedoraproject.org/pub/fedora/linux/development/rawhide/x86_64/o...
> >> ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~ ~~~~ ~~~ ~~ ~ ~~
> ~~~ ~~~~
> >>
> >> # du -hs /lib/modules/4.1.0-0.rc8.git0.1.fc23.x86_64/ | awk '{print $1}'
> >> 438M
> >>
> >> kernel-modules-4.1.0-0.rc8.git0.1.fc23.x86_64.rpm  ...  218M
> >>
> >> https://kojipkgs.fedoraproject.org/work/tasks/9420/10059420/
> >> ~~~~~~~~~~~ ~~~~~~~~~~ ~~~~~~~~~ ~~~~~~~~ ~~~~~~~ ~~~~~~ ~~~~~
> >>
> >>
> >> 218 ÷ 18 = 12.111111111
> >> 439 ÷ 53 = 8.283018868
> >>
> >>> = 4.1.0-0.rc7.git1.1 cca. 10 x sizeof
> >
> > Those are debug builds.
>  
>  They aren't actually.
>  
>  It looks like something in the buildroot stopped stripping the kernel
>  modules between rc7.git0 and rc7.git1.  Both elf-utils and binutils
>  changed, so maybe something in there.  The same SRPM build on F22 has
>  the smaller size.
>  
>  I haven't looked into it more than that.  If someone wants to poke at
>  this and figure out what broke, that would be excellent.  From a
>  kernel package perspective, there isn't anything that we did that
>  should cause this change in behavior that I'm aware of.
>  
>  josh
>  _______________________________________________
>  kernel mailing list
>  kernel(a)lists.fedoraproject.org
>  https://admin.fedoraproject.org/mailman/listinfo/kernel
>

sizeof - kernel modules

[Mark Wielaard]

On Mon, 2015-06-15 at 14:28 -0700, Josh Stone wrote:
>  CC elfutils-devel, this appears to be a regression.  I'm
> top-posting my
>  own findings, but I'll leave the kernel report below...
>  
>  With elfutils-0.162, it appears that "strip" doesn't reduce the file
>  size when operating in-place (without -o).  This isn't particular to
>  funny kernel ELF -- I reproduced it on a local stap binary.
>  
>  My full stap is 31MB.  With strip -o, I get a new 2.5MB file.  With
>  strip in-place, stap remains 31MB even though the ELF section data does
>  indicate all the DWARF sections are gone.
>  
>  I bisected it to commit 77482c4bf63a.
>
> https://git.fedorahosted.org/cgit/elfutils.git/commit/?id=77482c4bf63a921...
>  
>     if (elf->parent == NULL
>         && (elf->maximum_size == ~((size_t) 0)
>   	  || (size_t) size > elf->maximum_size)
>  -      && unlikely (ftruncate (elf->fildes, size) != 0))
>  +      && unlikely (posix_fallocate (elf->fildes, 0, size) != 0))
>  
>  In elf_begin() w/ ELF_C_RDWR, we get maximum_size == ~0, but now this
>  posix_fallocate can only increase the file size.  The later part of
>  write_file() to reduce the size is only called for maximum_size != ~0. 
Yes, you are right. Of course it is the smallest one-liner patch of the
release with the most comments and discussions across multiple lists.
And still people (me) failed to spot the maximum_size == -1 case. Sigh.
Sorry about that.

Simplest probably is to just revert this one patch and do a elfutils
0.163 release with just that (and the translation updates). The issue
that this was preventing (disk full with ELF_C_RDWR_MMAP, ftruncate
appear to succeed, write in memory map causes SIGBUS) is probably less
likely to trigger than this regression.

I have to get me some sleep now. I look over the other options tomorrow.

Cheers,

Mark

sizeof - kernel modules

[Mark Wielaard]

On Tue, 2015-06-16 at 00:20 +0200, Mark Wielaard wrote:
>  Simplest probably is to just revert this one patch and do a elfutils
>  0.163 release with just that (and the translation updates). The issue
>  that this was preventing (disk full with ELF_C_RDWR_MMAP, ftruncate
>  appear to succeed, write in memory map causes SIGBUS) is probably less
>  likely to trigger than this regression. 
Even simpler is to always just (also) call ftruncate to set the right
file size. Which is what the attached patch does. It also includes some
test cases to check in-place-eu-strip actually reduces the file size.

I'll update the elfutils fedora package so there is a fix right now. But
unless people complain I would like to just make a 0.163 release end of
the week with this fix and any other urgent fixes because I think this
is a somewhat bad regression and we should urge people to not just
patch, but to upgrade to a known good version.

Thanks,

Mark

sizeof - kernel modules

[Dennis Gilmore]

On Tuesday, June 16, 2015 02:17:10 PM Mark Wielaard wrote:
>  On Tue, 2015-06-16 at 00:20 +0200, Mark Wielaard wrote:
>  > Simplest probably is to just revert this one patch and do a elfutils
>  > 0.163 release with just that (and the translation updates). The issue
>  > that this was preventing (disk full with ELF_C_RDWR_MMAP, ftruncate
>  > appear to succeed, write in memory map causes SIGBUS) is probably less
>  > likely to trigger than this regression.
>  
>  Even simpler is to always just (also) call ftruncate to set the right
>  file size. Which is what the attached patch does. It also includes some
>  test cases to check in-place-eu-strip actually reduces the file size.
>  
>  I'll update the elfutils fedora package so there is a fix right now. But
>  unless people complain I would like to just make a 0.163 release end of
>  the week with this fix and any other urgent fixes because I think this
>  is a somewhat bad regression and we should urge people to not just
>  patch, but to upgrade to a known good version. 
We need this fixed for the mass rebuild that is due to start right now. so 
please apply the patch ASAP and let me know so I can start the rebuild

