kernel/input: fix failure on an old kernel

PATCH New Xiao Yang 2 participants 8 messages Oct 9, 2017 – Dec 4, 2017

Share & tools

Xiao Yang

author

2017-10-09 09:57 UTC #0 lore raw

PATCH 1/1 kernel/input: fix failure on an old kernel

On some old kernels(e.g. v2.6.32), we always triggered the following error when running input04 and input05:
--------------------------------------------------------------------

TINFO  :  Found uinput dev at /dev/uinput
TINFO  :  Unexpected ev type=0 code=0 value=0
TFAIL  :  Data received /dev/inputX

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

After investigation, the single EV_SYN event can be processed on an old kernel. However, the EV_SYN event handling has been modfied by the following patch:
'4369c64c79a2("Input: Send events one packet at a time")'

The patch is designed to reduce the time taken for processing a lot of events. As a side effect, the single EV_SYN event handling depends on one or more other events(e.g. EV_REL), so it should be processed together with others, since this action is atomic.

    if (disposition & INPUT_FLUSH) {
            // num_vals: number of values queued in the current frame
            if (dev->num_vals >= 2)
                    input_pass_values(dev, dev->vals, dev->num_vals);
                    dev->num_vals = 0;

...
--------------------------------------------------------------------

For example, steps to send an empty move and EV_SYN:
-----------------------------------------------------
send_event(fd, EV_REL, REL_X, 0); send_event(fd, EV_SYN, 0, 0);
-----------------------------------------------------

On an old kernel(before v3.7.0), the EV_SYN/SYN_REPORT event can be processed independently, so it is passed into /dev/input/eventX.

On a new kernel(since v3.7.0), the EV_SYN/SYN_REPORT event depends on EV_REL/REL_X which is an empty move, so it is ignored.

We feel this failure is due to the different mechanism of processing EV_SYN instead of a kernel bug, so we update no_events_queued() to fix it.

Signed-off-by Xiao Yang <[email protected]>

---
 testcases/kernel/input/input02.c      |  2 +-
 testcases/kernel/input/input04.c      |  2 +-
 testcases/kernel/input/input05.c      |  2 +-
 testcases/kernel/input/input_helper.c | 26 +++++++++++++++++++++-----
 testcases/kernel/input/input_helper.h |  2 +-
 5 files changed, 25 insertions(+), 9 deletions(-)

testcases/kernel/input/input02.c +1 -1

index df3b257..4f77758 100644
@@ -61,7 +61,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 1))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");

testcases/kernel/input/input04.c +1 -1

index 4b16ca7..e20b4f6 100644
@@ -57,7 +57,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 0))
 				tst_resm(TPASS,
 					"No data received in /dev/inputX");
 			else

testcases/kernel/input/input05.c +1 -1

index 5ca01d8..e8251a3 100644
@@ -63,7 +63,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 0))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");

testcases/kernel/input/input_helper.c +21 -5

index 08fa81c..68b8b5f 100644
@@ -229,7 +229,15 @@ void destroy_device(int fd)
 		unload_uinput();
 }
 
-int no_events_queued(int fd)
+/*
+ * the value of no_syn_evs_strict:
+ * 1: check that nothing is received in /dev/input/eventX
+ * 0: check that nothing or a EV_SYN/SYN_REPORT event is
+ *    received in /dev/input/eventX
+ * On an old kernel(before v3.7.0), a EV_SYN/SYN_REPORT event
+ * is always received even though we send empty events
+ */
+int no_events_queued(int fd, int no_syn_evs_strict)
 {
 	struct pollfd fds = {.fd = fd, .events = POLLIN};
 	int ret, res;

@@ -241,13 +249,21 @@ int no_events_queued(int fd)
 		res = read(fd, &ev, sizeof(ev));
 
 		if (res == sizeof(ev)) {
-			tst_resm(TINFO,
-			         "Unexpected ev type=%i code=%i value=%i",
-			         ev.type, ev.code, ev.value);
+			if (no_syn_evs_strict) {
+				tst_resm(TINFO, "Unexpected ev type=%i code=%i "
+					"value=%i", ev.type, ev.code, ev.value);
+				return 0;
+			}
+
+			if (!no_syn_evs_strict && ev.type != EV_SYN) {
+				tst_resm(TINFO, "Unexpected event type %i "
+					"expected %i", ev.type, EV_SYN);
+				return 0;
+			}
 		}
 	}
 
-	return ret == 0;
+	return 1;
 }
 
 static int check_device(void)

testcases/kernel/input/input_helper.h +1 -1

index 9b85dc9..95c20f2 100644
@@ -29,6 +29,6 @@ int open_uinput(void);
 void create_device(int fd);
 void setup_mouse_events(int fd);
 void destroy_device(int fd);
-int no_events_queued(int fd);
+int no_events_queued(int fd, int no_syn_evs_strict);
 
