Skip to contentSkip to conversation

iscsi_add_session() warning

DISCUSSION Andrew Morton 4 participants 6 messages Feb 4, 2009 SCSI

lore
Share & tools
lore
#0 lore raw

Looking at this:

drivers/scsi/scsi_transport_iscsi.c: In function 'iscsi_add_session':
drivers/scsi/scsi_transport_iscsi.c:703: warning: 'err' may be used uninitialized in this function

the compiler is wrong - `err' will always have been initialised to _something_.

But to what?

: int iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id) : {
: struct Scsi_Host *shost = iscsi_session_to_shost(session);
: struct iscsi_cls_host *ihost;
: unsigned long flags;
: unsigned int id = target_id;
: int err;
:
: ihost = shost->shost_data;
: session->sid = atomic_add_return(1, &iscsi_session_nr);
:
: if (id == ISCSI_MAX_TARGET) {
: for (id = 0; id < ISCSI_MAX_TARGET; id++) {
: err = device_for_each_child(&shost->shost_gendev, &id,
: iscsi_get_next_target_id);
: if (!err)
: break;
: }
:
: if (id == ISCSI_MAX_TARGET) {
: iscsi_cls_session_printk(KERN_ERR, session,
: "Too many iscsi targets. Max "
: "number of targets is %d.\n",
: ISCSI_MAX_TARGET - 1);

At this point, err is probably zero. Or it might be the return value from the most recent call to device_for_each_child(iscsi_get_next_target_id).

Should we not assign some error code to `err' here?

: goto release_host;
: }
: }
: session->target_id = id;
:
: dev_set_name(&session->dev, "session%u", session->sid);
: err = device_add(&session->dev);
: if (err) {
: iscsi_cls_session_printk(KERN_ERR, session,
: "could not register session's dev\n");
: goto release_host;
: }
: transport_register_device(&session->dev);
:
: spin_lock_irqsave(&sesslock, flags);
: list_add(&session->sess_list, &sesslist);
: spin_unlock_irqrestore(&sesslock, flags);
:
: iscsi_session_event(session, ISCSI_KEVENT_CREATE_SESSION);
: return 0;
:
: release_host:
: scsi_host_put(shost);
: return err;
: }

#1 lore raw
Quoted 3 lines · Andrew Morton

: if (id == ISCSI_MAX_TARGET) {
: for (id = 0; id < ISCSI_MAX_TARGET; id++) {
: err = device_for_each_child(&shost->shost_gendev, &id,

Possibly GCC just can't figure out that id was reinitialized, so it thinks that the for loop won't be executed?

Just a guess...
Kyle

#2 lore raw
Quoted 8 lines · On Tue, 3 Feb 2009 19:12:38 -0500 Kyle McMartin <[email protected]> wrote:
Quoted 3 lines · Andrew Morton

: if (id == ISCSI_MAX_TARGET) {
: for (id = 0; id < ISCSI_MAX_TARGET; id++) {
: err = device_for_each_child(&shost->shost_gendev, &id,

Possibly GCC just can't figure out that id was reinitialized, so it thinks that the for loop won't be executed?

Could be. But that isn't the point...

The point, dear Kyle, is that I misread the code :(

In fact on that codepath the function _will_ return the most recent return value from device_for_each_child(iscsi_get_next_target_id), and that value will be non-zero (-EEXIST). So it looks non-buggy, albeit rather obscure.

But did we intend to return -EEXIST in the "Too many iscsi targets" case?

#3 lore raw
Quoted 13 lines · Andrew Morton
Quoted 7 lines · On Tue, 3 Feb 2009 19:12:38 -0500 Kyle McMartin <[email protected]> wrote:
Quoted 3 lines · Andrew Morton

: if (id == ISCSI_MAX_TARGET) {
: for (id = 0; id < ISCSI_MAX_TARGET; id++) {
: err = device_for_each_child(&shost->shost_gendev, &id,

Possibly GCC just can't figure out that id was reinitialized, so it thinks that the for loop won't be executed?

Could be. But that isn't the point...

Quoted 8 lines

The point, dear Kyle, is that I misread the code :(

In fact on that codepath the function _will_ return the most recent return value from device_for_each_child(iscsi_get_next_target_id), and that value will be non-zero (-EEXIST). So it looks non-buggy, albeit rather obscure.

But did we intend to return -EEXIST in the "Too many iscsi targets" case?

I did not. The callers just check for non zero so it will work. It would be best to change it to something meaningful.

When there are no ids for some object left, what is the stadard Exxxx value to return? Would ENOSPC make sense? Here ENOSPC would mean no space in id space left instead of space on a device?

#4 lore raw
Quoted 3 lines · On Wed, 04 Feb 2009 01:40:11 -0600 Mike Christie <[email protected]> wrote:

When there are no ids for some object left, what is the stadard Exxxx value to return? Would ENOSPC make sense? Here ENOSPC would mean no space in id space left instead of space on a device?

Drivers often do that sort of thing. I don't like it personally - ENOSPC means "No space left on device". IOW, "your disk is full".

Co-opting that code just because something kinda-sorta-similar happened seems abusive to me. Plus the first thing poor old user is going to do is run `df' and ask wtf?

So what's left? EINVAL is a sort of generic something-went-wrong error code (in Linux, at least). But perhaps EBUSY ("Device or resource busy") is suitable here.

Dunno, hard.

#5 lore raw
In reply to Kyle McMartin
Quoted 7 lines · Kyle McMartin
Quoted 3 lines · Andrew Morton

: if (id == ISCSI_MAX_TARGET) {
: for (id = 0; id < ISCSI_MAX_TARGET; id++) {
: err = device_for_each_child(&shost->shost_gendev, &id,

Possibly GCC just can't figure out that id was reinitialized, so it thinks that the for loop won't be executed?

Would restructuring the code slightly be ok?

  if (target_id ==  ISCSI_MAX_TARGET) {
     unsigned int id = 0;
...
     session->target_id = id;
  } else {
     session->target_id = target_id;

}
...

ie. reduce the scope of "id" . Not quite a concise as existing code but just as readable.

hth,
grant

Quoted 7 lines

Just a guess...
Kyle
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html