Issue observed with chmod05, fchmod05 test cases

Hi
I am observing this anomaly with these 2 test cases chmod05.c and fchmod05.c.
These 2 test cases are written to test the POSIX standard which says:

"If the calling process is not privileged (Linux: does not have the CAP_FSETID capability), and the group of the file does not match the effective group ID of the process or one of its supplementary group IDs, the S_ISGID bit will be turned off, but this will not cause an error to be returned."

So the test case will try to setup a file in which the GID of the test/tmp file will be different from the effective group ID of the running process. But this criterion is found to be not sufficient condition for the clearing the S_ISGID bit of the file

------------------------------------------------------------
Ex:
There are 2 users nobody (uid=99, gid=99) and bin (uid=1, gid=1) let as assume temporary file as /tmp/zxcv

Test will Pass if

     Set UID/GID of /tmp/zxcv to            nobody->uid & nobody->GID
     Set eUID/eGID of process to           nobody->uid & bin->GID

Test will Fail if

     Set UID/GID of /tmp/zxcv to            nobody->uid & bin->GID
     Set eUID/eGID of process to           nobody->uid & nobody->GID

I verified supplementary group IDs for both 'nobody' and 'bin' and they are mutually exclusive and none of the other is supplementary group ID of the ID being tested
------------------------------------------------------------

Should we treat this as Test case issue or kernel bug?

Thanks
Yeehaw

------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

The different of 'nobody' and 'bin' is bin->GID has the CAP_FSETID capability, but nobody->GID has not, I think.

Regards.

------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Wei Yongjun

I am facing some issues with getcap, how else can I verify the USER capabilities If there is a simpler way to check user capabilities please let me know

My efforts to use getcap fails

getcap /tmp/test1
Failed to get capabilities for file `/tmp/test1'
(Function not implemented)

Thank you

------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

I tried with a small program to check out the capabilities of bin and nobody couldn't find any difference

This is what I tried
Program I used
------------------------------
int main()

{
	cap_t cap = cap_get_proc();
	printf("Running with uid %d\n", getuid());
	printf("Running with capabilities: %s\n", cap_to_text(cap, NULL));
	cap_free(cap);
	return 0;
}

------------------------------
Running the program under different user gave same information

mx3455a:/home/sharyathi # su bin
bin@mx3455a:/home/sharyathi> ./a.out
Running with uid 1
Running with capabilities: =
bin@mx3455a:/home/sharyathi> exit

mx3455a:/home/sharyathi # su nobody
nobody@mx3455a:/home/sharyathi> ./a.out
Running with uid 65534
Running with capabilities: =

Thanks
Sharyathi N

------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Hi
I was able to understand the problem...
This is what I learnt

Supplementary group IDs of a process is inherited from its parent process, so who ever spawns the process transfer its supplementary group ids also to its child process.

in our test case scenario checking the supplementary group IDs of root which is spawning the test cases fchmod5 and chmod5 shows this

---------------------------------------------------------
cat test.c
---------------------------------------------------------
int main()

{
        int size = 10;
        int i, index;
        gid_t list[10];
        index = getgroups(size, list);
        for ( i =0 ;i < index; i++)
                printf("\n list[%d]:%d", i, list[i]);
        printf("\n");
}

---------------------------------------------------------

[root@mx3350a sharyathi]# ./a.out

list[0]:0
list[1]:1
list[2]:2
list[3]:3
list[4]:4
list[5]:6
list[6]:10
---------------------------------------------------------
This shows that bin->GID (1) is one of the supplementary group IDs of the process while nobody->GID (99) is not.
So in the kernel code
--------------------------------------------------------

 if (ia_valid & ATTR_MODE) {
                umode_t mode = attr->ia_mode;

                if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
                        mode &= ~S_ISGID;
                inode->i_mode = mode;
        }

--------------------------------------------------------
S_ISGID is cleared in case file attribute is nobody and not in case of GID being bin

I was thinking of what could be the best solution as earlier solution proposed are not fool proof
Attaching a fix I though is the best way to handle the issue