 #endif /* INPUT_HELPER_H */

Signature · 1 line

1.8.3.1

Xiao Yang

author

2017-11-06 07:52 UTC #1 lore raw

PATCH 1/1 kernel/input: fix failure on an old kernel

Hi,

Ping :-)

Thanks,
Xiao Yang

Quoted 155 lines · Xiao Yang

On some old kernels(e.g. v2.6.32), we always triggered the following error when running input04 and input05:
--------------------------------------------------------------------
TINFO  :  Found uinput dev at /dev/uinput
TINFO  :  Unexpected ev type=0 code=0 value=0
TFAIL  :  Data received /dev/inputX
---------------------------------------------------------------------
After investigation, the single EV_SYN event can be processed on an old kernel. However, the EV_SYN event handling has been modfied by the following patch:
'4369c64c79a2("Input: Send events one packet at a time")'
The patch is designed to reduce the time taken for processing a lot of events. As a side effect, the single EV_SYN event handling depends on one or more other events(e.g. EV_REL), so it should be processed together with others, since this action is atomic.
Only when the number of queued events(include EV_SYN) is greater than one can these events be processed.
Please see the kenrel code in drivers/input/input.c, as below:
--------------------------------------------------------------------
static void input_handle_event(struct input_dev *dev,
...
    if (disposition & INPUT_FLUSH) {
            // num_vals: number of values queued in the current frame
            if (dev->num_vals >= 2)
                    input_pass_values(dev, dev->vals, dev->num_vals);
                    dev->num_vals = 0;
...
--------------------------------------------------------------------
For example, steps to send an empty move and EV_SYN:
-----------------------------------------------------
send_event(fd, EV_REL, REL_X, 0); send_event(fd, EV_SYN, 0, 0);
-----------------------------------------------------
On an old kernel(before v3.7.0), the EV_SYN/SYN_REPORT event can be processed independently, so it is passed into /dev/input/eventX.
On a new kernel(since v3.7.0), the EV_SYN/SYN_REPORT event depends on EV_REL/REL_X which is an empty move, so it is ignored.
We feel this failure is due to the different mechanism of processing EV_SYN instead of a kernel bug, so we update no_events_queued() to fix it.
Signed-off-by Xiao Yang <[email protected]>
---
 testcases/kernel/input/input02.c      |  2 +-
 testcases/kernel/input/input04.c      |  2 +-
 testcases/kernel/input/input05.c      |  2 +-
 testcases/kernel/input/input_helper.c | 26 +++++++++++++++++++++-----
 testcases/kernel/input/input_helper.h |  2 +-
 5 files changed, 25 insertions(+), 9 deletions(-)
testcases/kernel/input/input02.c +1 -1
index df3b257..4f77758 100644
@@ -61,7 +61,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 1))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");
testcases/kernel/input/input04.c +1 -1
index 4b16ca7..e20b4f6 100644
@@ -57,7 +57,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 0))
 				tst_resm(TPASS,
 					"No data received in /dev/inputX");
 			else
testcases/kernel/input/input05.c +1 -1
index 5ca01d8..e8251a3 100644
@@ -63,7 +63,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 0))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");
testcases/kernel/input/input_helper.c +21 -5
index 08fa81c..68b8b5f 100644
@@ -229,7 +229,15 @@ void destroy_device(int fd)
 		unload_uinput();
 }
 
-int no_events_queued(int fd)
+/*
+ * the value of no_syn_evs_strict:
+ * 1: check that nothing is received in /dev/input/eventX
+ * 0: check that nothing or a EV_SYN/SYN_REPORT event is
+ *    received in /dev/input/eventX
+ * On an old kernel(before v3.7.0), a EV_SYN/SYN_REPORT event
+ * is always received even though we send empty events
+ */
+int no_events_queued(int fd, int no_syn_evs_strict)
 {
 	struct pollfd fds = {.fd = fd, .events = POLLIN};
 	int ret, res;
@@ -241,13 +249,21 @@ int no_events_queued(int fd)
 		res = read(fd, &ev, sizeof(ev));
 
 		if (res == sizeof(ev)) {
-			tst_resm(TINFO,
-			         "Unexpected ev type=%i code=%i value=%i",
-			         ev.type, ev.code, ev.value);
+			if (no_syn_evs_strict) {
+				tst_resm(TINFO, "Unexpected ev type=%i code=%i "
+					"value=%i", ev.type, ev.code, ev.value);
+				return 0;
+			}
+
+			if (!no_syn_evs_strict && ev.type != EV_SYN) {
+				tst_resm(TINFO, "Unexpected event type %i "
+					"expected %i", ev.type, EV_SYN);
+				return 0;
+			}
 		}
 	}
 
