Re: [LTP] [PATCH v4 1/1] mount03: Convert to new API

[Richard Palethorpe <rpalethorpe@suse.de>]

Hello,

"xuyang2018.jy@fujitsu.com" <xuyang2018.jy@fujitsu.com> writes:

> Hi Richard, Petr
>
>> 
>> "xuyang2018.jy@fujitsu.com" <xuyang2018.jy@fujitsu.com> writes:
>> 
>>> Hi Petr
>>>
>>>> Hi all,
>>>>
>>>> when running in loop, I sometimes still get error:
>>>>
>>>> i=0; while true; do i=$((i+1)); echo "== $i =="; ./mount03  || break; done
>>>> ...
>>>> === 100 ===
>>>> ...
>>>> tst_device.c:434: TINFO: No device is mounted at mntpoint
>>>> tst_test.c:1599: TINFO: Testing on ext4
>>>> tst_test.c:1065: TINFO: Formatting /dev/loop0 with ext4 opts='' extra opts=''
>>>> mke2fs 1.46.5 (30-Dec-2021)
>>>> mount03.c:150: TINFO: Testing flag MS_RDONLY
>>>> mount03.c:48: TPASS: otfd = open(file, O_CREAT | O_RDWR, 0700) : EROFS (30)
>>>> mount03.c:150: TINFO: Testing flag MS_NODEV
>>>> mount03.c:153: TFAIL: mount(tst_device->dev, MNTPOINT, tst_device->fs_type, tc->flag, NULL) failed: EINVAL (22)
>>>> mount03.c:150: TINFO: Testing flag MS_NOEXEC
>>>> mount03.c:153: TFAIL: mount(tst_device->dev, MNTPOINT, tst_device->fs_type, tc->flag, NULL) failed: EINVAL (22)
>>>> mount03.c:150: TINFO: Testing flag MS_RDONLY
>>>> mount03.c:153: TFAIL: mount(tst_device->dev, MNTPOINT, tst_device->fs_type, tc->flag, NULL) failed: EINVAL (22)
>>>> mount03.c:150: TINFO: Testing flag MS_NOSUID
>>>> mount03.c:153: TFAIL: mount(tst_device->dev, MNTPOINT, tst_device->fs_type, tc->flag, NULL) failed: EINVAL (22)
>>>> mount03.c:150: TINFO: Testing flag MS_NOATIME
>>>> mount03.c:111: TPASS: st.st_atime == atime (1661183501)
>>>
>>> I also meet nosuid work not as expected.
>>>
>>> st_test.c:1599: TINFO: Testing on ext4
>>> tst_test.c:1064: TINFO: Formatting /dev/loop0 with ext4 opts='' extra
>>> opts=''
>>> mke2fs 1.46.5 (30-Dec-2021)
>>> mount03.c:151: TINFO: Testing flag MS_RDONLY
>>> mount03.c:48: TPASS: otfd = open(file, O_CREAT | O_RDWR, 0700) : EROFS (30)
>>> mount03.c:151: TINFO: Testing flag MS_NODEV
>>> mount03.c:55: TPASS: otfd = open(file, O_RDWR, 0700) : EACCES (13)
>>> mount03.c:151: TINFO: Testing flag MS_NOEXEC
>>> mount03.c:63: TPASS: execlp(file, basename(file), NULL) : EACCES (13)
>>> mount03.c:151: TINFO: Testing flag MS_RDONLY
>>> mount03.c:70: TPASS: otfd = open(file, O_CREAT | O_RDWR, 0700) returned fd 3
>>> mount03.c:151: TINFO: Testing flag MS_NOSUID
>>> mount03_suid_child.c:22: TFAIL: setreuid(getuid(), 0) succeeded
>>> mount03.c:151: TINFO: Testing flag MS_NOATIME
>>> mount03.c:112: TPASS: st.st_atime == atime (1661280863)
>>> tst_device.c:434: TINFO: No device is mounted at mntpoint
>>>
>> 
>> I added some more debug info:
>> 
>> @@ -86,7 +89,20 @@ static void test_nosuid(void)
>>   		if (st.st_mode != SUID_MODE)
>>   			SAFE_CHMOD(BIN_PATH, SUID_MODE);
>>   
>> -		SAFE_SETREUID(nobody_uid, nobody_uid);
