On my PC, gcry_cipher_encrypt() doesn't release all x87 FPU data registers .

[Mitsutoshi NAKANO]

1)
Description: Xubuntu 12.10 desktop i386
Release: 12.10
PC: Sony PCG-SRX7S/P Memory 512 M byte http://www.sony.jp/products/biz/vaio/PCG-SRX7S_P/spec.html
Chipset: Intel 815EM
$ cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 11
model name : Mobile Intel(R) Pentium(R) III CPU - M 900MHz
stepping : 4
microcode : 0x2
cpu MHz : 500.000
cache size : 512 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pse36 mmx fxsr sse up
bogomips : 993.64
clflush size : 32
cache_alignment : 32
address sizes : 36 bits physical, 32 bits virtual
power management:
$ uname -a
Linux loghost 3.5.0-22-generic #34-Ubuntu SMP Tue Jan 8 21:41:11 UTC 2013 i686 i686 i686 GNU/Linux

2)
libgcrypt11 1.5.0-3ubuntu1
gnome-keyring 3.6.1-0ubuntu1

3) What you expected to happen
$apt-get -b source gnome-keyring
$cd gnome-keyring-3.6.1/
$make -k check | tee make-k-check2.txt
 →all check OK

4) What happened instead
 → but /gnome2-store/import/pkcs12: was FAIL on the PC .
See also https://dl.dropbox.com/u/86335040/make-k-check2.txt .

5) I analyzed the gnome-keyring program.
So I found that gcry_cipher_encrypt() doesn't release all x87 FPU data registers .
Then -nan goes into a variable subsequent processing.

This problem found on the test of /gnome2-store/import/pkcs12 on
gnome-keyring-3.6.1,
function is encrypt_buffer() in gnome-keyring-3.6.1/pkcs11/gnome2-store/gkm-gnome2-file.c .

I wrote a reappearance program .
https://dl.dropbox.com/u/86335040/bug.tar.gz .
Please check it, and please fix this problem

See also https://forums.ubuntulinux.jp/viewtopic.php?id=14643
and http://d.hatena.ne.jp/Itisango/searchdiary?word=%2A%5Bgnome-keyring%5D .

Thanks .

[Mitsutoshi NAKANO (bkbin005)]

Hi .
I found the position of the problem, then I wrote a patch .
When adapted in this patch, the test of gnome-keyring was OK .
However, I cannot judge whether this correction is right .
And I cannot judge whether there is the same problem as other parts .
I want you to check these .

And although "libgcrypt11:i386" is the i386 architecture, I cannot judge whether it is allowed to call MMX instructions by the i386 architecture.
Then, I wish for the view from intellectuals .

Thanks .

[Ubuntu Foundations Team Bug Bot (crichton)]

The attachment "rijndael.c.diff" of this bug report has been identified as being a patch. The ubuntu-reviewers team has been subscribed to the bug report so that they can review the patch. In the event that this is in fact not a patch you can resolve this situation by removing the tag 'patch' from the bug report and editing the attachment so that it is not flagged as a patch. Additionally, if you are member of the ubuntu-reviewers team please also unsubscribe the team from this bug report.

[This is an automated message performed by a Launchpad user owned by Brian Murray. Please contact him regarding any issues with the action taken in this bug report.]

[Mitsutoshi NAKANO (bkbin005)]

The patch was released to Gcrypt-devel ML ( http://lists.gnupg.org/mailman/listinfo/gcrypt-devel ) .
For the people troubled by this problem, I wrote that how to apply the patch

*** How to apply the patch . ***

1. Get source .
$apt-get -b source libgcrypt11
$ cd libgcrypt11-1.5.0

NOTE: Additional packages may be required at this time and you may become an error.
      Follow directions then .
   ex) $ sudo apt-get install dpkg-dev
       $ sudo apt-get install debhelper texlive-latex-base texlive-generic-recommended texinfo cdbs libgpg-error-dev autotools-dev

2. Get the patch .
2.1. Access by your Web browser to http://marc.info/?l=gcrypt-devel&m=135944673925649&w=2
2.2. Start your editor.
$vi mail.patch
2.3. Copy&paste the mail message whole sentence .
2.4. Save it
NOTE: Don't use wget .

3. Apply the patch .
$ patch -p1 -b <mail.patch

4. Check updating .
$ diff cipher/rijndael.c.orig cipher/rijndael.c | less

5. Build it .
$ make

6. Test it .
$ make check
$ echo $?
0

7. Install it .
$ sudo make install
$ sudo ldconfig

8. Test the application with which the problem had occurred .
ex) $ chromium-browser --password-store=gnome

Thanks all .

[Andreas Metzler (k-launchpad-downhill-at-eu-org)]

This included in Debian's 1.5.0-5.

[Launchpad Janitor (janitor)]

This bug was fixed in the package libgcrypt11 - 1.5.3-2ubuntu1

---------------
libgcrypt11 (1.5.3-2ubuntu1) trusty; urgency=low

  * Merge from Debian unstable. Remaining changes:
    - no-global-init-thread-callbacks.diff: Do not call global_init when
      setting thread callbacks

libgcrypt11 (1.5.3-2) unstable; urgency=low

  * Convert to dh and move building of ps and html docs to
    override_dh_auto_build-indep. Enable parallel building.

libgcrypt11 (1.5.3-1) unstable; urgency=high

  * New upstream bugfix release. (CVE-2013-4242)

libgcrypt11 (1.5.2-3) unstable; urgency=low

  * Install libgcrypt.a and libgcrypt.so to /usr.
  * [15_multiarchpath_in_-L.diff] Do not print -L/lib/i386-linux-gnu on
    "libgcrypt-config --libs".
  * Use debhelper v9 mode. This allows us to mark libgcrypt11-dbg Multi-Arch:
    same.

libgcrypt11 (1.5.2-2) unstable; urgency=low

  * Upload to unstable.
  * Fix vcs-field-not-canonical lintian error by refering to anonscm instead
    of svn.debian.org.
  * Update info in debian/copyright from upstream's README, fixing typo 'teh'.
  * Delete some outdated and unused code in debian/rules.

libgcrypt11 (1.5.2-1) experimental; urgency=low

  * New upstream version.
    + IDEA support added.
  * Move list of supported algorithms to a separate paragraph in description
    to decrease work-load of translators. Closes: #640261
  * Move TeX-packages from b-d to Build-Depends-Indep. (Thanks, P. J.
    McDermott) Closes: #682597

libgcrypt11 (1.5.1-1) experimental; urgency=low

  * Point watchfile to stable release.
  * New upstream version.
  * Drop superfluous patches:
    29_Fix-a-problem-with-select-and-high-fds.patch
    30_Avoid-dereferencing-pointer-right-after-the-end.patch
    31_Fix-segv-with-AES-NI-on-some-platforms.patch
    32_libgcrypt-1.5-rinjdael-Fix-use-of-SSE2-outside-USE_A.patch
  * Bump version gcry_control@GCRYPT_1.2 in debian/libgcrypt11.symbols from
    1.4.5 to 1.5.1 since its argument enum has a new member.

libgcrypt11 (1.5.0-5) unstable; urgency=low

  * While we are at it also pick
    29_Fix-a-problem-with-select-and-high-fds.patch
    LP: #1084279

libgcrypt11 (1.5.0-4) unstable; urgency=low

  * Pull patches from upstream LIBGCRYPT-1-5-BRANCH:
      30_Avoid-dereferencing-pointer-right-after-the-end.patch
      31_Fix-segv-with-AES-NI-on-some-platforms.patch
         <https://bugs.g10code.com/gnupg/issue1452> LP: #1105758
      32_libgcrypt-1.5-rinjdael-Fix-use-of-SSE2-outside-USE_A.patch
    Closes: #699034
 -- Seth Arnold <email address hidden> Wed, 27 Nov 2013 10:36:27 -0800