* [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32.
@ 2018-10-01 17:11 kmagic
2018-10-08 9:47 ` Petr Vorel
2018-10-12 16:16 ` Petr Vorel
0 siblings, 2 replies; 5+ messages in thread
From: kmagic @ 2018-10-01 17:11 UTC (permalink / raw)
To: ltp
From: Kenneth Magic <kmagic@google.com>
Android runs all 32bit programs in a PER_LIINUX32 personality
even under a 64 bit kernel, causing the machine to show up as
"i686". Check for and correct this case so tests use the correct
kernel_bits value.
Signed-off-by: Kenneth Magic <kmagic@google.com>
---
lib/tst_kernel.c | 47 +++++++++++++++++++++++++++++++++++++++++------
1 file changed, 41 insertions(+), 6 deletions(-)
diff --git a/lib/tst_kernel.c b/lib/tst_kernel.c
index 4206e3c0d..d051f3b07 100644
--- a/lib/tst_kernel.c
+++ b/lib/tst_kernel.c
@@ -15,21 +15,28 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
+#include <linux/personality.h>
#include <sys/utsname.h>
#include "test.h"
#include "tst_kernel.h"
-int tst_kernel_bits(void)
+static int get_kernel_bits_from_uname(struct utsname *buf)
{
- struct utsname buf;
- int kernel_bits;
-
- if (uname(&buf)) {
+ if (uname(buf)) {
tst_brkm(TBROK | TERRNO, NULL, "uname()");
return -1;
}
- kernel_bits = strstr(buf.machine, "64") ? 64 : 32;
+ return strstr(buf->machine, "64") ? 64 : 32;
+}
+
+int tst_kernel_bits(void)
+{
+ struct utsname buf;
+ int kernel_bits = get_kernel_bits_from_uname(&buf);
+
+ if (kernel_bits == -1)
+ return -1;
/*
* ARM64 (aarch64) defines 32-bit compatibility modes as
@@ -40,6 +47,34 @@ int tst_kernel_bits(void)
|| !strcmp(buf.machine, "s390x"))
kernel_bits = 64;
+#ifdef __ANDROID__
+ /* Android's bionic libc sets the PER_LINUX32 personality for all 32-bit
+ * programs. This will cause buf.machine to report as i686 even though
+ * the kernel itself is 64-bit.
+ */
+ if (!strcmp(buf.machine, "i686") &&
+ (personality(0xffffffff) & PER_MASK) == PER_LINUX32) {
+ /* Set the personality back to the default. */
+ if (personality(PER_LINUX) == -1) {
+ tst_brkm(TBROK | TERRNO, NULL, "personality()");
+ return -1;
+ }
+
+ /* Redo the uname check without the PER_LINUX32 personality to
+ * determine the actual kernel bits value.
+ */
+ kernel_bits = get_kernel_bits_from_uname(&buf);
+ if (kernel_bits == -1)
+ return -1;
+
+ /* Set the personality back to PER_LINUX32. */
+ if (personality(PER_LINUX32) == -1) {
+ tst_brkm(TBROK | TERRNO, NULL, "personality()");
+ return -1;
+ }
+ }
+#endif /* __ANDROID__ */
+
tst_resm(TINFO, "uname.machine=%s kernel is %ibit",
buf.machine, kernel_bits);
--
2.19.0.605.g01d371f741-goog
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32.
2018-10-01 17:11 [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32 kmagic
@ 2018-10-08 9:47 ` Petr Vorel
2018-10-12 16:16 ` Petr Vorel
1 sibling, 0 replies; 5+ messages in thread
From: Petr Vorel @ 2018-10-08 9:47 UTC (permalink / raw)
To: ltp
Hi Kenneth,
> From: Kenneth Magic <kmagic@google.com>
> Android runs all 32bit programs in a PER_LIINUX32 personality
> even under a 64 bit kernel, causing the machine to show up as
> "i686". Check for and correct this case so tests use the correct
> kernel_bits value.
> Signed-off-by: Kenneth Magic <kmagic@google.com>
Reviewed-by: Petr Vorel <pvorel@suse.cz>
Kind regards,
Petr
^ permalink raw reply [flat|nested] 5+ messages in thread
* [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32.
2018-10-01 17:11 [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32 kmagic
2018-10-08 9:47 ` Petr Vorel
@ 2018-10-12 16:16 ` Petr Vorel
1 sibling, 0 replies; 5+ messages in thread
From: Petr Vorel @ 2018-10-12 16:16 UTC (permalink / raw)
To: ltp
Hi Kenneth,
> From: Kenneth Magic <kmagic@google.com>
> Android runs all 32bit programs in a PER_LIINUX32 personality
> even under a 64 bit kernel, causing the machine to show up as
> "i686". Check for and correct this case so tests use the correct
> kernel_bits value.