Thanks
Yeehaw

Index: ltp-full-20090430/testcases/kernel/syscalls/chmod/chmod05.c
===================================================================
@@ -240,6 +240,10 @@
 		tst_brkm(TBROK, cleanup, "mkdir(2) of %s failed: %s", TESTDIR,
 				strerror(errno));
 	}
+	
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
 
 	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't change owner of testdir: %s",

Index: ltp-full-20090430/testcases/kernel/syscalls/fchmod/fchmod05.c
===================================================================
@@ -218,6 +218,10 @@
 		tst_brkm(TBROK, cleanup, "mkdir(2) of %s failed", TESTDIR);
 	}
 
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
+
 	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't change owner of testdir: %s",
 				strerror(errno));

------------------------------------------------------------------------------
Enter the BlackBerry Developer Challenge
This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge

_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Hi Yeehaw,
1. You didn't provide sign-off, as noted in the DCO:
http://ltp.sourceforge.net/DCO.txt
2. This appears to be a legitimate documentation bug (at least as
far as the manpages for 2.6.29-r5 are concerned), because I can't find a single reference to nobody in the manpage, as far as special importance is concerned. Furthermore, looking at the OpenGroup's documentation, I don't see anything there either:
http://www.opengroup.org/onlinepubs/000095399/functions/getgroups.html
. So, I'm a bit stumped as to whether or not we'd be doing the right thing by `fixing' this test.
Thanks,
-Garrett

------------------------------------------------------------------------------
Enter the BlackBerry Developer Challenge
This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Garret

Excuse me for missing that out, if this patch format is okie I will resend it with the sign-off included in it

I am not very clear with what you are pointing at, but let me take a shot. bin and nobody
are 2 users used in the test case. I am not sure who or where it is getting created, but what I did observe is that they don't have the same privileges. GID of bin is in the list of supplementary group ids of root while nobody is not. If this is the case chmod/fchmod won't clear S_IGSID bit of the file, as you can read from man page of chmod/fchmod.
Nobody won't be in any man page as this User was created by LTP, I suppose, to use in this
test case.
Thanks
Sharyathi N

------------------------------------------------------------------------------
Enter the BlackBerry Developer Challenge
This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Sharyathi, please resend this clean.

Regards--
Subrata

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Subrata

Attaching it, with signed off added
Thanks
Sharyathi Nagesh

Index: ltp-full-20090430/testcases/kernel/syscalls/chmod/chmod05.c
===================================================================
@@ -240,6 +240,10 @@
 		tst_brkm(TBROK, cleanup, "mkdir(2) of %s failed: %s", TESTDIR,
 				strerror(errno));
 	}
+	
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
 
 	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't change owner of testdir: %s",

Index: ltp-full-20090430/testcases/kernel/syscalls/fchmod/fchmod05.c
===================================================================
@@ -218,6 +218,10 @@
 		tst_brkm(TBROK, cleanup, "mkdir(2) of %s failed", TESTDIR);
 	}
 
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
+
 	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't change owner of testdir: %s",
 				strerror(errno));

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july

_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Sharyathi,

This is good. But, what i wanted is:

1) Final and clear description of the purpose of the patch,
2) Signed-off-by: clause,
3) Patch itself,

Itś just like how you would resend a patch to LKML, saying:

[PATCH v2][RESEND] ...

Regards--
Subrata

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Subrata
Resending the patch, after the modification. Let me know if this is acceptable
Thanks
Yeehaw

This Patch fixes bug in the test cases chmod05.c and fchmod05.c. The test case uses 2 UserIds nobody and bin. The issue is observed when one of the UserId is the supplementary group Ids of the root. In that case the Posix standard mentions that S_IGSID bit should not be cleared by the systemcall causing the test to fail. This test case fixes the issue by specifically clearing the supplementary group ID of the process.The main intention of the test case is to test if S_IGSID bit is cleared in case the non privilage user tries to set file modes and hence clearing the supplementary group Ids.

Index: ltp-full-20090430/testcases/kernel/syscalls/chmod/chmod05.c
===================================================================
@@ -240,6 +240,10 @@
 		tst_brkm(TBROK, cleanup, "mkdir(2) of %s failed: %s", TESTDIR,
 				strerror(errno));
 	}
+	
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
 
 	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't change owner of testdir: %s",

Index: ltp-full-20090430/testcases/kernel/syscalls/fchmod/fchmod05.c
===================================================================
@@ -218,6 +218,10 @@
 		tst_brkm(TBROK, cleanup, "mkdir(2) of %s failed", TESTDIR);
 	}
 
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
+
 	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't change owner of testdir: %s",
 				strerror(errno));

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july

_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

It fails to Apply:

patching file testcases/kernel/syscalls/chmod/chmod05.c Hunk #1 succeeded at 240 with fuzz 2.
patching file testcases/kernel/syscalls/fchmod/fchmod05.c Hunk #1 FAILED at 218.
1 out of 1 hunk FAILED -- saving rejects to file testcases/kernel/syscalls/fchmod/fchmod05.c.rej

Regards--
Subrata

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Subrata
Excuse me for the late response, I have update the patch as well as the description.
Please re-verify
Thanks
Sharyathi

This Patch fixes bug in the test cases chmod05.c and fchmod05.c. The test case uses 2 UserIds nobody and bin. The issue is observed when one of the UserId is the supplementary group Ids of the rooti (either nobody or bin). In that case the Posix standard mentions that S_IGSID bit should not be cleared by the systemcall, chmod() or fchmod(), which causes the test to fail.
This patch fixes the issue by specifically clearing the supplementary group ID of the process. The main intention of the test case is to test if S_IGSID bit is cleared in case the non privilage user tries to set file modes. Hence felt clearing the supplementary group Ids is the best way to fix the issue. At the same time I am reverting the earlier patch submitted by Wei Yongjun <[email protected]>

Index: ltp-full-20090831/testcases/kernel/syscalls/fchmod/fchmod05.c
===================================================================
@@ -218,12 +218,16 @@
 		tst_brkm(TBROK, cleanup, "mkdir(2) of %s failed", TESTDIR);
 	}
 
-	if (chown(TESTDIR, nobody_u->pw_uid, nobody_u->pw_gid) == -1)
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
+
+	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't change owner of testdir: %s",
 				strerror(errno));
 
-	/* change to nobody:bin */
-	if (setegid(bin_group->gr_gid) == -1 || seteuid(nobody_u->pw_uid) == -1)
+	/* change to nobody:nobody */
+	if (setegid(nobody_u->pw_gid) == -1 || seteuid(nobody_u->pw_uid) == -1)
 		tst_brkm(TBROK, cleanup, "Couldn't switch to nobody:nobody: %s",
 				strerror(errno));
 

Index: ltp-full-20090831/testcases/kernel/syscalls/chmod/chmod05.c
===================================================================
@@ -236,11 +236,15 @@
 	if (mkdir(TESTDIR, MODE_RWX) < 0)
 		tst_brkm(TBROK|TERRNO, cleanup, "mkdir(%s) failed", TESTDIR);
 
-	if (chown(TESTDIR, nobody_u->pw_uid, nobody_u->pw_gid) == -1)
+	if(setgroups(1, &nobody_u->pw_gid) == -1)
+		tst_brkm(TBROK, cleanup, "Couldn't change supplementary group Id: %s",
+				strerror(errno));
+
+	if (chown(TESTDIR, nobody_u->pw_uid, bin_group->gr_gid) == -1)
 		tst_brkm(TBROK|TERRNO, cleanup, "chown() of testdir failed");
 
-	/* change to nobody:bin */
-	if (setegid(bin_group->gr_gid) == -1 ||
+	/* change to nobody:nobody */
+	if (setegid(nobody_u->pw_gid) == -1 || 
 		 seteuid(nobody_u->pw_uid) == -1)
 		tst_brkm(TBROK|TERRNO, cleanup, "Couldn't switch to nobody:nobody");
 }

------------------------------------------------------------------------------
Come build with us! The BlackBerry&reg; Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9&#45;12, 2009. Register now&#33;
http://p.sf.net/sfu/devconf

_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list

Thanks.

Regards--
Subrata

------------------------------------------------------------------------------
Come build with us! The BlackBerry&reg; Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9&#45;12, 2009. Register now&#33;
http://p.sf.net/sfu/devconf
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list