"C. Bergstr??m" writes:
> Hi Tom/David,
> > I've got the core file. I'll try to take a look at it sometime today.
> >
> > First, I'd like to back off a little bit to make sure that I understand
> > what we're doing here.
> >
> > As I understand things, you've made some changes to SMF code. Now when you
> > try to build you're running into problems in the usr/src/cmd/svc/seed
> > directory, because svccfg is dumping core. Is this pretty much correct?
> >
> > If it is correct, it would be helpful if you could give us a webrev showing
> > the code changes that you have made. Right now, we're sort of searching in
> > the dark.
> >
>
> I'm attaching an hg.diff.. I think this is what you mean by webrev.. The
> changes are pretty minor..
>
> 1) two duplicate lines of asm in libc (you can safely ignore this)
> 2) I've removed native from the build (will make sense/explain below)
> 3) There's a one line change in (I'm not sure these changes are correct)
> b/usr/src/cmd/svc/svccfg/svccfg_libscf.c
> b/usr/src/cmd/svc/startd/libscf.c
> b/usr/src/cmd/svc/configd/rc_node.c
> 4) In addition to the normal bldenv -d ./opensolaris.sh you'll need to
> export LDLIBS32="-m64"
> export CFLAGS="-m64 -g"
>
> Here's where I think I may lose help.. I'm trying to compile this all as
> x86_64. You'll notice there's no amd64/ directory for libsqlite or svc
> and thuse the reason for needing those two extra exports and removing
> native from the Makefiles
>
> I have an exact reproducible script against a fresh onnv-gate checkout,
> but it doesn't involve using bldenv to set things up. The execution
> flow of code shouldn't change between 32/64 bit, but if 64bit is running
> out of memory there's something strange going on I think. (Either a leak
> or change I've made)
>
> Warmest thanks,
>
> ./C
> diff -r 3b46d92554cb usr/src/cmd/svc/configd/rc_node.c
> --- a/usr/src/cmd/svc/configd/rc_node.c Thu Nov 06 08:38:48 2008 -0800
> +++ b/usr/src/cmd/svc/configd/rc_node.c Tue Dec 02 16:58:22 2008 +0000
> @@ -6249,7 +6249,7 @@
> */
> int
> rc_node_setup_iter(rc_node_ptr_t *npp, rc_node_iter_t **iterp,
> - uint32_t type, uint32_t flags, const char *pattern)
> + uint32_t type, size_t flags, const char *pattern)
> {
> rc_node_t *np;
> rc_iter_filter_func *f = NULL;
> diff -r 3b46d92554cb usr/src/cmd/svc/seed/Makefile
> --- a/usr/src/cmd/svc/seed/Makefile Thu Nov 06 08:38:48 2008 -0800
> +++ b/usr/src/cmd/svc/seed/Makefile Tue Dec 02 16:58:22 2008 +0000
> @@ -111,18 +111,18 @@
> SEEDFILEMODE = 0444 # seeds are not intended for editing, but may
> # be copied
>
> -CONFIGD = ../configd/svc.configd-native
> -SVCCFG = ../svccfg/svccfg-native
> +CONFIGD = ../configd/svc.configd
> +SVCCFG = ../svccfg/svccfg
>
> .KEEP_STATE:
>
> all: global.db nonglobal.db miniroot.db
>
> $(CONFIGD): FRC
> - @cd ../configd; pwd; $(MAKE) $(MFLAGS) native
> + @cd ../configd; pwd; $(MAKE) $(MFLAGS)
>