> Signed-off-by: Kenneth Magic <kmagic@google.com>
Pushed. Thanks for your patch!
Kind regards,
Petr
^ permalink raw reply [flat|nested] 5+ messages in thread
* [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32.
2018-09-14 18:40 Kenneth Magic
@ 2018-10-01 14:47 ` Cyril Hrubis
0 siblings, 0 replies; 5+ messages in thread
From: Cyril Hrubis @ 2018-10-01 14:47 UTC (permalink / raw)
To: ltp
Hi!
Looks like the patch was mangled by your mail client (all the
whitespaces are gone), can you please try adjust your mail client
configuration and resend?
See:
https://www.kernel.org/doc/html/latest/process/email-clients.html?highlight=email%20clients#some-email-client-mua-hints
--
Cyril Hrubis
chrubis@suse.cz
^ permalink raw reply [flat|nested] 5+ messages in thread
* [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32.
@ 2018-09-14 18:40 Kenneth Magic
2018-10-01 14:47 ` Cyril Hrubis
0 siblings, 1 reply; 5+ messages in thread
From: Kenneth Magic @ 2018-09-14 18:40 UTC (permalink / raw)
To: ltp
Android runs all 32bit programs in a PER_LIINUX32 personality
even under a 64 bit kernel, causing the machine to show up as
"i686". Check for and correct this case so tests use the correct
kernel_bits value.
Signed-off-by: Kenneth Magic <kmagic@google.com>
---
lib/tst_kernel.c | 47 +++++++++++++++++++++++++++++++++++++++++------
1 file changed, 41 insertions(+), 6 deletions(-)
diff --git a/lib/tst_kernel.c b/lib/tst_kernel.c
index 42d64cbdc..8edafb8ff 100644
--- a/lib/tst_kernel.c
+++ b/lib/tst_kernel.c
@@ -15,21 +15,28 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
+#include <linux/personality.h>
#include <sys/utsname.h>
#include "test.h"
#include "tst_kernel.h"
-int tst_kernel_bits(void)
+static int get_kernel_bits_from_uname(struct utsname *buf)
{
- struct utsname buf;
- int kernel_bits;
-
- if (uname(&buf)) {
+ if (uname(buf)) {
tst_brkm(TBROK | TERRNO, NULL, "uname()");
return -1;
}
- kernel_bits = strstr(buf.machine, "64") ? 64 : 32;
+ return strstr(buf->machine, "64") ? 64 : 32;
+}
+
+int tst_kernel_bits(void)
+{
+ struct utsname buf;
+ int kernel_bits = get_kernel_bits_from_uname(&buf);
+
+ if (kernel_bits == -1)
+ return -1;
/*
* ARM64 (aarch64) defines 32-bit compatibility modes as
@@ -40,6 +47,34 @@ int tst_kernel_bits(void)
|| !strcmp(buf.machine, "s390x"))
kernel_bits = 64;
+#ifdef __ANDROID__
+ /* Android's bionic libc sets the PER_LINUX32 personality for all 32-bit
+ * programs. This will cause buf.machine to report as i686 even though
+ * the kernel itself is 64-bit.
+ */
+ if (!strcmp(buf.machine, "i686") &&
+ (personality(0xffffffff) & PER_MASK) == PER_LINUX32) {
+ /* Set the personality back to the default. */
+ if (personality(PER_LINUX) == -1) {
+ tst_brkm(TBROK | TERRNO, NULL, "personality()");
+ return -1;
+ }
+
+ /* Redo the uname check without the PER_LINUX32 personality to
+ * determine the actual kernel bits value.
+ */
+ kernel_bits = get_kernel_bits_from_uname(&buf);
+ if (kernel_bits == -1)
+ return -1;
+
+ /* Set the personality back to PER_LINUX32. */
+ if (personality(PER_LINUX32) == -1) {
+ tst_brkm(TBROK | TERRNO, NULL, "personality()");
+ return -1;
+ }
+ }
+#endif /* __ANDROID__ */
+
tst_resm(TINFO, "uname.machine=%s kernel is %ibit",
buf.machine, kernel_bits);
--
2.19.0.397.gdd90340f6a-goog
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20180914/48e093ea/attachment-0001.html>
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2018-10-12 16:16 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-10-01 17:11 [LTP] [PATCH] Correct the kernel_bits value in x86 PER_LINUX32 kmagic
2018-10-08 9:47 ` Petr Vorel
2018-10-12 16:16 ` Petr Vorel
-- strict thread matches above, loose matches on Subject: below --
2018-09-14 18:40 Kenneth Magic
2018-10-01 14:47 ` Cyril Hrubis
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.