-	return ret == 0;
+	return 1;
 }
 
 static int check_device(void)
testcases/kernel/input/input_helper.h +1 -1
index 9b85dc9..95c20f2 100644
@@ -29,6 +29,6 @@ int open_uinput(void);
 void create_device(int fd);
 void setup_mouse_events(int fd);
 void destroy_device(int fd);
-int no_events_queued(int fd);
+int no_events_queued(int fd, int no_syn_evs_strict);
 
 #endif /* INPUT_HELPER_H */

Cyril Hrubis

2017-11-07 14:20 UTC #2 lore raw

PATCH 1/1 kernel/input: fix failure on an old kernel

In reply to Xiao Yang

Hi!

Quoted 41 lines

On some old kernels(e.g. v2.6.32), we always triggered the following error when running input04 and input05:
--------------------------------------------------------------------
TINFO  :  Found uinput dev at /dev/uinput
TINFO  :  Unexpected ev type=0 code=0 value=0
TFAIL  :  Data received /dev/inputX
---------------------------------------------------------------------
After investigation, the single EV_SYN event can be processed on an old kernel. However, the EV_SYN event handling has been modfied by the following patch:
'4369c64c79a2("Input: Send events one packet at a time")'
The patch is designed to reduce the time taken for processing a lot of events. As a side effect, the single EV_SYN event handling depends on one or more other events(e.g. EV_REL), so it should be processed together with others, since this action is atomic.
Only when the number of queued events(include EV_SYN) is greater than one can these events be processed.
Please see the kenrel code in drivers/input/input.c, as below:
--------------------------------------------------------------------
static void input_handle_event(struct input_dev *dev,
...
    if (disposition & INPUT_FLUSH) {
            // num_vals: number of values queued in the current frame
            if (dev->num_vals >= 2)
                    input_pass_values(dev, dev->vals, dev->num_vals);
                    dev->num_vals = 0;
...
--------------------------------------------------------------------
For example, steps to send an empty move and EV_SYN:
-----------------------------------------------------
send_event(fd, EV_REL, REL_X, 0); send_event(fd, EV_SYN, 0, 0);
-----------------------------------------------------
On an old kernel(before v3.7.0), the EV_SYN/SYN_REPORT event can be processed independently, so it is passed into /dev/input/eventX.

Quoted 6 lines

On a new kernel(since v3.7.0), the EV_SYN/SYN_REPORT event depends on EV_REL/REL_X which is an empty move, so it is ignored.
We feel this failure is due to the different mechanism of processing EV_SYN instead of a kernel bug, so we update no_events_queued() to fix it.

Well, I guess that we can change the tests to ignore stray sync events on older kernels, but new ones empty sync events would be regression.

Hence we should enable the backward compatibility only on old kernels where the tests actually fails.

Signature · 2 lines

Cyril Hrubis
[email protected]

Xiao Yang

author

2017-11-08 10:37 UTC #3 lore raw

PATCH v2 1/1 kernel/input: fix failure on an old kernel

On some old kernels(e.g. v2.6.32), we always triggered the following error when running input04 and input05:
--------------------------------------------------------------------

TINFO  :  Found uinput dev at /dev/uinput
TINFO  :  Unexpected ev type=0 code=0 value=0
TFAIL  :  Data received /dev/inputX

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

    if (disposition & INPUT_FLUSH) {
            // num_vals: number of values queued in the current frame
            if (dev->num_vals >= 2)
                    input_pass_values(dev, dev->vals, dev->num_vals);
                    dev->num_vals = 0;

...
--------------------------------------------------------------------

On an old kernel(before v3.7.0), the EV_SYN/SYN_REPORT event can be processed independently, so it is passed into /dev/input/eventX.

On a new kernel(since v3.7.0), the EV_SYN/SYN_REPORT event depends on EV_REL/REL_X which is an empty move, so it is ignored.

We feel this failure is due to the different mechanism of processing EV_SYN instead of a kernel bug, so we ignore stray sync events on older kernels.

