[LTP] [PATCHv2 0/2] Fix for syscalls/utimensat01 test on Ubuntu 4.4 kernel

[LTP] [PATCHv2 0/2] Fix for syscalls/utimensat01 test on Ubuntu 4.4 kernel

[Po-Hsu Lin]

Ubuntu 4.4 kernel got the backported kernel patch (b3b4283) for
syscalls/utimensat01 since 4.4.0-48.69, therefore it will return
EPERM instead of EACCES.

Without this fix, the test will fail on Ubuntu 4.4 kernel with:
  FAIL: 18: utimensat() failed with incorrect error, expected EACCES: EPERM (1)
  FAIL: 19: utimensat() failed with incorrect error, expected EACCES: EPERM (1)

Add Ubuntu kernel version comparison to fix this.

Also add support to get the distname for different OS by parsing ID=
from /etc/os-release in tst_kvcmp_distname()

Po-Hsu Lin (2):
  lib/tst_kvercmp: Add support to get distname for different OS in
    tst_kvcmp_distname
  syscalls/utimensat01: add exception for Ubuntu 4.4 kernel

 lib/tst_kvercmp.c                                | 16 ++++++++++++++++
 .../kernel/syscalls/utimensat/utimensat01.c      |  7 ++++++-
 2 files changed, 22 insertions(+), 1 deletion(-)

-- 
2.25.1

[LTP] [PATCHv2 1/2] lib/tst_kvercmp: Add support to get distname for different OS in tst_kvcmp_distname

[Po-Hsu Lin]

The kver on Ubuntu will be something like these:
* 4.4.0-187-generic
* 5.4.0-1021-kvm
* 4.15.0-1093-azure

So it's better to parse OS name from ID= in /etc/os-release, instead
of doing this from checking kver substring like what we did for RHEL
and Oracle Linux here.

From the document [1] this string will alway be in lowercase. Example:
"ID=fedora" or "ID=debian". Thus it needs to be converted to uppercase
to make it consistent with other return values in tst_kvcmp_distname().

Note that if ID was not set, it will default to "ID=linux". Thus we
can expect to get LINUX on some distros.

[1] https://www.freedesktop.org/software/systemd/man/os-release.html

Signed-off-by: Po-Hsu Lin <po-hsu.lin@canonical.com>
---
 lib/tst_kvercmp.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/lib/tst_kvercmp.c b/lib/tst_kvercmp.c
index dc3bb669b..af6c6de69 100644
--- a/lib/tst_kvercmp.c
+++ b/lib/tst_kvercmp.c
@@ -18,6 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include <ctype.h>
 #include <stdlib.h>
 #include <unistd.h>
 #include <string.h>
@@ -25,6 +26,8 @@
 #include <sys/utsname.h>
 #include "test.h"
 
+#define OSRELEASE_PATH "/etc/os-release"
+
 static char *parse_digit(const char *str, int *d)
 {
 	unsigned long v;
@@ -127,6 +130,8 @@ int tst_kvexcmp(const char *tst_exv, const char *cur_ver)
 
 const char *tst_kvcmp_distname(const char *kver)
 {
+	static char distname[64];
+	char *tok;
 	if (strstr(kver, ".el5uek"))
 		return "OL5UEK";
 
@@ -139,6 +144,17 @@ const char *tst_kvcmp_distname(const char *kver)
 	if (strstr(kver, ".el6"))
 		return "RHEL6";
 
+	// Special case for other releases with the presencse of /etc/os-release
+	if (access(OSRELEASE_PATH, F_OK) != -1) {
+		SAFE_FILE_LINES_SCANF(NULL, OSRELEASE_PATH, "ID=%s", distname);
+		tok = strtok(distname,"\0");
+		while (*tok) {
+			*tok = toupper((unsigned char) *tok);
+			tok++;
+		}
+		return distname;
+	}
+
 	return NULL;
 }
 
-- 
2.25.1

[LTP] [PATCHv2 1/2] lib/tst_kvercmp: Add support to get distname for different OS in tst_kvcmp_distname

[Cyril Hrubis]

Hi!
>  static char *parse_digit(const char *str, int *d)
>  {
>  	unsigned long v;
> @@ -127,6 +130,8 @@ int tst_kvexcmp(const char *tst_exv, const char *cur_ver)
>  
>  const char *tst_kvcmp_distname(const char *kver)
>  {
> +	static char distname[64];
> +	char *tok;
>  	if (strstr(kver, ".el5uek"))
>  		return "OL5UEK";
>  
> @@ -139,6 +144,17 @@ const char *tst_kvcmp_distname(const char *kver)
>  	if (strstr(kver, ".el6"))
>  		return "RHEL6";
>  
> +	// Special case for other releases with the presencse of /etc/os-release
> +	if (access(OSRELEASE_PATH, F_OK) != -1) {
> +		SAFE_FILE_LINES_SCANF(NULL, OSRELEASE_PATH, "ID=%s", distname);
> +		tok = strtok(distname,"\0");

Isn't this strtok() useless?


Other than that the patchset looks fine.

