[PATCH 0/5] coresight: miscellaneous fix and cleanup

[PATCH 0/5] coresight: miscellaneous fix and cleanup

[Mathieu Poirier]

These patches present a few fix and some cleanup code for the coresight
subsystem. The set serves as a foundation for the upcoming 8th iteration
of the "coresight: integration with perf" patch set but not part of the
new functionality, hence choosing to send separately.

This set is based on [1].

Thanks,
Mathieu

Mathieu Poirier (5):
  coresight: fixing indentation problem
  coresight: fixing lockdep error
  coresight: coresight_unregister() function cleanup
  coresight: release reference taken by 'bus_find_device()'
  coresight: remove csdev's link from topology

 drivers/hwtracing/coresight/coresight.c | 57 +++++++++++++++++++++++++++++----
 1 file changed, 51 insertions(+), 6 deletions(-)

-- 
2.1.4

[PATCH 1/5] coresight: fixing indentation problem

[Mathieu Poirier]

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
---
 drivers/hwtracing/coresight/coresight.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/hwtracing/coresight/coresight.c b/drivers/hwtracing/coresight/coresight.c
index 93738dfbf631..4236a02cdab2 100644
--- a/drivers/hwtracing/coresight/coresight.c
+++ b/drivers/hwtracing/coresight/coresight.c
@@ -536,7 +536,7 @@ static void coresight_fixup_orphan_conns(struct coresight_device *csdev)
 	 * are hooked-up with each newly added component.
 	 */
 	bus_for_each_dev(&coresight_bustype, NULL,
-				 csdev, coresight_orphan_match);
+			 csdev, coresight_orphan_match);
 }
 
 
-- 
2.1.4

[PATCH 2/5] coresight: fixing lockdep error

[Mathieu Poirier]

On some platform the following lockdep error occurs when doing simple
manipulations:

    [   23.197021]
    [   23.198608] ======================================================
    [   23.205078] [ INFO: possible circular locking dependency detected ]
    [   23.211639] 4.4.0-rc8-00025-gbbf360b #172 Not tainted
    [   23.216918] -------------------------------------------------------
    [   23.223480] sh/858 is trying to acquire lock:
    [   23.228057]  (coresight_mutex){+.+.+.}, at: [<c0415d40>] coresight_enable+0x1c/0x1b4
    [   23.236206]
    [   23.236206] but task is already holding lock:
    [   23.242309]  (s_active#52){++++.+}, at: [<c01d4b40>] kernfs_fop_write+0x5c/0x1c0
    [   23.250122]
    [   23.250122] which lock already depends on the new lock.
    [   23.250122]
    [   23.258697]
    [   23.258697] the existing dependency chain (in reverse order) is:
    [   23.266510]
    -> #1 (s_active#52){++++.+}:
    [   23.270843]        [<c01d30ec>] __kernfs_remove+0x294/0x35c
    [   23.276672]        [<c01d3e44>] kernfs_remove_by_name_ns+0x44/0x8c
    [   23.283172]        [<c01d6318>] remove_files+0x3c/0x84
    [   23.288543]        [<c01d66b4>] sysfs_remove_group+0x48/0x9c
    [   23.294494]        [<c01d6734>] sysfs_remove_groups+0x2c/0x3c
    [   23.300506]        [<c030b658>] device_remove_attrs+0x5c/0x74
    [   23.306549]        [<c030c290>] device_del+0x110/0x218
    [   23.311950]        [<c030c3c4>] device_unregister+0x2c/0x6c
    [   23.317779]        [<c04156d8>] coresight_unregister+0x30/0x40
    [   23.323883]        [<c041a290>] etm_probe+0x228/0x2e8
    [   23.329193]        [<c02bc760>] amba_probe+0xe4/0x160
    [   23.334503]        [<c0310540>] driver_probe_device+0x23c/0x480
    [   23.340728]        [<c0310820>] __driver_attach+0x9c/0xa0
    [   23.346374]        [<c030e400>] bus_for_each_dev+0x70/0xa4
    [   23.352142]        [<c030fcf4>] driver_attach+0x24/0x28
    [   23.357604]        [<c030f86c>] bus_add_driver+0x1e0/0x278
    [   23.363372]        [<c0310d48>] driver_register+0x80/0x100
    [   23.369110]        [<c02bc508>] amba_driver_register+0x58/0x5c
    [   23.375244]        [<c0749514>] etm_driver_init+0x18/0x1c
    [   23.380889]        [<c0009918>] do_one_initcall+0xc4/0x20c
    [   23.386657]        [<c0715e7c>] kernel_init_freeable+0x160/0x208
    [   23.392974]        [<c052d7fc>] kernel_init+0x18/0xf0
    [   23.398254]        [<c0010850>] ret_from_fork+0x14/0x24
    [   23.403747]
    -> #0 (coresight_mutex){+.+.+.}:
    [   23.408447]        [<c008ed60>] lock_acquire+0xe4/0x210
    [   23.413909]        [<c0530a30>] mutex_lock_nested+0x74/0x450
    [   23.419860]        [<c0415d40>] coresight_enable+0x1c/0x1b4
    [   23.425689]        [<c0416030>] enable_source_store+0x58/0x68
    [   23.431732]        [<c030b358>] dev_attr_store+0x20/0x2c
    [   23.437286]        [<c01d55e8>] sysfs_kf_write+0x50/0x54
    [   23.442871]        [<c01d4ba8>] kernfs_fop_write+0xc4/0x1c0
    [   23.448699]        [<c015b60c>] __vfs_write+0x34/0xe4
    [   23.454040]        [<c015bf38>] vfs_write+0x98/0x174
    [   23.459228]        [<c015c7a8>] SyS_write+0x4c/0xa8
    [   23.464355]        [<c00107c0>] ret_fast_syscall+0x0/0x1c
    [   23.470031]
    [   23.470031] other info that might help us debug this:
    [   23.470031]
    [   23.478393]  Possible unsafe locking scenario:
    [   23.478393]
    [   23.484619]        CPU0                    CPU1
    [   23.489349]        ----                    ----
    [   23.494079]   lock(s_active#52);
    [   23.497497]                                lock(coresight_mutex);
    [   23.503906]                                lock(s_active#52);
    [   23.509918]   lock(coresight_mutex);
    [   23.513702]
    [   23.513702]  *** DEADLOCK ***
    [   23.513702]
    [   23.519897] 3 locks held by sh/858:
    [   23.523529]  #0:  (sb_writers#7){.+.+.+}, at: [<c015ec38>] __sb_start_write+0xa8/0xd4
    [   23.531799]  #1:  (&of->mutex){+.+...}, at: [<c01d4b38>] kernfs_fop_write+0x54/0x1c0
    [   23.539916]  #2:  (s_active#52){++++.+}, at: [<c01d4b40>] kernfs_fop_write+0x5c/0x1c0
    [   23.548156]
    [   23.548156] stack backtrace:
    [   23.552734] CPU: 0 PID: 858 Comm: sh Not tainted 4.4.0-rc8-00025-gbbf360b #172
    [   23.560302] Hardware name: Generic OMAP4 (Flattened Device Tree)
    [   23.566589] Backtrace:
    [   23.569152] [<c00154d4>] (dump_backtrace) from [<c00156d0>] (show_stack+0x18/0x1c)
    [   23.577087]  r7:ed4b8570 r6:c0936400 r5:c07ae71c r4:00000000
    [   23.583038] [<c00156b8>] (show_stack) from [<c027e69c>] (dump_stack+0x98/0xc0)
    [   23.590606] [<c027e604>] (dump_stack) from [<c008a750>] (print_circular_bug+0x21c/0x33c)
    [   23.599090]  r5:c0939d60 r4:c0936400
    [   23.602874] [<c008a534>] (print_circular_bug) from [<c008e370>] (__lock_acquire+0x1c98/0x1d88)
    [   23.611877]  r10:00000003 r9:c0fd7a5c r8:ed4b8550 r7:ed4b8570 r6:ed4b8000 r5:c0ff69e4
    [   23.620117]  r4:c0936400 r3:ed4b8550
    [   23.623901] [<c008c6d8>] (__lock_acquire) from [<c008ed60>] (lock_acquire+0xe4/0x210)
    [   23.632080]  r10:00000000 r9:00000000 r8:60000013 r7:c07cb7b4 r6:00000001 r5:00000000
    [   23.640350]  r4:00000000
    [   23.643005] [<c008ec7c>] (lock_acquire) from [<c0530a30>] (mutex_lock_nested+0x74/0x450)
    [   23.651458]  r10:ecc0bf80 r9:edbe7dcc r8:ed4b8000 r7:c0fd7a5c r6:c0415d40 r5:00000000
    [   23.659729]  r4:c07cb780
    [   23.662384] [<c05309bc>] (mutex_lock_nested) from [<c0415d40>] (coresight_enable+0x1c/0x1b4)
    [   23.671234]  r10:ecc0bf80 r9:edbe7dcc r8:ed733c00 r7:00000000 r6:ed733c00 r5:00000002
    [   23.679473]  r4:ed762140
    [   23.682128] [<c0415d24>] (coresight_enable) from [<c0416030>] (enable_source_store+0x58/0x68)
    [   23.691070]  r7:00000000 r6:ed733c00 r5:00000002 r4:ed762160
    [   23.697052] [<c0415fd8>] (enable_source_store) from [<c030b358>] (dev_attr_store+0x20/0x2c)
    [   23.705780]  r5:edbe7dc0 r4:c0415fd8
    [   23.709533] [<c030b338>] (dev_attr_store) from [<c01d55e8>] (sysfs_kf_write+0x50/0x54)
    [   23.717834]  r5:edbe7dc0 r4:c030b338
    [   23.721618] [<c01d5598>] (sysfs_kf_write) from [<c01d4ba8>] (kernfs_fop_write+0xc4/0x1c0)
    [   23.730163]  r7:00000000 r6:00000000 r5:00000002 r4:edbe7dc0
    [   23.736145] [<c01d4ae4>] (kernfs_fop_write) from [<c015b60c>] (__vfs_write+0x34/0xe4)
    [   23.744323]  r10:00000000 r9:ecc0a000 r8:c0010964 r7:ecc0bf80 r6:00000002 r5:c01d4ae4
    [   23.752593]  r4:ee385a40
    [   23.755249] [<c015b5d8>] (__vfs_write) from [<c015bf38>] (vfs_write+0x98/0x174)
    [   23.762908]  r9:ecc0a000 r8:c0010964 r7:ecc0bf80 r6:000ab0d8 r5:00000002 r4:ee385a40
    [   23.771057] [<c015bea0>] (vfs_write) from [<c015c7a8>] (SyS_write+0x4c/0xa8)
    [   23.778442]  r8:c0010964 r7:00000002 r6:000ab0d8 r5:ee385a40 r4:ee385a40
    [   23.785522] [<c015c75c>] (SyS_write) from [<c00107c0>] (ret_fast_syscall+0x0/0x1c)
    [   23.793457]  r7:00000004 r6:00000001 r5:000ab0d8 r4:00000002
    [   23.799652] coresight-etb10 54162000.etb: ETB enabled
    [   23.805084] coresight-funnel 54164000.funnel: FUNNEL inport 0 enabled
    [   23.811859] coresight-replicator 44000000.ocp:replicator: REPLICATOR enabled
    [   23.819335] coresight-funnel 54158000.funnel: FUNNEL inport 0 enabled
    [   23.826110] coresight-etm3x 5414c000.ptm: ETM tracing enabled