Signed-off-by Xiao Yang <[email protected]>

---
 testcases/kernel/input/input02.c      |  2 +-
 testcases/kernel/input/input04.c      |  5 +++-
 testcases/kernel/input/input05.c      |  6 +++-
 testcases/kernel/input/input_helper.c | 53 ++++++++++++++++++++++++++++++++---
 testcases/kernel/input/input_helper.h |  3 +-
 5 files changed, 61 insertions(+), 8 deletions(-)

testcases/kernel/input/input02.c +1 -1

index df3b257..6964ed7 100644
@@ -61,7 +61,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 0))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");

testcases/kernel/input/input04.c +4 -1

index 4b16ca7..94d39c1 100644
@@ -35,6 +35,7 @@ static void send_events(void);
 static void cleanup(void);
 
 static int fd, fd2;
+static int sync_events_ignored;
 
 char *TCID = "input04";

@@ -57,7 +58,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, sync_events_ignored))
 				tst_resm(TPASS,
 					"No data received in /dev/inputX");
 			else

@@ -82,6 +83,8 @@ static void setup(void)
 	create_device(fd);
 
 	fd2 = open_device();
+
+	sync_events_ignored = check_single_sync_event(fd, fd2);
 }
 
 static void send_events(void)

testcases/kernel/input/input05.c +5 -1

index 5ca01d8..af6d862 100644
@@ -42,6 +42,8 @@ static void cleanup(void);
 static int fd;
 static int fd2;
 
+static int sync_events_ignored;
+
 char *TCID = "input05";
 
 int main(int ac, char **av)

@@ -63,7 +65,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, sync_events_ignored))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");

@@ -89,6 +91,8 @@ static void setup(void)
 	create_device(fd);
 
 	fd2 = open_device();
+
+	sync_events_ignored = check_single_sync_event(fd, fd2);
 }
 
 static void send_events(void)

testcases/kernel/input/input_helper.c +49 -4

72 lines not on this page · raw message

testcases/kernel/input/input_helper.h +2 -1

index 9b85dc9..1ef4d32 100644
@@ -29,6 +29,7 @@ int open_uinput(void);
 void create_device(int fd);
 void setup_mouse_events(int fd);
 void destroy_device(int fd);
-int no_events_queued(int fd);
+int no_events_queued(int fd, int sync_events_ignored);
+int check_single_sync_event(int fd1, int fd2);
 
 #endif /* INPUT_HELPER_H */

Signature · 1 line

1.8.3.1

Cyril Hrubis

2017-11-09 13:45 UTC #4 lore raw

PATCH v2 1/1 kernel/input: fix failure on an old kernel

Hi!

Quoted 12 lines

testcases/kernel/input/input_helper.c +8 -0

+/*
+ * the value of sync_events_ignored:
+ * 0: cannot ignore EV_SYN/SYN_REPORT events received in /dev/input/eventX
+ * 1: ignore EV_SYN/SYN_REPORT events received in /dev/input/eventX
+ * On an old kernel(before v3.7.0), EV_SYN/SYN_REPORT events are always
+ * received even though we send empty moves.
+ */
+int no_events_queued(int fd, int sync_events_ignored)
 {
 	struct pollfd fds = {.fd = fd, .events = POLLIN};
 	int ret, res;

testcases/kernel/input/input_helper.c int no_events_queued(int fd) · line 252

Quoted 8 lines

testcases/kernel/input/input_helper.c +2 -3

@@ -241,9 +248,14 @@ int no_events_queued(int fd)
 		res = read(fd, &ev, sizeof(ev));
 
 		if (res == sizeof(ev)) {
-			tst_resm(TINFO,
-			         "Unexpected ev type=%i code=%i value=%i",
-			         ev.type, ev.code, ev.value);
+			if (sync_events_ignored && !ev.type &&
+			    !ev.code && !ev.value) {

The ev.value is undefined for sync events, it's 0 most of the time but I remember that we had to fix the input06.c because it was failing for somebody, see check_sync_event() there. Maybe we should just move the check_event* and check_sync_event() helpers to the library so that we can use them here as well.

Quoted 10 lines

testcases/kernel/input/input_helper.c +6 -0

+				ret = 0;
+			} else {
+				tst_resm(TINFO,
+					 "Unexpected ev type=%i code=%i value=%i",
+					 ev.type, ev.code, ev.value);
+			}
 		}
 	}

testcases/kernel/input/input_helper.c static int check_device(void) · line 279

Quoted 4 lines

testcases/kernel/input/input_helper.c +1 -0

@@ -264,7 +276,40 @@ static int check_device(void)
 			return 1;
 	}
 
+

^
Please do not add useless empty lines.