Dennis

sizeof - kernel modules

[Peter Robinson]

On Tue, Jun 16, 2015 at 10:26 PM, Dennis Gilmore <dennis(a)ausil.us&gt; wrote:
>  On Tuesday, June 16, 2015 02:17:10 PM Mark Wielaard wrote:
> > On Tue, 2015-06-16 at 00:20 +0200, Mark Wielaard wrote:
> > > Simplest probably is to just revert this one patch and do a elfutils
> > > 0.163 release with just that (and the translation updates). The issue
> > > that this was preventing (disk full with ELF_C_RDWR_MMAP, ftruncate
> > > appear to succeed, write in memory map causes SIGBUS) is probably less
> > > likely to trigger than this regression.
> >
> > Even simpler is to always just (also) call ftruncate to set the right
> > file size. Which is what the attached patch does. It also includes some
> > test cases to check in-place-eu-strip actually reduces the file size.
> >
> > I'll update the elfutils fedora package so there is a fix right now. But
> > unless people complain I would like to just make a 0.163 release end of
> > the week with this fix and any other urgent fixes because I think this
> > is a somewhat bad regression and we should urge people to not just
> > patch, but to upgrade to a known good version.
>
>  We need this fixed for the mass rebuild that is due to start right now. so
>  please apply the patch ASAP and let me know so I can start the rebuild 
It's already applied in NVR elfutils-0.162-2.fc23

http://koji.fedoraproject.org/koji/buildinfo?buildID=646258

sizeof - kernel modules

[Josh Stone]

On 06/16/2015 02:31 PM, Peter Robinson wrote:
>  On Tue, Jun 16, 2015 at 10:26 PM, Dennis Gilmore
> <dennis(a)ausil.us&gt; wrote:
> > On Tuesday, June 16, 2015 02:17:10 PM Mark Wielaard wrote:
> >> On Tue, 2015-06-16 at 00:20 +0200, Mark Wielaard wrote:
> >>> Simplest probably is to just revert this one patch and do a elfutils
> >>> 0.163 release with just that (and the translation updates). The issue
> >>> that this was preventing (disk full with ELF_C_RDWR_MMAP, ftruncate
> >>> appear to succeed, write in memory map causes SIGBUS) is probably less
> >>> likely to trigger than this regression.
> >>
> >> Even simpler is to always just (also) call ftruncate to set the right
> >> file size. Which is what the attached patch does. It also includes some
> >> test cases to check in-place-eu-strip actually reduces the file size.
> >>
> >> I'll update the elfutils fedora package so there is a fix right now. But
> >> unless people complain I would like to just make a 0.163 release end of
> >> the week with this fix and any other urgent fixes because I think this
> >> is a somewhat bad regression and we should urge people to not just
> >> patch, but to upgrade to a known good version.
> >
> > We need this fixed for the mass rebuild that is due to start right now. so
> > please apply the patch ASAP and let me know so I can start the rebuild
>  
>  It's already applied in NVR elfutils-0.162-2.fc23
>  
>  http://koji.fedoraproject.org/koji/buildinfo?buildID=646258 
Yes.  It doesn't appear to have landed on mirrors yet, but it is in
f23-build.  I just confirmed that it's reducing files properly now.

sizeof - kernel modules

[Peter Robinson]

On Tue, Jun 16, 2015 at 10:36 PM, Josh Stone <jistone(a)redhat.com&gt; wrote:
>  On 06/16/2015 02:31 PM, Peter Robinson wrote:
> > On Tue, Jun 16, 2015 at 10:26 PM, Dennis Gilmore <dennis(a)ausil.us&gt; wrote:
> >> On Tuesday, June 16, 2015 02:17:10 PM Mark Wielaard wrote:
> >>> On Tue, 2015-06-16 at 00:20 +0200, Mark Wielaard wrote:
> >>>> Simplest probably is to just revert this one patch and do a elfutils
> >>>> 0.163 release with just that (and the translation updates). The issue
> >>>> that this was preventing (disk full with ELF_C_RDWR_MMAP, ftruncate
> >>>> appear to succeed, write in memory map causes SIGBUS) is probably less
> >>>> likely to trigger than this regression.
> >>>
> >>> Even simpler is to always just (also) call ftruncate to set the right
> >>> file size. Which is what the attached patch does. It also includes some
> >>> test cases to check in-place-eu-strip actually reduces the file size.
> >>>
> >>> I'll update the elfutils fedora package so there is a fix right now. But
> >>> unless people complain I would like to just make a 0.163 release end of
> >>> the week with this fix and any other urgent fixes because I think this
> >>> is a somewhat bad regression and we should urge people to not just
> >>> patch, but to upgrade to a known good version.
> >>
> >> We need this fixed for the mass rebuild that is due to start right now. so
> >> please apply the patch ASAP and let me know so I can start the rebuild
> >
> > It's already applied in NVR elfutils-0.162-2.fc23
> >
> > http://koji.fedoraproject.org/koji/buildinfo?buildID=646258
>
>  Yes.  It doesn't appear to have landed on mirrors yet, but it is in 
Mirrors aren't used for mass rebuilds, it doesn't go out to the
mirrors until after the nightly build, so it'll be out there for
general consumption tomorrow.

Peter