Skip to contentSkip to conversation

Memory leak from acpi_ev_install_space_handler()

bug report John Garry 3 participants 5 messages Apr 6, 2021 – May 19, 2021

Subsystems (2)
lore
Share & tools
lore

John Garry

author
#0 lore raw

Hi guys,

On next-20210406, I enabled CONFIG_DEBUG_KMEMLEAK and CONFIG_DEBUG_TEST_DRIVER_REMOVE for my arm64 system, and see this:

root@debian:/home/john# more /sys/kernel/debug/kmemleak
unreferenced object 0xffff202803c11f00 (size 128):
comm "swapper/0", pid 1, jiffies 4294894325 (age 337.524s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 1f c1 03 28 20 ff ff............( ..
08 1f c1 03 28 20 ff ff 00 00 00 00 00 00 00 00....( ..........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000ba00abc5>] i2c_acpi_install_space_handler+0xd4/0x138
[<000000008da42058>] i2c_register_adapter+0x368/0x910
[<00000000c03f7142>] i2c_add_adapter+0x9c/0x100
[<0000000000ba2fcf>] i2c_add_numbered_adapter+0x44/0x58
[<000000007df22d67>] i2c_dw_probe_master+0x68c/0x900
[<00000000682dfc98>] dw_i2c_plat_probe+0x460/0x640
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
unreferenced object 0xffff00280452c100 (size 128):
comm "swapper/0", pid 1, jiffies 4294894558 (age 336.604s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 c1 52 04 28 00 ff ff..........R.(...
08 c1 52 04 28 00 ff ff 00 00 00 00 00 00 00 00..R.(...........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000988d4f61>] acpi_gpiochip_add+0x20c/0x4a0
[<0000000073d4faab>] gpiochip_add_data_with_key+0xd10/0x1680
[<000000001d50b98a>] devm_gpiochip_add_data_with_key+0x30/0x78
[<00000000fc3e7eaf>] dwapb_gpio_probe+0x828/0xb28
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
[<00000000d330caed>] bus_for_each_dev+0xec/0x160
[<00000000eebc5f04>] driver_attach+0x34/0x48
root@debian:/home/john#

Thanks,
John

#1 lore raw
Quoted 5 lines · John Garry

Hi guys,

On next-20210406, I enabled CONFIG_DEBUG_KMEMLEAK and CONFIG_DEBUG_TEST_DRIVER_REMOVE for my arm64 system, and see this:

Why exactly do you think that acpi_ev_install_space_handler() leaks memory?

Quoted 49 lines
root@debian:/home/john# more /sys/kernel/debug/kmemleak
unreferenced object 0xffff202803c11f00 (size 128):
comm "swapper/0", pid 1, jiffies 4294894325 (age 337.524s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 1f c1 03 28 20 ff ff............( ..
08 1f c1 03 28 20 ff ff 00 00 00 00 00 00 00 00....( ..........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000ba00abc5>] i2c_acpi_install_space_handler+0xd4/0x138
[<000000008da42058>] i2c_register_adapter+0x368/0x910
[<00000000c03f7142>] i2c_add_adapter+0x9c/0x100
[<0000000000ba2fcf>] i2c_add_numbered_adapter+0x44/0x58
[<000000007df22d67>] i2c_dw_probe_master+0x68c/0x900
[<00000000682dfc98>] dw_i2c_plat_probe+0x460/0x640
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
unreferenced object 0xffff00280452c100 (size 128):
comm "swapper/0", pid 1, jiffies 4294894558 (age 336.604s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 c1 52 04 28 00 ff ff..........R.(...
08 c1 52 04 28 00 ff ff 00 00 00 00 00 00 00 00..R.(...........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000988d4f61>] acpi_gpiochip_add+0x20c/0x4a0
[<0000000073d4faab>] gpiochip_add_data_with_key+0xd10/0x1680
[<000000001d50b98a>] devm_gpiochip_add_data_with_key+0x30/0x78
[<00000000fc3e7eaf>] dwapb_gpio_probe+0x828/0xb28
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
[<00000000d330caed>] bus_for_each_dev+0xec/0x160
[<00000000eebc5f04>] driver_attach+0x34/0x48
root@debian:/home/john#

Thanks,
John

author
#2 lore raw
Quoted 7 lines · Rafael J. Wysocki
Quoted 5 lines · John Garry

Hi guys,

On next-20210406, I enabled CONFIG_DEBUG_KMEMLEAK and CONFIG_DEBUG_TEST_DRIVER_REMOVE for my arm64 system, and see this:

Hi Rafael,

Quoted 2 lines

Why exactly do you think that acpi_ev_install_space_handler() leaks memory?