Quoted 26 lines

testcases/kernel/input/input_helper.c +22 -0

 	fclose(file);
 
 	return 0;
 }
+
+/*
+ * Use this function to check if the current kernel can get single
+ * EV_SYN/SYN_REPORT events.
+ */
+int check_single_sync_event(int fd1, int fd2)
+{
+	pid_t pid;
+	struct pollfd fds = {.fd = fd2, .events = POLLIN};
+	struct input_event ev;
+	int ret, ignored;
+
+	pid = tst_fork();
+
+	if (pid == -1)
+		tst_brkm(TBROK | TERRNO, NULL, "fork() failed");
+
+	if (!pid) {
+		send_event(fd1, EV_SYN, 0, 0);
+		exit(0);
+	}
+

testcases/kernel/input/input_helper.c static int check_device(void) · line 315

Quoted 10 lines

testcases/kernel/input/input_helper.c +10 -0

+	ret = poll(&fds, 1, 30);
+	if (ret > 0 && SAFE_READ(NULL, 1, fd2, &ev, sizeof(ev)) == sizeof(ev)) {
+		if (!ev.type && !ev.code && !ev.value)
+			ignored = 1;
+	}
+
+	SAFE_WAITPID(NULL, pid, NULL, 0);
+
+	return ignored;
+}

This still does not fix the problem. Even if new kernel started to produce stray sync events these testcases will still pass. What I think is a good solution is simply to define first fixed kernel version and expect that no stray sync events are produced on newer kernels.

Quoted 18 lines

testcases/kernel/input/input_helper.h +2 -1

index 9b85dc9..1ef4d32 100644
@@ -29,6 +29,7 @@ int open_uinput(void);
 void create_device(int fd);
 void setup_mouse_events(int fd);
 void destroy_device(int fd);
-int no_events_queued(int fd);
+int no_events_queued(int fd, int sync_events_ignored);
+int check_single_sync_event(int fd1, int fd2);
 
 #endif /* INPUT_HELPER_H */

Signature · 1 line

1.8.3.1

Signature · 2 lines

Cyril Hrubis
[email protected]

Xiao Yang

author

2017-11-13 06:45 UTC #5 lore raw

PATCH v3 1/1 kernel/input: fix failure on an old kernel

1) On some old kernels(e.g. v2.6.32), we always triggered the following error when running input04 and input05:
--------------------------------------------------------------------

   TINFO  :  Found uinput dev at /dev/uinput
   TINFO  :  Unexpected ev type=0 code=0 value=0
   TFAIL  :  Data received /dev/inputX

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

Only when the number of queued events(include EV_SYN) is greater than one can these events be processed.
Please see the kenrel code in drivers/input/input.c, as below:
--------------------------------------------------------------------
static void input_handle_event(struct input_dev *dev,
...
if (disposition & INPUT_FLUSH) {
// num_vals: number of values queued in the current frame if (dev->num_vals >= 2)
input_pass_values(dev, dev->vals, dev->num_vals); dev->num_vals = 0;
...
--------------------------------------------------------------------

   For example, steps to send an empty move and EV_SYN:
   -----------------------------------------------------
   send_event(fd, EV_REL, REL_X, 0);
   send_event(fd, EV_SYN, 0, 0);
   -----------------------------------------------------

On an old kernel(before v3.7.0), the EV_SYN/SYN_REPORT event can be processed independently, so it is passed into /dev/input/eventX.

On a new kernel(since v3.7.0), the EV_SYN/SYN_REPORT event depends on EV_REL/REL_X which is an empty move, so it is ignored.

We feel this failure is due to the different mechanism of processing EV_SYN instead of a kernel bug, so we ignore stray sync events on older kernels.

2) We move the check_event_code and check_sync_event() helpers to the library so that we can use them.

3) Fix compiler warning, as below:
------------------------------------------------------------------------
input06.c:213: warning: ‘ret’ may be used uninitialized in this function
------------------------------------------------------------------------

Signed-off-by Xiao Yang <[email protected]>

---
 testcases/kernel/input/input02.c      |  2 +-
 testcases/kernel/input/input04.c      |  2 +-
 testcases/kernel/input/input05.c      |  2 +-
 testcases/kernel/input/input06.c      | 18 ++++--------------
 testcases/kernel/input/input_helper.c | 34 +++++++++++++++++++++++++++++-----
 testcases/kernel/input/input_helper.h |  4 +++-
 6 files changed, 39 insertions(+), 23 deletions(-)

testcases/kernel/input/input02.c +1 -1

index df3b257..6964ed7 100644
@@ -61,7 +61,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 0))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");

testcases/kernel/input/input04.c +1 -1

index 4b16ca7..e57b76b 100644
@@ -57,7 +57,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 1))
 				tst_resm(TPASS,
 					"No data received in /dev/inputX");
 			else

testcases/kernel/input/input05.c +1 -1

index 5ca01d8..46b4fe8 100644
@@ -63,7 +63,7 @@ int main(int ac, char **av)
 		case -1:
 			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
 		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 1))
 				tst_resm(TPASS, "No data received in eventX");
 			else
 				tst_resm(TFAIL, "Data received in eventX");

testcases/kernel/input/input06.c +4 ~~-14~~

index 4c5f1a8..14141b7 100644
@@ -115,11 +115,6 @@ static int check_event(struct input_event *iev, int event, int code, int value)
 	return iev->type == event && iev->code == code && iev->value == value;
 }
 
