scsi: scsi_transport_iscsi.c initialize the variable before it get used
PATCH 1/1 scsi: scsi_transport_iscsi.c initialize the variable before it get used
I am watching this problem from long time in -tip.
Subject: [PATCH] scsi: scsi_transport_iscsi.c initialize the variable before it get used
In some cases, err will be used uninitialized.
Also fixed compilation warning :
CC drivers/scsi/scsi_transport_iscsi.odrivers/scsi/scsi_transport_iscsi.c: In function ‘iscsi_add_session’:drivers/scsi/scsi_transport_iscsi.c:678: warning: ‘err’ may be used uninitialized in this function
--- drivers/scsi/scsi_transport_iscsi.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/scsi/scsi_transport_iscsi.c +1 -1
@@ -675,7 +675,7 @@ int iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id)struct iscsi_cls_host *ihost;unsigned long flags;unsigned int id = target_id; -int err; +int err =0 ; ihost = shost->shost_data; session->sid =atomic_add_return (1 , &iscsi_session_nr);
Signature · 8 lines
1.6.0.6 -- 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
James Bottomley
maintainerQuoted 5 lines · Jaswinder Singh Rajput
I am watching this problem from long time in -tip.
Subject: [PATCH] scsi:
scsi_transport_iscsi.cinitialize the variable before it get usedIn some cases, err will be used uninitialized.
What cases? A simple theoretical analysis shows that this variable is always set.
Quoted 4 lines
CC
drivers/scsi/scsi_transport_iscsi.odrivers/scsi/scsi_transport_iscsi.c: In function ‘iscsi_add_session’:drivers/scsi/scsi_transport_iscsi.c:678: warning: ‘err’ may be used uninitialized in this function
My version of gcc (4.3.3) correctly sees that the variable cannot be uninitialised ... what version are you using?
If it's a popular version, we can always do the uninitialised_var() thing, but if it's just a non-standard compiler, I'd rather not mess up the source code to please gcc.
James
--
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
Hello James,
Quoted 10 lines · James Bottomley
Quoted 5 lines · Jaswinder Singh Rajput
I am watching this problem from long time in -tip.
Subject: [PATCH] scsi:
scsi_transport_iscsi.cinitialize the variable before it get usedIn some cases, err will be used uninitialized.
What cases? A simple theoretical analysis shows that this variable is always set.
Ok lets takes your tree:
672 int iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id) 673 {
674 struct Scsi_Host *shost = iscsi_session_to_shost(session); 675 struct iscsi_cls_host *ihost; 676 unsigned long flags; 677 unsigned int id = target_id; 678 int err;
679
680 ihost = shost->shost_data;
681 session->sid = atomic_add_return(1, &iscsi_session_nr);
682
683 if (id == ISCSI_MAX_TARGET) {
684 for (id = 0; id < ISCSI_MAX_TARGET; id++) {
685 err = device_for_each_child(&shost->shost_gendev, &id,
686 iscsi_get_next_target_id);
687 if (!err)
688 break;
689 }
690
691 if (id == ISCSI_MAX_TARGET) {
692 iscsi_cls_session_printk(KERN_ERR, session,
693 "Too many iscsi targets. Max " 694 "number of targets is %d.\n", 695 ISCSI_MAX_TARGET - 1);
696 goto release_host;
697 }
698 }
699 session->target_id = id;
700
701 dev_set_name(&session->dev, "session%u", session->sid);
702 err = device_add(&session->dev);
703 if (err) {704 iscsi_cls_session_printk(KERN_ERR, session,
705 "could not register session's dev\n");
706 goto release_host;
707 }
708 transport_register_device(&session->dev);
709
710 spin_lock_irqsave(&sesslock, flags); 711 list_add(&session->sess_list, &sesslist); 712 spin_unlock_irqrestore(&sesslock, flags);
713
714 iscsi_session_event(session, ISCSI_KEVENT_CREATE_SESSION);
715 return 0;
716
717 release_host:
718 scsi_host_put(shost);
719 return err;
720 }
You are right but compiler is confused and giving warning like this :
1. do not go inside :
684 for (id = 0; id < ISCSI_MAX_TARGET; id++) {
2. and then choose path :
696 goto release_host;
then we return err uninitialized ;-)
Quoted 5 lines
Quoted 4 lines
CC
drivers/scsi/scsi_transport_iscsi.odrivers/scsi/scsi_transport_iscsi.c: In function ‘iscsi_add_session’:drivers/scsi/scsi_transport_iscsi.c:678: warning: ‘err’ may be used uninitialized in this function
Quoted 7 lines
My version of gcc (4.3.3) correctly sees that the variable cannot be uninitialised ... what version are you using?
