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

[Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr>]

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