The locking in coresight_unregister() is not required as the only customers of
the function are drivers themselves when an initialisation failure has been
encoutered.

Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
---
 drivers/hwtracing/coresight/coresight.c | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/drivers/hwtracing/coresight/coresight.c b/drivers/hwtracing/coresight/coresight.c
index 4236a02cdab2..41b42018b660 100644
--- a/drivers/hwtracing/coresight/coresight.c
+++ b/drivers/hwtracing/coresight/coresight.c
@@ -713,12 +713,8 @@ EXPORT_SYMBOL_GPL(coresight_register);
 
 void coresight_unregister(struct coresight_device *csdev)
 {
-	mutex_lock(&coresight_mutex);
-
 	kfree(csdev->conns);
 	device_unregister(&csdev->dev);
-
-	mutex_unlock(&coresight_mutex);
 }
 EXPORT_SYMBOL_GPL(coresight_unregister);
 
-- 
2.1.4

[PATCH 3/5] coresight: coresight_unregister() function cleanup

[Mathieu Poirier]

In its current form the code never frees csdev->refcnt allocated
in coresight_register().  There is also a problem with csdev->conns
that is freed before device_unregister() rather than in the device
release function.

This patch addresses both issues by moving kfree(csdev->conns) to
coresight_device_release() and freeing csdev->refcnt, also in
the same function.

Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
---
 drivers/hwtracing/coresight/coresight.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/hwtracing/coresight/coresight.c b/drivers/hwtracing/coresight/coresight.c
index 41b42018b660..8872db4410eb 100644
--- a/drivers/hwtracing/coresight/coresight.c
+++ b/drivers/hwtracing/coresight/coresight.c
@@ -481,6 +481,8 @@ static void coresight_device_release(struct device *dev)
 {
 	struct coresight_device *csdev = to_coresight_device(dev);
 
+	kfree(csdev->conns);
+	kfree(csdev->refcnt);
 	kfree(csdev);
 }
 
@@ -713,7 +715,6 @@ EXPORT_SYMBOL_GPL(coresight_register);
 
 void coresight_unregister(struct coresight_device *csdev)
 {
-	kfree(csdev->conns);
 	device_unregister(&csdev->dev);
 }
 EXPORT_SYMBOL_GPL(coresight_unregister);
-- 
2.1.4

[PATCH 4/5] coresight: release reference taken by 'bus_find_device()'

[Mathieu Poirier]

The reference count taken by function bus_find_device() needs
to be released if a child device is found, something this patch
is adding.

Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
---
 drivers/hwtracing/coresight/coresight.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight.c b/drivers/hwtracing/coresight/coresight.c
index 8872db4410eb..a35ca54a76c9 100644
--- a/drivers/hwtracing/coresight/coresight.c
+++ b/drivers/hwtracing/coresight/coresight.c
@@ -570,6 +570,8 @@ static void coresight_fixup_device_conns(struct coresight_device *csdev)
 
 		if (dev) {
 			conn->child_dev = to_coresight_device(dev);
+			/* and put reference from 'bus_find_device()' */
+			put_device(dev);
 		} else {
 			csdev->orphan = true;
 			conn->child_dev = NULL;
-- 
2.1.4

[PATCH 5/5] coresight: remove csdev's link from topology

[Mathieu Poirier]

In function 'coresight_unregister()', all references to the csdev that
is being taken away need to be removed from the topology.  Otherwise
building the next coresight path from source to sink may use memory
that has been released.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
---
 drivers/hwtracing/coresight/coresight.c | 46 +++++++++++++++++++++++++++++++++
 1 file changed, 46 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight.c b/drivers/hwtracing/coresight/coresight.c
index a35ca54a76c9..7e6e9ff27dd1 100644
--- a/drivers/hwtracing/coresight/coresight.c
+++ b/drivers/hwtracing/coresight/coresight.c
@@ -579,6 +579,50 @@ static void coresight_fixup_device_conns(struct coresight_device *csdev)
 	}
 }
 
+static int coresight_remove_match(struct device *dev, void *data)
+{
+	int i;
+	struct coresight_device *csdev, *iterator;
+	struct coresight_connection *conn;
+
+	csdev = data;
+	iterator = to_coresight_device(dev);
+
+	/* No need to check oneself */
+	if (csdev == iterator)
+		return 0;
+
+	/*
+	 * Circle throuch all the connection of that component.  If we find
+	 * a connection whose name matches @csdev, remove it.
+	 */
+	for (i = 0; i < iterator->nr_outport; i++) {
+		conn = &iterator->conns[i];
+
+		if (conn->child_dev == NULL)
+			continue;
+
+		if (!strcmp(dev_name(&csdev->dev), conn->child_name)) {
+			iterator->orphan = true;
+			conn->child_dev = NULL;
+			/* No need to continue */
+			break;
+		}
+	}
+
+	/*
+	 * Returning '0' ensures that all known component on the
+	 * bus will be checked.
+	 */
+	return 0;
+}
+
+static void coresight_remove_conns(struct coresight_device *csdev)
+{
+	bus_for_each_dev(&coresight_bustype, NULL,
+			 csdev, coresight_remove_match);
+}
+
 /**
  * coresight_timeout - loop until a bit has changed to a specific state.
  * @addr: base address of the area of interest.
@@ -717,6 +761,8 @@ EXPORT_SYMBOL_GPL(coresight_register);
 
 void coresight_unregister(struct coresight_device *csdev)
 {
+	/* Remove references of that device in the topology */
+	coresight_remove_conns(csdev);
 	device_unregister(&csdev->dev);
 }
 EXPORT_SYMBOL_GPL(coresight_unregister);
-- 
2.1.4