I don't think that acpi_ev_install_space_handler() itself leaks memory, but it seems that there is something missing in the code which is meant to undo/clean up after that on the uninstall path - I did make the point in writing "memory leak from", but maybe still not worded clearly.

Anyway, I have not analyzed the problem fully - I'm just reporting.

I don't mind looking further if requested.

Thanks,
John

Quoted 51 lines
Quoted 47 lines
root@debian:/home/john# more /sys/kernel/debug/kmemleak
unreferenced object 0xffff202803c11f00 (size 128):
comm "swapper/0", pid 1, jiffies 4294894325 (age 337.524s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 1f c1 03 28 20 ff ff............( ..
08 1f c1 03 28 20 ff ff 00 00 00 00 00 00 00 00....( ..........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000ba00abc5>] i2c_acpi_install_space_handler+0xd4/0x138
[<000000008da42058>] i2c_register_adapter+0x368/0x910
[<00000000c03f7142>] i2c_add_adapter+0x9c/0x100
[<0000000000ba2fcf>] i2c_add_numbered_adapter+0x44/0x58
[<000000007df22d67>] i2c_dw_probe_master+0x68c/0x900
[<00000000682dfc98>] dw_i2c_plat_probe+0x460/0x640
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
unreferenced object 0xffff00280452c100 (size 128):
comm "swapper/0", pid 1, jiffies 4294894558 (age 336.604s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 c1 52 04 28 00 ff ff..........R.(...
08 c1 52 04 28 00 ff ff 00 00 00 00 00 00 00 00..R.(...........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000988d4f61>] acpi_gpiochip_add+0x20c/0x4a0
[<0000000073d4faab>] gpiochip_add_data_with_key+0xd10/0x1680
[<000000001d50b98a>] devm_gpiochip_add_data_with_key+0x30/0x78
[<00000000fc3e7eaf>] dwapb_gpio_probe+0x828/0xb28
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
[<00000000d330caed>] bus_for_each_dev+0xec/0x160
[<00000000eebc5f04>] driver_attach+0x34/0x48
root@debian:/home/john#
Quoted 2 lines

Thanks,
John

.

#3 lore raw
Quoted 28 lines

-----Original Message-----
From: John Garry <[email protected]>
Sent: Tuesday, April 6, 2021 9:48 AM
To: Rafael J. Wysocki <[email protected]>
Cc: ACPI Devel Maling List <[email protected]>; Len Brown <[email protected]>; Rafael J. Wysocki <[email protected]>; linux- [email protected]
Subject: Re: [bug report] Memory leak from acpi_ev_install_space_handler()

Quoted 1 line · Rafael J. Wysocki

On Tue, Apr 6, 2021 at 5:51 PM John Garry <[email protected]>

Quoted 5 lines · wrote:
Quoted 5 lines

Hi guys,

On next-20210406, I enabled CONFIG_DEBUG_KMEMLEAK and CONFIG_DEBUG_TEST_DRIVER_REMOVE for my arm64 system, and see

this:

Quoted 1 line

Hi Rafael,

Quoted 1 line

Why exactly do you think that acpi_ev_install_space_handler() leaks

memory?

Quoted 1 line
Quoted 7 lines

I don't think that acpi_ev_install_space_handler() itself leaks memory, but it seems that there is something missing in the code which is meant to undo/clean up after that on the uninstall path - I did make the point in writing "memory leak from", but maybe still not worded clearly.

Anyway, I have not analyzed the problem fully - I'm just reporting.

Hi John,

Quoted 1 line

I don't mind looking further if requested.

Someone else reported this as well. Could you try the patch below? I think it might help fix this issue..