> $(SVCCFG): FRC
> - @cd ../svccfg; pwd; $(MAKE) $(MFLAGS) native
> + @cd ../svccfg; pwd; $(MAKE) $(MFLAGS)
>
> ../milestone/console-login.xml:
> @cd ../milestone; pwd; $(MAKE) $(MFLAGS) console-login.xml
> @@ -134,7 +134,7 @@
> SVCCFG_DTD=../dtd/service_bundle.dtd.1 \
> SVCCFG_REPOSITORY=$(SRC)/cmd/svc/seed/common.db \
> SVCCFG_CONFIGD_PATH=$(CONFIGD) \
> - $(SVCCFG) import $$m; \
> + $(SVCCFG) -v import $$m; \
> done
>
> global.db: common.db $(GLOBAL_ZONE_DESCRIPTIONS) $(CONFIGD) $(SVCCFG)
> diff -r 3b46d92554cb usr/src/cmd/svc/startd/libscf.c
> --- a/usr/src/cmd/svc/startd/libscf.c Thu Nov 06 08:38:48 2008 -0800
> +++ b/usr/src/cmd/svc/startd/libscf.c Tue Dec 02 16:58:22 2008 +0000
> @@ -3182,7 +3182,7 @@
> */
> int
> libscf_unset_action(scf_handle_t *h, scf_propertygroup_t *pg,
> - admin_action_t a, hrtime_t ts)
> + admin_action_t a, int64_t ts)
> {
> scf_transaction_t *t;
> scf_transaction_entry_t *e;
> diff -r 3b46d92554cb usr/src/cmd/svc/svccfg/Makefile
> --- a/usr/src/cmd/svc/svccfg/Makefile Thu Nov 06 08:38:48 2008 -0800
> +++ b/usr/src/cmd/svc/svccfg/Makefile Tue Dec 02 16:58:22 2008 +0000
> @@ -114,7 +114,9 @@
>
> all debug: $(PROG)
>
> -native: FRC
> +native:
> +
> +native_old: FRC
> @cd $(LIBUUTIL)/native; pwd; $(MAKE) $(MFLAGS) install
> @cd $(LIBSCF)/native; pwd; $(MAKE) $(MFLAGS) install
> @NATIVE_BUILD= $(MAKE) $(MFLAGS) all
> diff -r 3b46d92554cb usr/src/cmd/svc/svccfg/svccfg_libscf.c
> --- a/usr/src/cmd/svc/svccfg/svccfg_libscf.c Thu Nov 06 08:38:48 2008 -0800
> +++ b/usr/src/cmd/svc/svccfg/svccfg_libscf.c Tue Dec 02 16:58:22 2008 +0000
> @@ -1882,6 +1882,7 @@
>
> default:
> bad_error("scf_transaction_commit", r);
> + /* { ( void ) fprintf ( ( & __iob [ 2 ] ) , "%s:%d: %s() failed
> with unexpected " "error %d. Aborting.\n" , "svccfg_libscf.c" , 1884 , (
> "scf_transaction_commit" ) , ( r ) ) ; }; */
> }
>
> scf_transaction_destroy_children(imp_tx);
> diff -r 3b46d92554cb usr/src/common/atomic/amd64/atomic.s
> --- a/usr/src/common/atomic/amd64/atomic.s Thu Nov 06 08:38:48 2008 -0800
> +++ b/usr/src/common/atomic/amd64/atomic.s Tue Dec 02 16:58:22 2008 +0000
> @@ -432,8 +432,6 @@
> jne 1b
> movzbl %cl, %eax / return new value
> ret
> - SET_SIZE(atomic_and_uchar_nv)
> - SET_SIZE(atomic_and_8_nv)
>
> ENTRY(atomic_or_16_nv)
> ALTENTRY(atomic_or_ushort_nv)
> diff -r 3b46d92554cb usr/src/lib/libscf/Makefile
> --- a/usr/src/lib/libscf/Makefile Thu Nov 06 08:38:48 2008 -0800
> +++ b/usr/src/lib/libscf/Makefile Tue Dec 02 16:58:22 2008 +0000
> @@ -53,7 +53,7 @@
>
> .KEEP_STATE:
>
> -all clean clobber install: native $(SUBDIRS)
> +all clean clobber install: $(SUBDIRS)
>
> lint: $(SUBDIRS)
>
> @@ -65,7 +65,7 @@
>
> _msg: $(MSGDOMAINPOFILE)
>
> -$(SUBDIRS) native: FRC
> +$(SUBDIRS) : FRC
> @cd $@; pwd; $(MAKE) $(TARGET)
>
> FRC:
I haven't had a chance to look at your diffs yet, but here is what I found
by looking at your core file.
