mlxsw: thermal: Fix out-of-bounds memory accesses
authorIdo Schimmel <idosch@nvidia.com>
Tue, 12 Oct 2021 17:49:55 +0000 (20:49 +0300)
committerJakub Kicinski <kuba@kernel.org>
Thu, 14 Oct 2021 14:13:26 +0000 (07:13 -0700)
commit332fdf951df8b870e3da86b122ae304e2aabe88c
treeaf2f19768febc47ec638b06ca316013db39edfb1
parent40507e7aada8422c38aafa0c8a1a09e4623c712a
mlxsw: thermal: Fix out-of-bounds memory accesses

Currently, mlxsw allows cooling states to be set above the maximum
cooling state supported by the driver:

 # cat /sys/class/thermal/thermal_zone2/cdev0/type
 mlxsw_fan
 # cat /sys/class/thermal/thermal_zone2/cdev0/max_state
 10
 # echo 18 > /sys/class/thermal/thermal_zone2/cdev0/cur_state
 # echo $?
 0

This results in out-of-bounds memory accesses when thermal state
transition statistics are enabled (CONFIG_THERMAL_STATISTICS=y), as the
transition table is accessed with a too large index (state) [1].

According to the thermal maintainer, it is the responsibility of the
driver to reject such operations [2].

Therefore, return an error when the state to be set exceeds the maximum
cooling state supported by the driver.

To avoid dead code, as suggested by the thermal maintainer [3],
partially revert commit a421ce088ac8 ("mlxsw: core: Extend cooling
device with cooling levels") that tried to interpret these invalid
cooling states (above the maximum) in a special way. The cooling levels
array is not removed in order to prevent the fans going below 20% PWM,
which would cause them to get stuck at 0% PWM.

[1]
BUG: KASAN: slab-out-of-bounds in thermal_cooling_device_stats_update+0x271/0x290
Read of size 4 at addr ffff8881052f7bf8 by task kworker/0:0/5

CPU: 0 PID: 5 Comm: kworker/0:0 Not tainted 5.15.0-rc3-custom-45935-gce1adf704b14 #122
Hardware name: Mellanox Technologies Ltd. "MSN2410-CB2FO"/"SA000874", BIOS 4.6.5 03/08/2016
Workqueue: events_freezable_power_ thermal_zone_device_check
Call Trace:
 dump_stack_lvl+0x8b/0xb3
 print_address_description.constprop.0+0x1f/0x140
 kasan_report.cold+0x7f/0x11b
 thermal_cooling_device_stats_update+0x271/0x290
 __thermal_cdev_update+0x15e/0x4e0
 thermal_cdev_update+0x9f/0xe0
 step_wise_throttle+0x770/0xee0
 thermal_zone_device_update+0x3f6/0xdf0
 process_one_work+0xa42/0x1770
 worker_thread+0x62f/0x13e0
 kthread+0x3ee/0x4e0
 ret_from_fork+0x1f/0x30

Allocated by task 1:
 kasan_save_stack+0x1b/0x40
 __kasan_kmalloc+0x7c/0x90
 thermal_cooling_device_setup_sysfs+0x153/0x2c0
 __thermal_cooling_device_register.part.0+0x25b/0x9c0
 thermal_cooling_device_register+0xb3/0x100
 mlxsw_thermal_init+0x5c5/0x7e0
 __mlxsw_core_bus_device_register+0xcb3/0x19c0
 mlxsw_core_bus_device_register+0x56/0xb0
 mlxsw_pci_probe+0x54f/0x710
 local_pci_probe+0xc6/0x170
 pci_device_probe+0x2b2/0x4d0
 really_probe+0x293/0xd10
 __driver_probe_device+0x2af/0x440
 driver_probe_device+0x51/0x1e0
 __driver_attach+0x21b/0x530
 bus_for_each_dev+0x14c/0x1d0
 bus_add_driver+0x3ac/0x650
 driver_register+0x241/0x3d0
 mlxsw_sp_module_init+0xa2/0x174
 do_one_initcall+0xee/0x5f0
 kernel_init_freeable+0x45a/0x4de
 kernel_init+0x1f/0x210
 ret_from_fork+0x1f/0x30

The buggy address belongs to the object at ffff8881052f7800
 which belongs to the cache kmalloc-1k of size 1024
The buggy address is located 1016 bytes inside of
 1024-byte region [ffff8881052f7800ffff8881052f7c00)
The buggy address belongs to the page:
page:0000000052355272 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x1052f0
head:0000000052355272 order:3 compound_mapcount:0 compound_pincount:0
flags: 0x200000000010200(slab|head|node=0|zone=2)
raw: 0200000000010200 ffffea0005034800 0000000300000003 ffff888100041dc0
raw: 0000000000000000 0000000000100010 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff8881052f7a80: 00 00 00 00 00 00 04 fc fc fc fc fc fc fc fc fc
 ffff8881052f7b00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>ffff8881052f7b80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
                                                                ^
 ffff8881052f7c00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 ffff8881052f7c80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc

[2] https://lore.kernel.org/linux-pm/9aca37cb-1629-5c67-1895-1fdc45c0244e@linaro.org/
[3] https://lore.kernel.org/linux-pm/af9857f2-578e-de3a-e62b-6baff7e69fd4@linaro.org/

CC: Daniel Lezcano <daniel.lezcano@linaro.org>
Fixes: a50c1e35650b ("mlxsw: core: Implement thermal zone")
Fixes: a421ce088ac8 ("mlxsw: core: Extend cooling device with cooling levels")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Vadim Pasternak <vadimp@nvidia.com>
Link: https://lore.kernel.org/r/20211012174955.472928-1-idosch@idosch.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
drivers/net/ethernet/mellanox/mlxsw/core_thermal.c