-static int check_event_code(struct input_event *iev, int event, int code)
-{
-	return iev->type == event && iev->code == code;
-}
-
 static void read_events(void)
 {
 	int rd = read(fd2, events, sizeof(events));

@@ -151,11 +146,6 @@ static struct input_event *next_event(void)
 	return &events[ev_iter++];
 }
 
-static int check_sync_event(void)
-{
-	return check_event_code(next_event(), EV_SYN, SYN_REPORT);
-}
-
 static int parse_autorepeat_config(struct input_event *iev)
 {
 	if (!check_event_code(iev, EV_REP, REP_DELAY)) {

@@ -177,13 +167,13 @@ static int parse_key(struct input_event *iev)
 {
 	int autorep_count = 0;
 
-	if (!check_event(iev, EV_KEY, KEY_X, 1) || !check_sync_event()) {
+	if (!check_event(iev, EV_KEY, KEY_X, 1) || !check_sync_event(next_event())) {
 		tst_resm(TFAIL, "Didn't get expected key press for KEY_X");
 		return 0;
 	}
 
 	iev = next_event();
-	while (check_event(iev, EV_KEY, KEY_X, 2) && check_sync_event()) {
+	while (check_event(iev, EV_KEY, KEY_X, 2) && check_sync_event(next_event())) {
 		autorep_count++;
 		iev = next_event();
 	}

@@ -195,7 +185,7 @@ static int parse_key(struct input_event *iev)
 		return 0;
 	}
 
-	if (!check_event(iev, EV_KEY, KEY_X, 0) || !check_sync_event()) {
+	if (!check_event(iev, EV_KEY, KEY_X, 0) || !check_sync_event(next_event())) {
 		tst_resm(TFAIL,
 			 "Didn't get expected key release for KEY_X");
 		return 0;

@@ -210,7 +200,7 @@ static int parse_key(struct input_event *iev)
 static int check_events(void)
 {
 	struct input_event *iev;
-	int ret;
+	int ret = 0;
 	int rep_config_done = 0;
 	int rep_keys_done = 0;

testcases/kernel/input/input_helper.c +29 -5

index 08fa81c..13a6cd8 100644
@@ -229,21 +229,45 @@ void destroy_device(int fd)
 		unload_uinput();
 }
 
-int no_events_queued(int fd)
+int check_event_code(struct input_event *iev, int event, int code)
+{
+	return iev->type == event && iev->code == code;
+}
+
+int check_sync_event(struct input_event *iev)
+{
+	return check_event_code(iev, EV_SYN, SYN_REPORT);
+}
+
+/*
+ * the value of stray_sync_event:
+ * 0: EV_SYN/SYN_REPORT events should not be received in /dev/input/eventX
+ * 1: EV_SYN/SYN_REPORT events may be received in /dev/input/eventX
+ * On an old kernel(before v3.7.0), EV_SYN/SYN_REPORT events are always
+ * received even though we send empty moves.
+ */
+int no_events_queued(int fd, int stray_sync_event)
 {
 	struct pollfd fds = {.fd = fd, .events = POLLIN};
-	int ret, res;
+	int ret, res, sync_event_ignored;
 	struct input_event ev;
 
+	if (tst_kvercmp(3, 7, 0) < 0 && stray_sync_event)
+		sync_event_ignored = 1;
+
 	ret = poll(&fds, 1, 30);
 
 	if (ret > 0) {
 		res = read(fd, &ev, sizeof(ev));
 
 		if (res == sizeof(ev)) {
-			tst_resm(TINFO,
-			         "Unexpected ev type=%i code=%i value=%i",
-			         ev.type, ev.code, ev.value);
+			if (sync_event_ignored && check_sync_event(&ev)) {
+				ret = 0;
+			} else {
+				tst_resm(TINFO,
+					 "Unexpected ev type=%i code=%i value=%i",
+					 ev.type, ev.code, ev.value);
+			}
 		}
 	}

testcases/kernel/input/input_helper.h +3 -1

index 9b85dc9..7f61be1 100644
@@ -29,6 +29,8 @@ int open_uinput(void);
 void create_device(int fd);
 void setup_mouse_events(int fd);
 void destroy_device(int fd);
-int no_events_queued(int fd);
+int check_event_code(struct input_event *iev, int event, int code);
+int check_sync_event(struct input_event *iev);
+int no_events_queued(int fd, int stray_sync_event);
 