If it's a popular version, we can always do the
uninitialised_var()thing, but if it's just a non-standard compiler, I'd rather not mess up the source code to please gcc.
gcc version 4.3.2 20081105 (Red Hat 4.3.2-7) (GCC)
--
JSR
--
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
James Bottomley
maintainerQuoted 69 lines · Jaswinder Singh Rajput
Hello James,
Quoted 10 lines · James Bottomley
Quoted 5 lines · Jaswinder Singh Rajput
I am watching this problem from long time in -tip.
Subject: [PATCH] scsi:
scsi_transport_iscsi.cinitialize the variable before it get usedIn some cases, err will be used uninitialized.
What cases? A simple theoretical analysis shows that this variable is always set.
Ok lets takes your tree:
672 int
iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id)673 {674 struct Scsi_Host *shost = iscsi_session_to_shost(session); 675 struct iscsi_cls_host *ihost; 676 unsigned long flags; 677 unsigned int id = target_id; 678 int err;679
680 ihost =shost->shost_data;
681session->sid=atomic_add_return(1, &iscsi_session_nr);
682
683 if (id ==ISCSI_MAX_TARGET) {
684 for (id = 0; id <ISCSI_MAX_TARGET; id++) {
685 err = device_for_each_child(&shost->shost_gendev, &id,
686iscsi_get_next_target_id);
687 if (!err)
688 break;
689 }
690
691 if (id ==ISCSI_MAX_TARGET) {
692 iscsi_cls_session_printk(KERN_ERR, session,693 "Too many iscsi targets. Max " 694 "number of targets is %d.\n", 695 ISCSI_MAX_TARGET - 1);696 goto
release_host;
697 }
698 }
699session->target_id= id;
700701 dev_set_name(&session->dev, "session%u", session->sid); 702 err = device_add(&session->dev); 703 if (err) {704 iscsi_cls_session_printk(KERN_ERR, session,
705 "could not register session's dev\n");
706 gotorelease_host;
707 }
708transport_register_device(&session->dev);
709710 spin_lock_irqsave(&sesslock, flags); 711 list_add(&session->sess_list, &sesslist); 712 spin_unlock_irqrestore(&sesslock, flags);713
714iscsi_session_event(session, ISCSI_KEVENT_CREATE_SESSION);
715 return 0;
716
717release_host:
718scsi_host_put(shost);
719 return err;
720 }
Quoted 4 lines
You are right but compiler is confused and giving warning like this :
1. do not go inside :
684 for (id = 0; id <ISCSI_MAX_TARGET; id++) {
How? id is an unsigned int and ISCSI_MAX_TARGET is defined to be -1 in include/scsi/scsi_transport_iscsi.h (might be better if it were UINT_MAX, but -1 is just as good).
James
Quoted 70 lines · James Bottomley
Quoted 69 lines · Jaswinder Singh Rajput
Hello James,
Quoted 10 lines · James Bottomley
Quoted 5 lines · Jaswinder Singh Rajput
I am watching this problem from long time in -tip.
Subject: [PATCH] scsi:
scsi_transport_iscsi.cinitialize the variable before it get usedIn some cases, err will be used uninitialized.
What cases? A simple theoretical analysis shows that this variable is always set.
Ok lets takes your tree:
672 int
iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id)673 {674 struct Scsi_Host *shost = iscsi_session_to_shost(session); 675 struct iscsi_cls_host *ihost; 676 unsigned long flags; 677 unsigned int id = target_id; 678 int err;679
680 ihost =shost->shost_data;
681session->sid=atomic_add_return(1, &iscsi_session_nr);
682
683 if (id ==ISCSI_MAX_TARGET) {
684 for (id = 0; id <ISCSI_MAX_TARGET; id++) {
685 err = device_for_each_child(&shost->shost_gendev, &id,
686iscsi_get_next_target_id);
687 if (!err)
688 break;
689 }
690
691 if (id ==ISCSI_MAX_TARGET) {
692 iscsi_cls_session_printk(KERN_ERR, session,693 "Too many iscsi targets. Max " 694 "number of targets is %d.\n", 695 ISCSI_MAX_TARGET - 1);696 goto
release_host;
697 }
698 }
699session->target_id= id;
700701 dev_set_name(&session->dev, "session%u", session->sid); 702 err = device_add(&session->dev); 703 if (err) {704 iscsi_cls_session_printk(KERN_ERR, session,
705 "could not register session's dev\n");
706 gotorelease_host;
707 }
708transport_register_device(&session->dev);
709710 spin_lock_irqsave(&sesslock, flags); 711 list_add(&session->sess_list, &sesslist); 712 spin_unlock_irqrestore(&sesslock, flags);713
714iscsi_session_event(session, ISCSI_KEVENT_CREATE_SESSION);
715 return 0;
716
717release_host:
718scsi_host_put(shost);
719 return err;
720 }
Quoted 9 lines
Quoted 4 lines
You are right but compiler is confused and giving warning like this :
1. do not go inside :
684 for (id = 0; id <ISCSI_MAX_TARGET; id++) {How? id is an unsigned int and
ISCSI_MAX_TARGETis defined to be -1 ininclude/scsi/scsi_transport_iscsi.h(might be better if it wereUINT_MAX, but -1 is just as good).
Yes, you are right, I was telling you how compiler was confused in 4.3.2 but I am glad to hear that this problem is fixed in 4.3.3 as you told me :-)
So now its upto you. 4.3.2 is not soo old, So either initialize the variable or stay with stray warning in the kernel ;-)
Thanks,
--
JSR
Quoted 6 lines · Jaswinder Singh Rajput
I am watching this problem from long time in -tip.
Subject: [PATCH] scsi:
scsi_transport_iscsi.cinitialize the variable before it get usedIn some cases, err will be used uninitialized.
As you guys already discussed it will not get used uninitialized.
Quoted 6 lines
Also fixed compilation warning :
CC
drivers/scsi/scsi_transport_iscsi.odrivers/scsi/scsi_transport_iscsi.c: In function ‘iscsi_add_session’:drivers/scsi/scsi_transport_iscsi.c:678: warning: ‘err’ may be used uninitialized in this function
I did the attached patch to silence the compile warning. I goofed and forgot to send it before, because I had change my compiler and was not seeing the error message. I was planning on sending the attached with some other changes for the next feature window. I am going to fix up the patch to use UINT_MAX like James mentioned in the thread.
Quoted 1 line
From
32ffa5248bf11f4df76ebf3c4a023958a9131f9cMon Sep 17 00:00:00 2001
From: Mike Christie <[email protected]>
Date: Mon, 18 May 2009 11:54:37 -0500
Subject: [PATCH] iscsi class: quiet iscsi_add_session compilation warning.
If ISCSI_MAX_TARGET was zero then iscsi_add_session could return success when it had failed. This will currently not happen because ISCSI_MAX_TARGET is a macro and not a variable. But gcc likes to complain about it and people keep reporting it. This patch just initializes err to be -EINVAL so if we ever defined ISCSI_MAX_TARGET incorrectly we would fail gracefully.
--- drivers/scsi/scsi_transport_iscsi.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/scsi/scsi_transport_iscsi.c +1 -1
@@ -675,7 +675,7 @@ int iscsi_add_session(struct iscsi_cls_session *session, unsigned int target_id)struct iscsi_cls_host *ihost;unsigned long flags;unsigned int id = target_id; -int err; +int err = -EINVAL ; ihost = shost->shost_data; session->sid =atomic_add_return (1 , &iscsi_session_nr);
Signature · 1 line
1.6.0.6