>> +		tst_res(TINFO, "dev %d:%d, rdev: %d:%d",
>> +			major(st.st_dev), minor(st.st_dev), major(st.st_rdev), minor(st.st_rdev));
>> +		SAFE_SETRESGID(nobody_gid, nobody_gid, nobody_gid);
>> +		SAFE_SETRESUID(nobody_uid, nobody_uid, nobody_uid);
>> +
>> +		uid_t uid, gid, euid, egid, suid, sgid;
>> +
>> +		getresuid(&uid, &euid, &suid);
>> +		getresgid(&gid, &egid, &sgid);
>> +		tst_res(TINFO, "uid: %d, gid: %d, euid: %d, egid: %d, suid: %d, sgid: %d",
>> +			uid, gid, euid, egid, suid, sgid);
>> +
>> +		tst_system("pwd");
>> +		tst_system("ls -l " MNTPOINT);
>>   		SAFE_EXECL(BIN_PATH, BIN_PATH, NULL);
>>   		tst_brk(TFAIL | TTERRNO, "Failed to execute %s", BIN_PATH);
>> 
>> and in the child
>> 
>>   int main(void)
>>   {
>> +	uid_t uid, gid, euid, egid, suid, sgid;
>>   	tst_reinit();
>>   
>> +	getresuid(&uid, &euid, &suid);
>> +	getresgid(&gid, &egid, &sgid);
>> +	tst_res(TINFO, "uid: %d, gid: %d, euid: %d, egid: %d, suid: %d, sgid: %d",
>> +		uid, gid, euid, egid, suid, sgid);
>> +
>>   	TST_EXP_FAIL(setreuid(getuid(), 0), EPERM);
>>   
>> 
>> This shows that BIN_PATH is being accessed from the wrong device AFAICT
>> 
>> mount03.c:166: TINFO: Testing flag MS_NOSUID
>> 
>> ...
>> 
>> 30 1 0:26 / /tmp rw,relatime - tmpfs ltp rw,size=512000k,inode64
>> 31 30 0:27 / /tmp/mouLkSN2v/mntpoint rw,nosuid,relatime - tmpfs /dev/loop0 rw,inode64
>> mount03.c:92: TINFO: dev 0:26, rdev: 0:0
>> mount03.c:101: TINFO: uid: 65534, gid: 65534, euid: 65534, egid: 65534, suid: 65534, sgid: 65534
>> /tmp/mouLkSN2v
>> total 688
>> -r-s--x--x 1 root 0 702952 Oct 13 11:29 mount03_suid_child
>> mount03_suid_child.c:25: TINFO: uid: 65534, gid: 65534, euid: 0, egid: 65534, suid: 0, sgid: 65534
>> mount03_suid_child.c:28: TFAIL: setreuid(getuid(), 0) succeeded
>> 
>> The device should be 0:27 not 0:26.
>> 
>> Indeed if I remove SAFE_CHMOD then the euid and suid are set to
>> nobody. Also if I mount /tmp as nosuid then the test passes. Possibly we
>> need to use absoute paths to ensure we are going through the mount
>> point.
>
> I use statfs struct f_flags member to debug, then found the mounted 
> filesystem doesn't have nosuid flag.
>
> It seems parent has umounted this dev and so child doesn't have nosuid 
> flag, so this case failed.
>
> --- a/testcases/kernel/syscalls/mount/mount03.c
> +++ b/testcases/kernel/syscalls/mount/mount03.c
> @@ -90,6 +90,7 @@ static void test_nosuid(void)
>                  SAFE_EXECL(BIN_PATH, BIN_PATH, NULL);
>                  tst_brk(TFAIL | TTERRNO, "Failed to execute %s", BIN_PATH);
>          }
> +       tst_reap_children();
>   }
>
>
> After this change, this case can pass when I run with -i 10.

Good catch; merged with this fix and a couple of others.

-- 
Thank you,
Richard.

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp