On 4/15/21 14:58, Kees Cook wrote:
> On Wed, Apr 14, 2021 at 06:45:15PM -0500, Gustavo A. R. Silva wrote:
>> Fix the following out-of-bounds warnings by adding a new structure
>> wl3501_req instead of duplicating the same members in structure
>> wl3501_join_req and wl3501_scan_confirm:
>>
>> arch/x86/include/asm/string_32.h:182:25: warning: '__builtin_memcpy' offset
>> [39, 108] from the object at 'sig' is out of the bounds of referenced
>> subobject 'beacon_period' with type 'short unsigned int' at offset 36
>> [-Warray-bounds]
>> arch/x86/include/asm/string_32.h:182:25: warning: '__builtin_memcpy' offset
>> [25, 95] from the object at 'sig' is out of the bounds of referenced
>> subobject 'beacon_period' with type 'short unsigned int' at offset 22
>> [-Warray-bounds]
>>
>> Refactor the code, accordingly:
>>
>> $ pahole -C wl3501_req drivers/net/wireless/wl3501_cs.o
>> struct wl3501_req {
>> u16 beacon_period; /* 0 2 */
>> u16 dtim_period; /* 2 2 */
>> u16 cap_info; /* 4 2 */
>> u8 bss_type; /* 6 1 */
>> u8 bssid[6]; /* 7 6 */
>> struct iw_mgmt_essid_pset ssid; /* 13 34 */
>> struct iw_mgmt_ds_pset ds_pset; /* 47 3 */
>> struct iw_mgmt_cf_pset cf_pset; /* 50 8 */
>> struct iw_mgmt_ibss_pset ibss_pset; /* 58 4 */
>> struct iw_mgmt_data_rset bss_basic_rset; /* 62 10 */
>>
>> /* size: 72, cachelines: 2, members: 10 */
>> /* last cacheline: 8 bytes */
>> };
>>
>> $ pahole -C wl3501_join_req drivers/net/wireless/wl3501_cs.o
>> struct wl3501_join_req {
>> u16 next_blk; /* 0 2 */
>> u8 sig_id; /* 2 1 */
>> u8 reserved; /* 3 1 */
>> struct iw_mgmt_data_rset operational_rset; /* 4 10 */
>> u16 reserved2; /* 14 2 */
>> u16 timeout; /* 16 2 */
>> u16 probe_delay; /* 18 2 */
>> u8 timestamp[8]; /* 20 8 */
>> u8 local_time[8]; /* 28 8 */
>> struct wl3501_req req; /* 36 72 */
>>
>> /* size: 108, cachelines: 2, members: 10 */
>> /* last cacheline: 44 bytes */
>> };
>>
>> $ pahole -C wl3501_scan_confirm drivers/net/wireless/wl3501_cs.o
>> struct wl3501_scan_confirm {
>> u16 next_blk; /* 0 2 */
>> u8 sig_id; /* 2 1 */
>> u8 reserved; /* 3 1 */
>> u16 status; /* 4 2 */
>> char timestamp[8]; /* 6 8 */
>> char localtime[8]; /* 14 8 */
>> struct wl3501_req req; /* 22 72 */
>> /* --- cacheline 1 boundary (64 bytes) was 30 bytes ago --- */
>> u8 rssi; /* 94 1 */
>>
>> /* size: 96, cachelines: 2, members: 8 */
>> /* padding: 1 */
>> /* last cacheline: 32 bytes */
>> };
>>
>> The problem is that the original code is trying to copy data into a
>> bunch of struct members adjacent to each other in a single call to
>> memcpy(). Now that a new struct wl3501_req enclosing all those adjacent
>> members is introduced, memcpy() doesn't overrun the length of
>> &sig.beacon_period and &this->bss_set[i].beacon_period, because the
>> address of the new struct object _req_ is used as the destination,
>> instead.
>>
>> This helps with the ongoing efforts to globally enable -Warray-bounds
>> and get us closer to being able to tighten the FORTIFY_SOURCE routines
>> on memcpy().
>>
>> Link: https://github.com/KSPP/linux/issues/109
>> Reported-by: kernel test robot <l...@intel.com>
>> Signed-off-by: Gustavo A. R. Silva <gustavo...@kernel.org>
>
> Awesome! Thank you for this solution.
>
> Reviewed-by: Kees Cook <keesc...@chromium.org>
Thanks, Kees!
--
Gustavo