In svccfg.h, bad_error() is defined:
#define bad_error(func, err) { \
(void) fprintf(stderr, "%s:%d: %s() failed with unexpected " \
"error %d. Aborting.\n", __FILE__, __LINE__, (func), (err)); \
abort(); \
}
Now lets look at your core file with mdb.
$ mdb core
Loading modules: [ libumem.so.1 libc.so.1 libuutil.so.1 ld.so.1 ]
> ::status
debugging core file of svccfg (64-bit) from
file: /usr/sbin/svccfg
initial argv: ../svccfg/svccfg -v import ../milestone/global.xml
threading model: native threads
status: process terminated by SIGABRT (Abort), pid=7362 uid=0 code=-1
The stack trace is:
> $c
libc.so.1`_lwp_kill+0xa()
libc.so.1`raise+0x19()
libc.so.1`abort+0xca()
entity_pgroup_import+0xa59()
libuutil.so.1`uu_list_walk+0xa4()
lscf_service_import+0x55b()
libuutil.so.1`uu_list_walk+0xa4()
lscf_bundle_import+0x11e4()
engine_import+0x4e2()
yyparse+0x70c()
engine_exec+0x65()
main+0x268()
Let's disassemble the code in entity_proup_import() around the abort call:
> entity_pgroup_import+0xa59::dis
entity_pgroup_import+0xa20: movl -0x78(%rbp),%r9d
entity_pgroup_import+0xa24: movq $0x45bfe8,%r8
entity_pgroup_import+0xa2b: movl $0x75c,%ecx
entity_pgroup_import+0xa30: movq $0x45b338,%rdx
entity_pgroup_import+0xa37: movq $0x45b348,%rsi
entity_pgroup_import+0xa3e: movq $0x478f20,%rdi <svccfg`_iob+0x100>
entity_pgroup_import+0xa45: movl $0x0,%eax
entity_pgroup_import+0xa4a: call -0xcf57 <PLT=libc.so.1`fprintf>
entity_pgroup_import+0xa4f: movl $0x0,%eax
entity_pgroup_import+0xa54: call -0xce11 <PLT=libc.so.1`abort>
entity_pgroup_import+0xa59: jmp +0x2c <entity_pgroup_import+0xa87>
entity_pgroup_import+0xa5b: nop
entity_pgroup_import+0xa5c: movl -0x78(%rbp),%eax
entity_pgroup_import+0xa5f: addl $0x1,%eax
entity_pgroup_import+0xa62: cmpl $0x2,%eax
entity_pgroup_import+0xa65: ja -0x47 <entity_pgroup_import+0xa20>
entity_pgroup_import+0xa67: shll $0x2,%eax
entity_pgroup_import+0xa6a: movl %eax,%eax
entity_pgroup_import+0xa6c: cltq
entity_pgroup_import+0xa6e: movq %rax,%r8
entity_pgroup_import+0xa71: movl 0x41c670(%r8),%eax
At a4a there is a call to fprintf. Let's look at the parameters to the
fprintf call:
> 0x45b348/s
0x45b348: %s:%d: %s() failed with unexpected error %d. Aborting.
Looks like the format string in bad_error().
> 0x45b338/s
0x45b338: svccfg_libscf.c
That's the file name, and the line number is:
> 0x75c=d
1884
So bad_error() is being invoked at line 1884 in svccfg.c. The function
that failed is:
> 0x45bfe8/s
0x45bfe8: scf_transaction_commit
tom
