Skip to contentSkip to conversation

Possible uninitialized 'last' variable in usdt_copyin_data()

PATCH In review [email protected] 2 participants 6 messages Aug 12, 2025 – Aug 15, 2025

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 Possible uninitialized 'last' variable in usdt_copyin_data()

From: Eugene Loh <[email protected]>

Some compilers warn:

libcommon/usdt_parser.c: In function ‘usdt_copyin_data’:
libcommon/usdt_parser.c:191:15: warning:
‘last’ may be used uninitialized in this function [-Wmaybe-uninitialized]

    last->next = blk;
    ~~~~~~~~~~~^~~~~

Change the "if" check to make it easier for compilers to recognize that "last" will be initialized (and non-NULL even!).

---
 libcommon/usdt_parser.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
libcommon/usdt_parser.c +2 -2
index 864198098..d8cb9b7ba 100644
@@ -163,7 +163,7 @@ usdt_destroy_data(usdt_data_t *data)
 usdt_data_t *
 usdt_copyin_data(int in, int out, int *ok)
 {
-	usdt_data_t *first = NULL, *last;
+	usdt_data_t *first = NULL, *last = NULL;
 	size_t cnt;
 
 	*ok = 1;
@@ -185,7 +185,7 @@ usdt_copyin_data(int in, int out, int *ok)
 		if ((blk = usdt_copyin_block(in, out, ok)) == NULL)
 			goto err;
 
-		if (first == NULL)
+		if (last == NULL)
 			first = last = blk;
 		else {
 			last->next = blk;
Signature · 1 line
2.47.3
#1 lore raw

PATCH 2/2 Use a consistent type for dtrace_consume()

From: Eugene Loh <[email protected]>

For a long time, the type of dtrace_consume() has been
- int according to libdtrace/dtrace.h
- dtrace_workstatus_t according to libdtrace/dt_consume.c

With some recent compilers, however, this triggers the warning (redacted here):

libdtrace/dt_consume.c:3041:1: warning: conflicting types for ‘dtrace_consume’ due to enum/integer mismatch; have ‘dtrace_workstatus_t(...)’ [-Wenum-int-mismatch]
3041 | dtrace_consume(...)
| ^~~~~~~~~~~~~~
In file included from libdtrace/dt_impl.h:14,
from libdtrace/dt_consume.c:16:
libdtrace/dtrace.h:210:12: note: previous declaration of ‘dtrace_consume’ with type ‘int(...)’
210 | extern int dtrace_consume(...)
| ^~~~~~~~~~~~~~

which is a nuisance.

Note that dtrace_consume() is called from only one site, where its value is compared to DTRACE_WORKSTATUS_ERROR, which is an argument for the dtrace_workstatus_t type.

On the other hand, dtrace_consume() is defined to return a variety of values, like 0, dt_set_errno(), and dt_consume_cpu(), all of which are int, but also DTRACE_WORKSTATUS_OKAY, DTRACE_WORKSTATUS_ERROR, and rval, all of which are dtrace_workstatus_t. But then rval itself is set to dtrace_workstatus_t dt_consume_begin() or int dt_consume_cpu(). So, there is simply no consistency here.

Having the prototype be dtrace_workstatus_t requires some amount of code refactoring.

Just change the definition to int and clean up the warning.

---
 libdtrace/dt_consume.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
libdtrace/dt_consume.c +1 -1
index 07b19d498..d9be563d9 100644
@@ -3037,7 +3037,7 @@ dt_consume_fini(dtrace_hdl_t *dtp)
 	dt_htab_destroy(dtp->dt_spec_bufs);
 }
 
-dtrace_workstatus_t
+int
 dtrace_consume(dtrace_hdl_t *dtp, FILE *fp, dtrace_consume_probe_f *pf,
 	       dtrace_consume_rec_f *rf, void *arg)
 {
Signature · 1 line
2.47.3
Reviewed-by on v1 2/2
#2 lore raw

Re: [DTrace-devel] [PATCH 2/2] Use a consistent type for dtrace_consume()

Quoted 34 lines · eugene.loh--- via DTrace-devel

From: Eugene Loh <[email protected]>

For a long time, the type of dtrace_consume() has been
- int according to libdtrace/dtrace.h
- dtrace_workstatus_t according to libdtrace/dt_consume.c

With some recent compilers, however, this triggers the warning (redacted here):

    libdtrace/dt_consume.c:3041:1: warning: conflicting types
    for ???dtrace_consume??? due to enum/integer mismatch;
    have ???dtrace_workstatus_t(...)??? [-Wenum-int-mismatch]
     3041 | dtrace_consume(...)
          | ^~~~~~~~~~~~~~
    In file included from libdtrace/dt_impl.h:14,
                     from libdtrace/dt_consume.c:16:
    libdtrace/dtrace.h:210:12: note: previous declaration of
    ???dtrace_consume??? with type ???int(...)???
      210 | extern int dtrace_consume(...)
          |            ^~~~~~~~~~~~~~

which is a nuisance.