 #endif /* INPUT_HELPER_H */

Signature · 1 line

1.8.3.1

Xiao Yang

author

2017-12-04 07:45 UTC #6 lore raw

PATCH v3 1/1 kernel/input: fix failure on an old kernel

Hi Cyril,

Ping. :-)

Thanks,
Xiao Yang

Quoted 236 lines · Xiao Yang

1) On some old kernels(e.g. v2.6.32), we always triggered the following error when running input04 and input05:
--------------------------------------------------------------------
TINFO : Found uinput dev at /dev/uinput
TINFO : Unexpected ev type=0 code=0 value=0
TFAIL : Data received /dev/inputX
---------------------------------------------------------------------

Only when the number of queued events(include EV_SYN) is greater than one can these events be processed.
Please see the kenrel code in drivers/input/input.c, as below:
--------------------------------------------------------------------
static void input_handle_event(struct input_dev *dev,
...
if (disposition& INPUT_FLUSH) {
// num_vals: number of values queued in the current frame if (dev->num_vals>= 2)
input_pass_values(dev, dev->vals, dev->num_vals); dev->num_vals = 0;
...
--------------------------------------------------------------------

On an old kernel(before v3.7.0), the EV_SYN/SYN_REPORT event can be processed independently, so it is passed into /dev/input/eventX.

On a new kernel(since v3.7.0), the EV_SYN/SYN_REPORT event depends on EV_REL/REL_X which is an empty move, so it is ignored.

We feel this failure is due to the different mechanism of processing EV_SYN instead of a kernel bug, so we ignore stray sync events on older kernels.

2) We move the check_event_code and check_sync_event() helpers to the library so that we can use them.

3) Fix compiler warning, as below:

    ------------------------------------------------------------------------
    input06.c:213: warning: ‘ret’ may be used uninitialized in this function
    ------------------------------------------------------------------------

Signed-off-by Xiao Yang<[email protected]>

---
  testcases/kernel/input/input02.c      |  2 +-
  testcases/kernel/input/input04.c      |  2 +-
  testcases/kernel/input/input05.c      |  2 +-
  testcases/kernel/input/input06.c      | 18 ++++--------------
  testcases/kernel/input/input_helper.c | 34 +++++++++++++++++++++++++++++-----
  testcases/kernel/input/input_helper.h |  4 +++-
  6 files changed, 39 insertions(+), 23 deletions(-)

testcases/kernel/input/input02.c +1 -1

index df3b257..6964ed7 100644
@@ -61,7 +61,7 @@ int main(int ac, char **av)
  		case -1:
  			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
  		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 0))
  				tst_resm(TPASS, "No data received in eventX");
  			else
  				tst_resm(TFAIL, "Data received in eventX");

testcases/kernel/input/input04.c +1 -1

index 4b16ca7..e57b76b 100644
@@ -57,7 +57,7 @@ int main(int ac, char **av)
  		case -1:
  			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
  		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 1))
  				tst_resm(TPASS,
  					"No data received in /dev/inputX");
  			else

testcases/kernel/input/input05.c +1 -1

index 5ca01d8..46b4fe8 100644
@@ -63,7 +63,7 @@ int main(int ac, char **av)
  		case -1:
  			tst_brkm(TBROK | TERRNO, cleanup, "fork() failed");
  		default:
-			if (no_events_queued(fd2))
+			if (no_events_queued(fd2, 1))
  				tst_resm(TPASS, "No data received in eventX");
  			else
  				tst_resm(TFAIL, "Data received in eventX");

testcases/kernel/input/input06.c +4 ~~-14~~

index 4c5f1a8..14141b7 100644
@@ -115,11 +115,6 @@ static int check_event(struct input_event *iev, int event, int code, int value)
  	return iev->type == event&&  iev->code == code&&  iev->value == value;
  }
 
-static int check_event_code(struct input_event *iev, int event, int code)
-{
-	return iev->type == event&&  iev->code == code;
-}
-
  static void read_events(void)
  {
  	int rd = read(fd2, events, sizeof(events));

@@ -151,11 +146,6 @@ static struct input_event *next_event(void)
  	return&events[ev_iter++];
  }
 
-static int check_sync_event(void)
-{
-	return check_event_code(next_event(), EV_SYN, SYN_REPORT);
-}
-
  static int parse_autorepeat_config(struct input_event *iev)
  {
  	if (!check_event_code(iev, EV_REP, REP_DELAY)) {

@@ -177,13 +167,13 @@ static int parse_key(struct input_event *iev)
  {
  	int autorep_count = 0;
 
-	if (!check_event(iev, EV_KEY, KEY_X, 1) || !check_sync_event()) {
+	if (!check_event(iev, EV_KEY, KEY_X, 1) || !check_sync_event(next_event())) {
  		tst_resm(TFAIL, "Didn't get expected key press for KEY_X");
  		return 0;
  	}
 
  	iev = next_event();
-	while (check_event(iev, EV_KEY, KEY_X, 2)&&  check_sync_event()) {
+	while (check_event(iev, EV_KEY, KEY_X, 2)&&  check_sync_event(next_event())) {
  		autorep_count++;
  		iev = next_event();
  	}

@@ -195,7 +185,7 @@ static int parse_key(struct input_event *iev)
  		return 0;
  	}
 
-	if (!check_event(iev, EV_KEY, KEY_X, 0) || !check_sync_event()) {
+	if (!check_event(iev, EV_KEY, KEY_X, 0) || !check_sync_event(next_event())) {
  		tst_resm(TFAIL,
  			 "Didn't get expected key release for KEY_X");
  		return 0;

@@ -210,7 +200,7 @@ static int parse_key(struct input_event *iev)
  static int check_events(void)
  {
  	struct input_event *iev;
-	int ret;
+	int ret = 0;
  	int rep_config_done = 0;
  	int rep_keys_done = 0;

testcases/kernel/input/input_helper.c +29 -5

index 08fa81c..13a6cd8 100644
@@ -229,21 +229,45 @@ void destroy_device(int fd)
  		unload_uinput();
  }
 
-int no_events_queued(int fd)
+int check_event_code(struct input_event *iev, int event, int code)
+{
+	return iev->type == event&&  iev->code == code;
+}
+
+int check_sync_event(struct input_event *iev)
+{
+	return check_event_code(iev, EV_SYN, SYN_REPORT);
+}
+
+/*
+ * the value of stray_sync_event:
+ * 0: EV_SYN/SYN_REPORT events should not be received in /dev/input/eventX
+ * 1: EV_SYN/SYN_REPORT events may be received in /dev/input/eventX
+ * On an old kernel(before v3.7.0), EV_SYN/SYN_REPORT events are always
+ * received even though we send empty moves.
+ */
+int no_events_queued(int fd, int stray_sync_event)
  {
  	struct pollfd fds = {.fd = fd, .events = POLLIN};
-	int ret, res;
+	int ret, res, sync_event_ignored;
  	struct input_event ev;
 
+	if (tst_kvercmp(3, 7, 0)<  0&&  stray_sync_event)
+		sync_event_ignored = 1;
+
  	ret = poll(&fds, 1, 30);
 
  	if (ret>  0) {
  		res = read(fd,&ev, sizeof(ev));
 
  		if (res == sizeof(ev)) {
-			tst_resm(TINFO,
-			         "Unexpected ev type=%i code=%i value=%i",
-			         ev.type, ev.code, ev.value);
+			if (sync_event_ignored&&  check_sync_event(&ev)) {
+				ret = 0;
+			} else {
+				tst_resm(TINFO,
+					 "Unexpected ev type=%i code=%i value=%i",
+					 ev.type, ev.code, ev.value);
+			}
  		}
  	}

testcases/kernel/input/input_helper.h +3 -1

index 9b85dc9..7f61be1 100644
@@ -29,6 +29,8 @@ int open_uinput(void);
  void create_device(int fd);
  void setup_mouse_events(int fd);
  void destroy_device(int fd);
-int no_events_queued(int fd);
+int check_event_code(struct input_event *iev, int event, int code);
+int check_sync_event(struct input_event *iev);
+int no_events_queued(int fd, int stray_sync_event);
 
  #endif /* INPUT_HELPER_H */

Cyril Hrubis

2017-12-04 15:33 UTC #7 lore raw

PATCH v3 1/1 kernel/input: fix failure on an old kernel

In reply to Xiao Yang

Hi!

testcases/kernel/input/input_helper.c void destroy_device(int fd) · line 265

Quoted 2 lines

testcases/kernel/input/input_helper.c +2 -0
+			if (sync_event_ignored && check_sync_event(&ev)) {
+				ret = 0;

I've added a TINFO message here that we ignored the empty sync event here and pushed, thanks.

Quoted 7 lines

testcases/kernel/input/input_helper.c +5 -0

+			} else {
+				tst_resm(TINFO,
+					 "Unexpected ev type=%i code=%i value=%i",
+					 ev.type, ev.code, ev.value);
+			}
 		}
 	}

Signature · 2 lines

Cyril Hrubis
[email protected]