> +		while (*tok) {
> +			*tok = toupper((unsigned char) *tok);
> +			tok++;
> +		}
> +		return distname;
> +	}
> +
>  	return NULL;
>  }
>  
> -- 
> 2.25.1
> 

-- 
Cyril Hrubis
chrubis@suse.cz

[LTP] [PATCHv2 1/2] lib/tst_kvercmp: Add support to get distname for different OS in tst_kvcmp_distname

[Po-Hsu Lin]

On Tue, Aug 18, 2020 at 11:29 PM Cyril Hrubis <chrubis@suse.cz> wrote:
>
> Hi!
> >  static char *parse_digit(const char *str, int *d)
> >  {
> >       unsigned long v;
> > @@ -127,6 +130,8 @@ int tst_kvexcmp(const char *tst_exv, const char *cur_ver)
> >
> >  const char *tst_kvcmp_distname(const char *kver)
> >  {
> > +     static char distname[64];
> > +     char *tok;
> >       if (strstr(kver, ".el5uek"))
> >               return "OL5UEK";
> >
> > @@ -139,6 +144,17 @@ const char *tst_kvcmp_distname(const char *kver)
> >       if (strstr(kver, ".el6"))
> >               return "RHEL6";
> >
> > +     // Special case for other releases with the presencse of /etc/os-release
> > +     if (access(OSRELEASE_PATH, F_OK) != -1) {
> > +             SAFE_FILE_LINES_SCANF(NULL, OSRELEASE_PATH, "ID=%s", distname);
> > +             tok = strtok(distname,"\0");
>
> Isn't this strtok() useless?
Hello,
oh indeed, the delimiter in strtok will become null character anyway.
I think just tok = distname can do the trick here.
I will send v3 tomorrow.

Thank you


>
>
> Other than that the patchset looks fine.
>
> > +             while (*tok) {
> > +                     *tok = toupper((unsigned char) *tok);
> > +                     tok++;
> > +             }
> > +             return distname;
> > +     }
> > +
> >       return NULL;
> >  }
> >
> > --
> > 2.25.1
> >
>
> --
> Cyril Hrubis
> chrubis@suse.cz

[LTP] [PATCHv2 1/2] lib/tst_kvercmp: Add support to get distname for different OS in tst_kvcmp_distname

[Cyril Hrubis]

Hi!
> oh indeed, the delimiter in strtok will become null character anyway.
> I think just tok = distname can do the trick here.
> I will send v3 tomorrow.

I can easily fix that before I push the patch, no need for v3.

-- 
Cyril Hrubis
chrubis@suse.cz

[LTP] [PATCHv2 1/2] lib/tst_kvercmp: Add support to get distname for different OS in tst_kvcmp_distname

[Po-Hsu Lin]

On Wed, Aug 19, 2020 at 2:29 AM Cyril Hrubis <chrubis@suse.cz> wrote:
>
> Hi!
> > oh indeed, the delimiter in strtok will become null character anyway.
> > I think just tok = distname can do the trick here.
> > I will send v3 tomorrow.
>
> I can easily fix that before I push the patch, no need for v3.

Hello Cyril,
that would be great, thanks for the review!

>
> --
> Cyril Hrubis
> chrubis@suse.cz

[LTP] [PATCHv2 1/2] lib/tst_kvercmp: Add support to get distname for different OS in tst_kvcmp_distname

[Cyril Hrubis]

Hi!
Both patches pushed with a minor changes, thanks.

-- 
Cyril Hrubis
chrubis@suse.cz

[LTP] [PATCHv2 2/2] syscalls/utimensat01: add exception for Ubuntu 4.4 kernel

[Po-Hsu Lin]

Ubuntu 4.4 kernel got this patch (b3b4283) since 4.4.0-48.69,
therefore it will return EPERM instead of EACCES.

Without this fix, the test will fail on Ubuntu 4.4 kernel with:
  FAIL: 18: utimensat() failed with incorrect error, expected EACCES: EPERM (1)
  FAIL: 19: utimensat() failed with incorrect error, expected EACCES: EPERM (1)

Signed-off-by: Po-Hsu Lin <po-hsu.lin@canonical.com>
---
 testcases/kernel/syscalls/utimensat/utimensat01.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/testcases/kernel/syscalls/utimensat/utimensat01.c b/testcases/kernel/syscalls/utimensat/utimensat01.c
index 7dabfed6d..de4f0601b 100644
--- a/testcases/kernel/syscalls/utimensat/utimensat01.c
+++ b/testcases/kernel/syscalls/utimensat/utimensat01.c
@@ -158,6 +158,11 @@ static void tst_multi_set_time(enum tst_ts_type type, struct mytime *mytime)
 
 static void update_error(struct test_case *tc)
 {
+	// Special case for Ubuntu kernel, which has patch b3b4283 since 4.4.0-48.69
+	static struct tst_kern_exv kvers[] = {
+		{ "UBUNTU", "4.4.0-48.69" },
+	};
+
 	if (tc->exp_err != -1)
 		return;
 
@@ -167,7 +172,7 @@ static void update_error(struct test_case *tc)
 	 * This patch has also been merged to stable 4.4 with
 	 * b3b4283 ("vfs: move permission checking into notify_change() for utimes(NULL)")
 	 */
-	if (tst_kvercmp(4, 4, 27) < 0)
+	if (tst_kvercmp2(4, 4, 27, kvers) < 0)
 		tc->exp_err = EACCES;
 	else
 		tc->exp_err = EPERM;
-- 
2.25.1