Howdy,
I recently patched a Solaris 10 server, and it looks like the patch update
broke the DTraceToolkit anonpgpid.d script:
$ anonpgpid.d
dtrace: failed to compile script ./anonpgpid.d:
"/usr/include/sys/kstat.h", line 439: invalid type combination
After doing a bit of digging, it looks like this is caused by the
"#include <sys/vnode.h>" statement in the script:
$ cat foo
#include <sys/vnode.h>
syscall:::entry
{}
$ dtrace -Cs foo
dtrace: failed to compile script foo: "/usr/include/sys/kstat.h", line
439: invalid type combination
Does anyone happen to know if anything changed?
Thanks for any insight,
- Ryan
--
UNIX Administrator
http://prefetch.net
> > Howdy, > > I recently patched a Solaris 10 server, and it looks like the patch update > broke the DTraceToolkit anonpgpid.d script: > > $ anonpgpid.d > dtrace: failed to compile script ./anonpgpid.d: > "/usr/include/sys/kstat.h", line 439: invalid type combination > > After doing a bit of digging, it looks like this is caused by the > "#include <sys/vnode.h>" statement in the script: > > $ cat foo > #include <sys/vnode.h> > > syscall:::entry > {} > > $ dtrace -Cs foo > dtrace: failed to compile script foo: "/usr/include/sys/kstat.h", line > 439: invalid type combination > > Does anyone happen to know if anything changed? > > Thanks for any insight, > - RyanWhat does this stuff report on the system: $ /usr/bin/uname -rv $ /usr/ccs/bin/what /usr/include/sys/kstat.h $ /usr/ccs/bin/what /usr/include/sys/vnode.h -Mike -- Mike Shapiro, Solaris Kernel Development. blogs.sun.com/mws/
On Mon, 28 Aug 2006, Michael Shapiro wrote:> What does this stuff report on the system: > > $ /usr/bin/uname -rv > $ /usr/ccs/bin/what /usr/include/sys/kstat.h > $ /usr/ccs/bin/what /usr/include/sys/vnode.hHere is the output from uname and what: $ /usr/bin/uname -rv 5.10 Generic_118833-18 $ /usr/ccs/bin/what /usr/include/sys/kstat.h /usr/include/sys/kstat.h: kstat.h 1.20 04/03/29 SMI $ /usr/ccs/bin/what /usr/include/sys/vnode.h /usr/include/sys/vnode.h: vnode.h 1.125 06/04/24 SMI Thanks for any insight, - Ryan -- UNIX Administrator http://prefetch.net
> > > On Mon, 28 Aug 2006, Michael Shapiro wrote: > > > What does this stuff report on the system: > > > > $ /usr/bin/uname -rv > > $ /usr/ccs/bin/what /usr/include/sys/kstat.h > > $ /usr/ccs/bin/what /usr/include/sys/vnode.h > > Here is the output from uname and what: > > $ /usr/bin/uname -rv > 5.10 Generic_118833-18 > > $ /usr/ccs/bin/what /usr/include/sys/kstat.h > /usr/include/sys/kstat.h: > kstat.h 1.20 04/03/29 SMI > > $ /usr/ccs/bin/what /usr/include/sys/vnode.h > /usr/include/sys/vnode.h: > vnode.h 1.125 06/04/24 SMI > > Thanks for any insight, > - Ryankstat.h version 1.20 is missing the fix for this bug: 6315039 including <sys/kstat.h> in a D script induces compile-time error wherein we changed its use of the D reserved "string" keyword, which is what is causing the problem. You''ll need to get a patch for that (or use latest SX or OpenSolaris bits or equivalent). -Mike -- Mike Shapiro, Solaris Kernel Development. blogs.sun.com/mws/
> On Mon, 28 Aug 2006, Michael Shapiro wrote: > > > What does this stuff report on the system: > > > > $ /usr/bin/uname -rv > > $ /usr/ccs/bin/what /usr/include/sys/kstat.h > > $ /usr/ccs/bin/what /usr/include/sys/vnode.h > > Here is the output from uname and what: > > $ /usr/bin/uname -rv > 5.10 Generic_118833-18 > > $ /usr/ccs/bin/what /usr/include/sys/kstat.h > /usr/include/sys/kstat.h: > kstat.h 1.20 04/03/29 SMI > > $ /usr/ccs/bin/what /usr/include/sys/vnode.h > /usr/include/sys/vnode.h: > vnode.h 1.125 06/04/24 SMI > > > kstat.h version 1.20 is missing the fix for this bug: > > 6315039 including <sys/kstat.h> in a D script induces compile-time error > > wherein we changed its use of the D reserved "string" keyword, which is > what is causing the problem. You''ll need to get a patch for that (or > use latest SX or OpenSolaris bits or equivalent). > > -MikeThanks for the information. I wasn''t able to find a patch on sunsolve.sun.com, so I opened a support case to see if one was available. According to the support engineer, this issue won''t be officially fixed in a GA release until Solaris 11: Reply from Sun support:> The bug 6315039 is set to be fixed in Solaris Nevada (solaris 11)--and > I''m not sure when the general audience release is set.> 118844-30 is a Dtrace patch (among other things)--but, it doesn''t state > that it will alleviate the aforementioned bug nor its sister bug > 6311947.> To the best of my knowledge this bug won''t be ''officially'' slain until > solaris 11.Since several scripts in the DTraceToolkit quit working because of this bug, I thought I would forward the reply from Sun support to the list. Hopefully Google will archive this message so people have easy access to Mike''s explanation and recommended fix (i.e., use opensolaris). Thanks, - Ryan -- UNIX Administrator http://prefetch.net
James C. McPherson
2006-Sep-02 12:15 UTC
[dtrace-discuss] Re: anonpgpid.d and sys/vnode.h issue
Matty wrote: ...> Thanks for the information. I wasn''t able to find a patch on > sunsolve.sun.com, so I opened a support case to see if one was > available. According to the support engineer, this issue won''t be > officially fixed in a GA release until Solaris 11: > > Reply from Sun support: > >> The bug 6315039 is set to be fixed in Solaris Nevada (solaris 11)--and >> I''m not sure when the general audience release is set. > >> 118844-30 is a Dtrace patch (among other things)--but, it doesn''t >> state that it will alleviate the aforementioned bug nor its sister bug >> 6311947. > >> To the best of my knowledge this bug won''t be ''officially'' slain until >> solaris 11.Matty, did you request that the fix be backported to Solaris 10? There''s no other way to get that fix on Solaris 10 unless you do so. James C. McPherson
Michael Shapiro
2006-Sep-02 20:42 UTC
[dtrace-discuss] Re: anonpgpid.d and sys/vnode.h issue
> > Thanks for the information. I wasn''t able to find a patch on > sunsolve.sun.com, so I opened a support case to see if one was available. > According to the support engineer, this issue won''t be officially fixed > in a GA release until Solaris 11: >That isn''t necessarily correct: that just means that absent any proactive effort to backport, it won''t show up until S11. But of course, we here at dtrace-discuss aim to please ... so we''ll add to this our list of things to try to get backported into S10 Update 4. Adam or I will post an update to that situation once we have the full U4 plan worked out. -Mike -- Mike Shapiro, Solaris Kernel Development. blogs.sun.com/mws/