Note that dtrace_consume() is called from only one site, where its value is compared to DTRACE_WORKSTATUS_ERROR, which is an argument for the dtrace_workstatus_t type.

On the other hand, dtrace_consume() is defined to return a variety of values, like 0, dt_set_errno(), and dt_consume_cpu(), all of which are int, but also DTRACE_WORKSTATUS_OKAY, DTRACE_WORKSTATUS_ERROR, and rval, all of which are dtrace_workstatus_t. But then rval itself is set to dtrace_workstatus_t dt_consume_begin() or int dt_consume_cpu(). So, there is simply no consistency here.

Quoted 6 lines

Having the prototype be dtrace_workstatus_t requires some amount of code refactoring.

Just change the definition to int and clean up the warning.

... I admit this was my fault. I tried to refactor the code so that we could
have dtrace_consume() return dtrace_workstatus_t, but only got this far. I agree that for now, changing the return type of dtrace_consume() is the best action to get rid of the compiler warning. Maybe we can revisit the actual refactoring work later to clean this up properly.

Quoted 25 lines
---
 libdtrace/dt_consume.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
libdtrace/dt_consume.c +1 -1
index 07b19d498..d9be563d9 100644
@@ -3037,7 +3037,7 @@ dt_consume_fini(dtrace_hdl_t *dtp)
 	dt_htab_destroy(dtp->dt_spec_bufs);
 }
 
-dtrace_workstatus_t
+int
 dtrace_consume(dtrace_hdl_t *dtp, FILE *fp, dtrace_consume_probe_f *pf,
 	       dtrace_consume_rec_f *rf, void *arg)
 {
Signature · 7 lines
2.47.3


_______________________________________________
DTrace-devel mailing list
[email protected]
https://oss.oracle.com/mailman/listinfo/dtrace-devel
#3 lore raw
In reply to [email protected]
Quoted 4 lines · On Tue, Aug 12, 2025 at 06:46:05PM -0400, [email protected] wrote:

From: Eugene Loh <[email protected]>

Some compilers warn:

Quoted 8 lines

libcommon/usdt_parser.c: In function ???usdt_copyin_data???:
libcommon/usdt_parser.c:191:15: warning:

   ???last??? may be used uninitialized in this function [-Wmaybe-uninitialized]
    last->next = blk;
    ~~~~~~~~~~~^~~~~

Change the "if" check to make it easier for compilers to recognize that "last" will be initialized (and non-NULL even!).

I disagree... What compiler version reported this as a warning? The warning shows a limitation of the compiler to see that last can actually never be used uninitialized.

I don't think we should make changes like these to accomodate compielrs that are less advanced. We generally expect systems to be updated to the most recent version of packages so that would include the compiler.

Quoted 30 lines
---
 libcommon/usdt_parser.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
libcommon/usdt_parser.c +2 -2
index 864198098..d8cb9b7ba 100644
@@ -163,7 +163,7 @@ usdt_destroy_data(usdt_data_t *data)
 usdt_data_t *
 usdt_copyin_data(int in, int out, int *ok)
 {
-	usdt_data_t *first = NULL, *last;
+	usdt_data_t *first = NULL, *last = NULL;
 	size_t cnt;
 
 	*ok = 1;
@@ -185,7 +185,7 @@ usdt_copyin_data(int in, int out, int *ok)
 		if ((blk = usdt_copyin_block(in, out, ok)) == NULL)
 			goto err;
 
-		if (first == NULL)
+		if (last == NULL)
 			first = last = blk;
 		else {
 			last->next = blk;
Signature · 1 line
2.47.3
author
#4 lore raw
Quoted 11 lines · Kris Van Hees
Quoted 10 lines · On Tue, Aug 12, 2025 at 06:46:05PM -0400, [email protected] wrote:

From: Eugene Loh <[email protected]>

Some compilers warn:

libcommon/usdt_parser.c: In function ???usdt_copyin_data???:
libcommon/usdt_parser.c:191:15: warning:

    ???last??? may be used uninitialized in this function [-Wmaybe-uninitialized]
     last->next = blk;
     ~~~~~~~~~~~^~~~~
Quoted 5 lines
Quoted 2 lines

Change the "if" check to make it easier for compilers to recognize that "last" will be initialized (and non-NULL even!).

I disagree... What compiler version reported this as a warning? The warning shows a limitation of the compiler to see that last can actually never be used uninitialized.

OL8 with "yum update" then "make".  Looks like
gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-26.0.1)
Should I be using a different recipe?

At least in my opinion, the new code with this patch is simply cleaner.

Quoted 34 lines

I don't think we should make changes like these to accomodate compielrs that are less advanced. We generally expect systems to be updated to the most recent version of packages so that would include the compiler.

Quoted 30 lines
---
  libcommon/usdt_parser.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
libcommon/usdt_parser.c +2 -2
index 864198098..d8cb9b7ba 100644
@@ -163,7 +163,7 @@ usdt_destroy_data(usdt_data_t *data)
  usdt_data_t *
  usdt_copyin_data(int in, int out, int *ok)
  {
-	usdt_data_t *first = NULL, *last;
+	usdt_data_t *first = NULL, *last = NULL;
  	size_t cnt;
 
  	*ok = 1;
@@ -185,7 +185,7 @@ usdt_copyin_data(int in, int out, int *ok)
  		if ((blk = usdt_copyin_block(in, out, ok)) == NULL)
  			goto err;
 
-		if (first == NULL)
+		if (last == NULL)
  			first = last = blk;
  		else {
  			last->next = blk;
Signature · 1 line
2.47.3
#5 lore raw
Quoted 15 lines · Eugene Loh
Quoted 13 lines · Kris Van Hees
Quoted 12 lines · On Tue, Aug 12, 2025 at 06:46:05PM -0400, [email protected] wrote:

From: Eugene Loh <[email protected]>

Some compilers warn:

libcommon/usdt_parser.c: In function ???usdt_copyin_data???:
libcommon/usdt_parser.c:191:15: warning:

    ???last??? may be used uninitialized in this function [-Wmaybe-uninitialized]
     last->next = blk;
     ~~~~~~~~~~~^~~~~

Change the "if" check to make it easier for compilers to recognize that "last" will be initialized (and non-NULL even!).

Quoted 7 lines
Quoted 3 lines

I disagree... What compiler version reported this as a warning? The warning shows a limitation of the compiler to see that last can actually never be used uninitialized.

OL8 with "yum update" then "make".  Looks like
gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-26.0.1)
Should I be using a different recipe?

8.5.0 is definitely an old compiler. So that is not entirely unexpected.

Quoted 1 line

At least in my opinion, the new code with this patch is simply cleaner.

I guess that will always be subjective... I really prefer the original code, because it captures IMHO better that last depends on first, i.e. there is no concept of a last block until there is at least a first block, since last is the most recently added block (which cannot exist until there has been a first). With the new code, while it avoids a warning on an older compiler, you make the conditional operate on last, which seems counter-intuitive to me for a construct where you have a 'first block' and 'more recently added block'.

YMMV

Quoted 32 lines
Quoted 32 lines

I don't think we should make changes like these to accomodate compielrs that are less advanced. We generally expect systems to be updated to the most recent version of packages so that would include the compiler.

Quoted 28 lines
---
  libcommon/usdt_parser.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
libcommon/usdt_parser.c +2 -2
index 864198098..d8cb9b7ba 100644
@@ -163,7 +163,7 @@ usdt_destroy_data(usdt_data_t *data)
  usdt_data_t *
  usdt_copyin_data(int in, int out, int *ok)
  {
-	usdt_data_t *first = NULL, *last;
+	usdt_data_t *first = NULL, *last = NULL;
  	size_t cnt;
  	*ok = 1;
@@ -185,7 +185,7 @@ usdt_copyin_data(int in, int out, int *ok)
  		if ((blk = usdt_copyin_block(in, out, ok)) == NULL)
  			goto err;
-		if (first == NULL)
+		if (last == NULL)
  			first = last = blk;
  		else {
  			last->next = blk;
Signature · 1 line
2.47.3