I got the following memory leak report while doing fault injection test:
unreferenced object 0xffff888108cff080 (size 16):
comm "77-i2c-ptn5150", pid 253, jiffies 4294732125 (age 33.960s)
hex dump (first 16 bytes):
65 78 74 63 6f 6e 38 32 00 f0 cf 08 81 88 ff ff extcon82........
backtrace:
[<000000001811fc5c>] __kmalloc_node_track_caller+0x44/0x1b0
[<000000005bac260d>] kvasprintf+0xb5/0x140
[<000000000dd1d389>] kvasprintf_const+0x55/0x110
[<000000000ea8369c>] kobject_set_name_vargs+0x43/0xf0
[<00000000d89b28c6>] dev_set_name+0xab/0xe0
[<00000000a55cf990>] extcon_dev_register+0x170/0xea0 [extcon_core]
[<00000000eb2f186e>] devm_extcon_dev_register+0x48/0xb0 [extcon_core]
[<0000000077933346>] ptn5150_i2c_probe+0x2f2/0x543 [extcon_ptn5150]
In the error case before device_register(), the name allocated
by dev_set_name() is leaked. Fix this by moving dev_set_name()
to front of device_register(), so it's managed by device core,
and it is freed when the refcount is 0.
Fixes: 806d9dd71ff5 ("Extcon: support multiple states at a device.")
Signed-off-by: Yang Yingliang <[email protected]>
---
drivers/extcon/extcon.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/extcon/extcon.c b/drivers/extcon/extcon.c
index e1c71359b605..f978e5e8e6cf 100644
--- a/drivers/extcon/extcon.c
+++ b/drivers/extcon/extcon.c
@@ -1115,8 +1115,6 @@ int extcon_dev_register(struct extcon_dev *edev)
"extcon device name is null\n");
return -EINVAL;
}
- dev_set_name(&edev->dev, "extcon%lu",
- (unsigned long)atomic_inc_return(&edev_no));
if (edev->max_supported) {
char *str;
@@ -1252,6 +1250,8 @@ int extcon_dev_register(struct extcon_dev *edev)
dev_set_drvdata(&edev->dev, edev);
edev->state = 0;
+ dev_set_name(&edev->dev, "extcon%lu",
+ (unsigned long)atomic_inc_return(&edev_no));
ret = device_register(&edev->dev);
if (ret) {
put_device(&edev->dev);
--
2.25.1