Re: BUG: bisected: thermald regression (MEMLEAK) in commit c7ff29763989bd09c433f73fae3c1e1c15d9cda4

From: Mirsad Goran Todorovac
Date: Mon Oct 24 2022 - 16:31:27 EST

Hi Srinivas,

On 24. 10. 2022. 17:51, srinivas pandruvada wrote:
Hi Mirsad,

Thanks for the bisect.

On Mon, 2022-10-24 at 15:13 +0200, Mirsad Goran Todorovac wrote:
Dear all,

Around Sep 27th 2022 I've noticed in a mainline kernel built with
CONFIG_DEBUG_KMEMLEAK=y
that there actually is a leak:

sudo cat /sys/kernel/debug/kmemleak unreferenced object
0xffff8881095f3ee0 (size 80): comm "thermald", pid 837, jiffies
4294896698 (age 9867.428s) hex dump (first 32 bytes): 00 00 00 00 00
00
00 00 0d 01 2d 00 00 00 00 00 ..........-..... af 07 01 00 00 c9 ff
ff
00 00 00 00 00 00 00 00 ................ backtrace:
[<00000000b50b9dd6>]
kmem_cache_alloc+0x184/0x380 [<00000000fa8428c0>]
acpi_os_acquire_object+0x2c/0x32 [<000000002cc0099f>]
acpi_ps_alloc_op+0x65/0xe6 [<00000000335faf1b>]
acpi_ps_get_next_arg+0x842/0x9ed [<000000007afa2dee>]
acpi_ps_parse_loop+0x718/0xee1 [<0000000010ce490e>]
acpi_ps_parse_aml+0x261/0x7b2 [<00000000278d4c5f>]
acpi_ps_execute_method+0x360/0x459 [<00000000ff7ad4ba>]
acpi_ns_evaluate+0x595/0x810 [<0000000037ce3488>]
acpi_evaluate_object+0x28b/0x5b2 [<000000001a800bbf>]
acpi_run_osc+0x209/0x3d0 [<00000000776fbd43>]
int3400_thermal_run_osc+0xed/0x180 [int3400_thermal]
[<00000000d6ec2302>] current_uuid_store+0x17c/0x1d0 [int3400_thermal]
[<00000000486cf3e6>] dev_attr_store+0x3e/0x60 [<00000000bf193027>]
sysfs_kf_write+0x88/0xa0 [<00000000820b5cce>]
kernfs_fop_write_iter+0x1c9/0x270 [<0000000062f8d35e>]
vfs_write+0x5a5/0x750 Mr. Pandruvada required a bug bisect from me,
so I
have eventually made one. # first bad commit:
[c7ff29763989bd09c433f73fae3c1e1c15d9cda4] thermal: int340x: Update
OS
This will say this patch as this patch is calling acpi_run_osc in
response to thermald calls for the first time.

But looking at code, this is freeing the memory allocated by
acpi_run_osc() call chain as any other caller.

status = acpi_run_osc(handle, &context);
if (ACPI_SUCCESS(status)) {
ret = *((u32 *)(context.ret.pointer + 4));
if (ret != *enable)
result = -EPERM;

kfree(context.ret.pointer);
} else
result = -EPERM;

There is no kfree when call failed as at other places.
I think you are failing, you can search for "_OSC" in dmesg.
On some Dell systems this OSC setting fails because of some BIOS issue.
May be you are hitting that case.
Just for the sake of test, please apply the diff and see if the issue
is gone.

Thank you for the patch. Unfortunately, when applied to v6.0.3 it didn't fix the issue.

marvin@marvin-IdeaPad-3-15ITL6:~$ uname -rms
Linux 6.0.3-18-fix01-mlk+ x86_64
marvin@marvin-IdeaPad-3-15ITL6:~$ sudo bash
[sudo] password for marvin:
root@marvin-IdeaPad-3-15ITL6:/home/marvin# cat /sys/kernel/debug/kmemleak
root@marvin-IdeaPad-3-15ITL6:/home/marvin# echo scan > /sys/kernel/debug/kmemleak
root@marvin-IdeaPad-3-15ITL6:/home/marvin# cat /sys/kernel/debug/kmemleak
unreferenced object 0xffff998b030c3370 (size 80):
  comm "thermald", pid 824, jiffies 4294893654 (age 67.080s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 0d 01 2d 00 00 00 00 00 ..........-.....
    af 07 01 c0 6f bc ff ff 00 00 00 00 00 00 00 00 ....o...........
  backtrace:
    [<00000000490225c2>] slab_post_alloc_hook+0x80/0x2e0
    [<00000000dc142b33>] kmem_cache_alloc+0x166/0x2e0
    [<00000000168f1071>] acpi_os_acquire_object+0x2c/0x32
    [<00000000fcc615e1>] acpi_ps_alloc_op+0x4a/0x99
    [<00000000fb475bb4>] acpi_ps_get_next_arg+0x611/0x761
    [<000000009048d529>] acpi_ps_parse_loop+0x494/0x8d7
    [<000000005b0bf086>] acpi_ps_parse_aml+0x1bb/0x561
    [<000000007ab7e288>] acpi_ps_execute_method+0x20f/0x2d5
    [<00000000c12fa6b7>] acpi_ns_evaluate+0x34d/0x4f3
    [<000000001be94719>] acpi_evaluate_object+0x180/0x3ae
    [<00000000423a7ad5>] acpi_run_osc+0x128/0x250
    [<0000000040a72af8>] int3400_thermal_run_osc+0x6f/0xc0 [int3400_thermal]
    [<00000000f8d59987>] current_uuid_store+0xe3/0x120 [int3400_thermal]
    [<000000007e2e2d17>] dev_attr_store+0x14/0x30
    [<00000000b824b589>] sysfs_kf_write+0x38/0x50
    [<00000000beae69c1>] kernfs_fop_write_iter+0x146/0x1d0
root@marvin-IdeaPad-3-15ITL6:/home/marvin#

The build process was as follows:

 1573  10/24/2022 06:41:53 PM  cd linux_stable
 1574  10/24/2022 06:42:03 PM  git checkout v6.0.3
 1575  10/24/2022 06:42:44 PM  cd ..
 1576  10/24/2022 06:42:50 PM  time rm -rf linux_stable_build; time cp -rp linux_stable linux_stable_build; \
                    time diff -ur linux_stable linux_stable_build; cd linux_stable_build
 1577  10/24/2022 06:46:19 PM  git apply ../thermald-20221024-01.diff
 1578  10/24/2022 06:46:28 PM  vi ../config-5.15.0-50-memleak
 1579  10/24/2022 06:47:08 PM  cp ../config-5.15.0-50-memleak .config
 1580  10/24/2022 06:47:16 PM  make olddefconfig
 1581  10/24/2022 06:48:42 PM  time nice make CC="ccache gcc" KBUILD_BUILD_TIMESTAMP="" -j10 deb-pkg; date

I think your patch definitively makes sense, but there's more to this than meets the eye :-/

Hope this helps.

Thanks
Mirsad

--
Mirsad Goran Todorovac
Sistem inženjer
Grafički fakultet | Akademija likovnih umjetnosti
Sveučilište u Zagrebu
--
System engineer
Faculty of Graphic Arts | Academy of Fine Arts
University of Zagreb, Republic of Croatia
The European Union