Re: linuxnext-20191127 edac warns (was Re: edac KASAN warning in experimental arm64 allmodconfig boot)

[John Garry]

Hi Robert,

> thank you for testing.

I'm just stumbling across these, TBH.

> On 27.11.19 17:07:33, John Garry wrote:
>
> > [snip]
> >
> > I have test enabled:
> > +CONFIG_DEBUG_TEST_DRIVER_REMOVE=y
> > +CONFIG_KASAN=y
> > +CONFIG_DEBUG_KMEMLEAK=y
>
> Is this a regression (did it work before?), or a new test that you
> newly run?

linuxnext-20191119 does not look to have the issue - that's when I 
cherry-pick your refcount fix - but has lots of memory leaks:

root@(none)$
root@(none)$ echo scan > /sys/kernel/debug/kmemleak
root@(none)$ [  121.639978] kmemleak: 128 new suspected memory leaks 
(see /sys/kernel/debug/kmemleak)

root@(none)$ cat /sys/kernel/debug/kmemleak
unreferenced object 0xffff00236c24ba00 (size 256):
  comm "swapper/0", pid 1, jiffies 4294897826 (age 107.824s)
  hex dump (first 32 bytes):
    00 40 2d 3c 23 00 ff ff 00 48 2d 3c 23 00 ff ff  .@-<#....H-<#...
    00 50 2d 3c 23 00 ff ff 00 58 2d 3c 23 00 ff ff  .P-<#....X-<#...
  backtrace:
    [<0000000009aed8e3>] __kmalloc+0x1e0/0x2c0
    [<00000000bf599427>] edac_mc_alloc+0x31c/0x888
    [<00000000c070e314>] ghes_edac_register+0x15c/0x390
    [<00000000e4aad1c2>] ghes_probe+0x28c/0x5f0
    [<0000000079c357cb>] platform_drv_probe+0x70/0xd8
    [<00000000d4ab9188>] really_probe+0x118/0x548
    [<00000000763d50f1>] driver_probe_device+0x7c/0x148
    [<0000000058e623c3>] device_driver_attach+0x94/0xa0
    [<00000000d7cb679d>] __driver_attach+0xa4/0x110
    [<000000007d0942a0>] bus_for_each_dev+0xe8/0x158
    [<000000004cf734d1>] driver_attach+0x30/0x40
    [<000000009aa3536e>] bus_add_driver+0x234/0x2f0
    [<00000000d163cfe0>] driver_register+0xbc/0x1d0
    [<000000007e4f0ac1>] __platform_driver_register+0x7c/0x88
    [<00000000a63c8dd0>] ghes_init+0xbc/0x14c
    [<00000000356c8a7f>] do_one_initcall+0xb4/0x254
unreferenced object 0xffff00233c2d4000 (size 1024):
  comm "swapper/0", pid 1, jiffies 4294897826 (age 107.824s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<000000004945469f>] kmem_cache_alloc+0x188/0x260
    [<0000000032ea779d>] edac_mc_alloc+0x38c/0x888

Unfortunately v5.4 has similar memory leaks.

Thanks,
John