Hello,
Yes, there was some problems with the linker patches later last
year. The attached is a note on explaining the problem.
Seizo
|
| Seizo Sakurai wrote:
| > Please do not use the patches which appear in between the fix for
| > the bug 6332983 and 5080443. They are all bad-patched, or in the
| > process of being bad-padched.
|
| So, to be clear, are you saying that these patches are bad and
| should not be used?
|
| > Solaris/SunOS 5.10_sparc patch T117461-03, -04, -07
| > Solaris/SunOS 5.10_x86 patch T118345-03, -04, -11
| > Solaris/SunOS 5.9_sparc patch T112963-21, -22, -23, -24
| > Solaris/SunOS 5.9_x86 patch T113986-17, -18, -19, -20
| > Solaris/SunOS 5.8_sparc patch T109147-36, -37, -38, -39
| > Solaris/SunOS 5.8_x86 patch T109148-36, -37, -38, -39
|
| Thanks for the clarification!
|
| glen
| This message posted from opensolaris.org
| _______________________________________________
| tools-linking mailing list
| tools-linking at opensolaris.org
-------------- next part --------------
======================================
=
======================================
The attacked file SUNWonld-README lists all the bugs fixed
in the Solaris link-editors from Solaris 5.5 till
the current Solaris 11.
The file also lists all the patches delivered since
Solaris 5.5.
-------------------
- bug information -
-------------------
Please see the attached SUNonld-README file.
The bugs and patches appear later are putbacked later.
The fix for the bug:
5056867 $ISALIST/$HWCAP expansion should be more flexible.
was integrated into the Solaris 10.
The fix for the bug:
5080443 lazy loading failure doesn't clean up after itself
was integrated into Nevada gate earlier this year.
We got hit by a security bug:
6291547 ld.so mishandles LD_AUDIT causing security problems
This bug was introduced by the fix for 5056867. The fix for this
bug 6291547 was fixed in Nevada gate and back ported to Solaris 10/9/8.
Since this was a security bug, many customers using Solaris 9/8
installed the patch including the fix for 6291547. The patches
also included the fix for 5080443.
The fix for 5080553 cleaned up many runtime-linker's problems but
it introduced regressions. Since many Solaris 9/8 users installed
the patches including the fix for 5080553 because of the security
fix for 6291547, number of customers got hit by the regressions.
The regressions introcuded by 5080443 caused the following problems:
6301218 Matlab dumps core on startup when running on 112963-22
6314115 Checkpoint refuses to start, crashes on start, after application of
6332983 s9 linker patches 112963-24/113986-20 causing cluster machines
We believe that the regressions caused by the 5080443 will now be fixed
by the above 3 bug fixes.
Any Solaris 10/9/8 patches which includes the fix for:
5056867 $ISALIST/$HWCAP expansion should be more flexible.
but which does not include the fix for:
6332983 s9 linker patches 112963-24/113986-20 causing cluster machines
should not be used and are basically bad-patched.
---------------------
- patch information -
---------------------
5056867 $ISALIST/$HWCAP expansion should be more flexible.
This bug fix was integrated into Solaris 10 gate when Solaris 10
was under development and first integrated into the follwoing
patches. This fix caused the security bug 6291547.
Solaris/SunOS 5.9_sparc patch T112963-16
Solaris/SunOS 5.9_x86 patch T113986-12
Solaris/SunOS 5.8_sparc patch T109147-31
Solaris/SunOS 5.8_x86 patch T109148-31
5080443 lazy loading failure doesn't clean up after itself
This bug fix was integrated into Solaris 11 gate. The Solaris 11
is the release which is under development. Because there was an escalation
for the bug:
6262789 Infosys wants a faster linker
6262789 was fixed in the Solaris 11 gate and the following patches were
generated for Solaris 10/9/8.
Solaris/SunOS 5.10_sparc patch T117461-03
Solaris/SunOS 5.10_x86 patch T118345-03
Solaris/SunOS 5.9_sparc patch T112963-21
Solaris/SunOS 5.9_x86 patch T113986-17
Solaris/SunOS 5.8_sparc patch T109147-36
Solaris/SunOS 5.8_x86 patch T109148-36
The above patches were not widely used by the customers but the following
security
bug was reported.
6291547 ld.so mishandles LD_AUDIT causing security problems.
The fix for this bug was putbacked into Solaris 11 and immediately the
following
patches were generated for Solaris 10/9/8.
Solaris/SunOS 5.10_sparc patch T117461-04
Solaris/SunOS 5.10_x86 patch T118345-04
Solaris/SunOS 5.9_sparc patch T112963-22
Solaris/SunOS 5.9_x86 patch T113986-18
Solaris/SunOS 5.8_sparc patch T109147-37
Solaris/SunOS 5.8_x86 patch T109148-37
Because the above patches were for a security bug fix, a lot of Solaris 9/8
users
installed the patches. The above patches also included the fix for
5080443 lazy loading failure doesn't clean up after itself
and number of users hit the regressions reported as:
6301218 Matlab dumps core on startup when running on 112963-22
6314115 Checkpoint refuses to start, crashes on start, after
application of
6332983 s9 linker patches 112963-24/113986-20 causing cluster machines
So all of the patches mentioned above which includes the fix for 5056867 should
not
be used.
The T-patches which includes the fix for 6332983, and 6301218/6314115, are
out and
will be out shortly. They are/(will be) the followings:
Solaris/SunOS 5.10_sparc patch T117461-08
Solaris/SunOS 5.10_x86 patch T121208-02
Solaris/SunOS 5.9_sparc patch T112963-25
Solaris/SunOS 5.9_x86 patch T113986-21
Solaris/SunOS 5.8_sparc patch T109147-40
Solaris/SunOS 5.8_x86 patch T109148-40
And the above patches will be the patches which will fix currently known
best linker bits.
For the above T-patches, which are already out, I am planning to release
the locks I have
this week after verifications. For the T-patches not out yet, I will do the
same as soon
as they are out.
Seizo,
10/31/2005