Thanks,
Erik

drivers/acpi/acpica/utdelete.c +8 -0
index 624a26794d55..e5ba9795ec69 100644
@@ -285,6 +285,14 @@ static void acpi_ut_delete_internal_obj(union acpi_operand_object *object)
                }
                break;
 
+       case ACPI_TYPE_LOCAL_ADDRESS_HANDLER:
+
+               ACPI_DEBUG_PRINT((ACPI_DB_ALLOCATIONS,
+                                 "***** Address handler %p\n", object));
+
+               acpi_os_delete_mutex(object->address_space.context_mutex);
+               break;
+
        default:
 
                break;
Quoted 55 lines

Thanks,
John

Quoted 51 lines
Quoted 47 lines
root@debian:/home/john# more /sys/kernel/debug/kmemleak
unreferenced object 0xffff202803c11f00 (size 128):
comm "swapper/0", pid 1, jiffies 4294894325 (age 337.524s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 1f c1 03 28 20 ff ff............( ..
08 1f c1 03 28 20 ff ff 00 00 00 00 00 00 00 00....( ..........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000ba00abc5>] i2c_acpi_install_space_handler+0xd4/0x138
[<000000008da42058>] i2c_register_adapter+0x368/0x910
[<00000000c03f7142>] i2c_add_adapter+0x9c/0x100
[<0000000000ba2fcf>] i2c_add_numbered_adapter+0x44/0x58
[<000000007df22d67>] i2c_dw_probe_master+0x68c/0x900
[<00000000682dfc98>] dw_i2c_plat_probe+0x460/0x640
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
unreferenced object 0xffff00280452c100 (size 128):
comm "swapper/0", pid 1, jiffies 4294894558 (age 336.604s)
hex dump (first 32 bytes):
00 00 00 00 02 00 00 00 08 c1 52 04 28 00 ff ff..........R.(...
08 c1 52 04 28 00 ff ff 00 00 00 00 00 00 00 00..R.(...........
backtrace:
[<00000000670a0938>] slab_post_alloc_hook+0x9c/0x2f8
[<00000000a3f47b39>] kmem_cache_alloc+0x198/0x2a8
[<000000002bdba864>] acpi_os_create_semaphore+0x54/0xe0
[<00000000bcd513fe>] acpi_ev_install_space_handler+0x24c/0x300
[<0000000002e116e2>] acpi_install_address_space_handler+0x64/0xb0
[<00000000988d4f61>] acpi_gpiochip_add+0x20c/0x4a0
[<0000000073d4faab>] gpiochip_add_data_with_key+0xd10/0x1680
[<000000001d50b98a>] devm_gpiochip_add_data_with_key+0x30/0x78
[<00000000fc3e7eaf>] dwapb_gpio_probe+0x828/0xb28
[<00000000ad2dd3ee>] platform_probe+0x8c/0x108
[<00000000dd183e3f>] really_probe+0x190/0x670
[<0000000066017341>] driver_probe_device+0x8c/0xf8
[<00000000c441e843>] device_driver_attach+0x9c/0xa8
[<00000000f91dc709>] __driver_attach+0x88/0x138
[<00000000d330caed>] bus_for_each_dev+0xec/0x160
[<00000000eebc5f04>] driver_attach+0x34/0x48
root@debian:/home/john#
Quoted 2 lines

Thanks,
John

.

author
#4 lore raw
Quoted 4 lines · Erik

Hi John,

Quoted 1 line

I don't mind looking further if requested.

Someone else reported this as well.

I noticed. By chance, it was my close colleague Xiang Chen.

Could you try the patch below? I think it might help fix this issue..

It looks like the same problem, and you provided the same solution, so we can look to test it.

Thanks

Quoted 24 lines

Thanks,
Erik

drivers/acpi/acpica/utdelete.c +8 -0
index 624a26794d55..e5ba9795ec69 100644
@@ -285,6 +285,14 @@ static void acpi_ut_delete_internal_obj(union acpi_operand_object *object)
                 }
                 break;
 
+       case ACPI_TYPE_LOCAL_ADDRESS_HANDLER:
+
+               ACPI_DEBUG_PRINT((ACPI_DB_ALLOCATIONS,
+                                 "***** Address handler %p\n", object));
+
+               acpi_os_delete_mutex(object->address_space.context_mutex);
+               break;
+
         default:
 
                 break;