Hi Eugeniu,

On Tue, Dec 3, 2019 at 11:51 AM Eugeniu Rosca [off-list ref] wrote:

On Wed, Nov 27, 2019 at 09:42:51AM +0100, Geert Uytterhoeven wrote:

quoted

+static int gpio_aggregator_probe(struct platform_device *pdev)
+{
+     struct device *dev = &pdev->dev;
+     struct gpio_desc **descs;
+     struct gpiochip_fwd *fwd;
+     int i, n;

FWIW/FTR, doing some blind creation and deletion of gpio aggregator
chips [1] on R-Car H3ULCB overnight, kmemleak reported once [2]. Not
sure this is something 100% reproducible.

[1] while true; do \
   echo e6055400.gpio 12,13 > /sys/bus/platform/drivers/gpio-aggregator/new_device; \
   echo gpio-aggregator.0 > /sys/bus/platform/drivers/gpio-aggregator/delete_device; \
   done

[2] unreferenced object 0xffff0006d2c2e000 (size 128):
  comm "kworker/3:1", pid 55, jiffies 4294676978 (age 38546.676s)
  hex dump (first 32 bytes):
    00 d9 d2 d3 06 00 ff ff 0c 00 e0 0f ff ff ff ff  ................
    01 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<00000000a8e18c13>] slab_post_alloc_hook+0x8c/0x94
    [<000000006f419a4f>] __kmalloc+0x170/0x218
    [<0000000060d185ea>] kobj_map+0x78/0x1c0
    [<00000000c96645f3>] cdev_add+0x68/0x94
    [<00000000a7a5a8ac>] cdev_device_add+0x74/0x90
    [<00000000497871d3>] gpiochip_setup_dev+0x84/0x1f0
    [<00000000b993f95f>] gpiochip_add_data_with_key+0xbcc/0x11f0
    [<00000000fd728c0e>] devm_gpiochip_add_data+0x60/0xa8
    [<00000000442e34c1>] gpio_aggregator_probe+0x210/0x3c8
    [<00000000076e13fb>] platform_drv_probe+0x70/0xe4
    [<00000000de84b58b>] really_probe+0x2d8/0x434
    [<00000000c95c9784>] driver_probe_device+0x15c/0x16c
    [<00000000afb7dd4f>] __device_attach_driver+0xdc/0x120
    [<00000000efa40cae>] bus_for_each_drv+0x12c/0x154
    [<00000000c149acef>] __device_attach+0x148/0x1e0
    [<00000000a74fd158>] device_initial_probe+0x24/0x30

This is the allocation of the GPIO character device, which is allocated
in response to the creation of the GPIO chip, from .probe().
As that is done using devm_gpiochip_add_data(), the chardev should be
deallocated automatically by devm_gpio_chip_release() when
platform_device_unregister() is called.

Weird...

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds