From: "Gustavo A. R. Silva" <gustavoars@kernel.org> Date: 2021-03-31 22:44:46
Fix the a couple of out-of-bounds warnings by making the code
a bit more structured.
This helps with the ongoing efforts to enable -Warray-bounds and
avoid confusing the compiler.
Link: https://github.com/KSPP/linux/issues/109
Changes in v2:
- Update changelog text in patch 1/2.
- Replace a couple of magic numbers with new variable sig_addr_len.
Gustavo A. R. Silva (2):
wl3501_cs: Fix out-of-bounds warning in wl3501_send_pkt
wl3501_cs: Fix out-of-bounds warning in wl3501_mgmt_join
drivers/net/wireless/wl3501.h | 28 ++++++++++++++++------------
drivers/net/wireless/wl3501_cs.c | 11 ++++++-----
2 files changed, 22 insertions(+), 17 deletions(-)
--
2.27.0
From: "Gustavo A. R. Silva" <gustavoars@kernel.org> Date: 2021-03-31 22:45:17
Fix the following out-of-bounds warning by enclosing
structure members daddr and saddr into new struct addr:
arch/x86/include/asm/string_32.h:182:25: warning: '__builtin_memcpy' offset [18, 23] from the object at 'sig' is out of the bounds of referenced subobject 'daddr' with type 'u8[6]' {aka 'unsigned char[6]'} at offset 11 [-Warray-bounds]
Refactor the code, accordingly:
$ pahole -C wl3501_md_req drivers/net/wireless/wl3501_cs.o
struct wl3501_md_req {
u16 next_blk; /* 0 2 */
u8 sig_id; /* 2 1 */
u8 routing; /* 3 1 */
u16 data; /* 4 2 */
u16 size; /* 6 2 */
u8 pri; /* 8 1 */
u8 service_class; /* 9 1 */
struct {
u8 daddr[6]; /* 10 6 */
u8 saddr[6]; /* 16 6 */
} addr; /* 10 12 */
/* size: 22, cachelines: 1, members: 8 */
/* last cacheline: 22 bytes */
};
The problem is that the original code is trying to copy data into a
couple of arrays adjacent to each other in a single call to memcpy().
Now that a new struct _addr_ enclosing those two adjacent arrays
is introduced, memcpy() doesn't overrun the length of &sig.daddr[0],
because the address of the new struct object _addr_ is used as
destination, instead.
Also, this helps with the ongoing efforts to enable -Warray-bounds and
avoid confusing the compiler.
Link: https://github.com/KSPP/linux/issues/109
Reported-by: kernel test robot <redacted>
Build-tested-by: kernel test robot [off-list ref]
Link: https://lore.kernel.org/lkml/60641d9b.2eNLedOGSdcSoAV2%25lkp@intel.com/
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
---
Changes in v2:
- Update changelog text.
- Replace a couple of magic numbers with new variable sig_addr_len.
drivers/net/wireless/wl3501.h | 6 ++++--
drivers/net/wireless/wl3501_cs.c | 7 ++++---
2 files changed, 8 insertions(+), 5 deletions(-)
From: "Gustavo A. R. Silva" <gustavoars@kernel.org> Date: 2021-03-31 22:46:21
Fix the following out-of-bounds warning by enclosing
some structure members into new struct req:
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]
Refactor the code, accordingly:
$ 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 {
u16 beacon_period; /* 36 2 */
u16 dtim_period; /* 38 2 */
u16 cap_info; /* 40 2 */
u8 bss_type; /* 42 1 */
u8 bssid[6]; /* 43 6 */
struct iw_mgmt_essid_pset ssid; /* 49 34 */
/* --- cacheline 1 boundary (64 bytes) was 19 bytes ago --- */
struct iw_mgmt_ds_pset ds_pset; /* 83 3 */
struct iw_mgmt_cf_pset cf_pset; /* 86 8 */
struct iw_mgmt_ibss_pset ibss_pset; /* 94 4 */
struct iw_mgmt_data_rset bss_basic_rset; /* 98 10 */
} req; /* 36 72 */
/* size: 108, cachelines: 2, members: 10 */
/* last cacheline: 44 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 _req_ enclosing all those adjacent
members is introduced, memcpy() doesn't overrun the length of
&sig.beacon_period, because the address of the new struct object
_req_ is used as the destination, instead.
Also, this helps with the ongoing efforts to enable -Warray-bounds and
avoid confusing the compiler.
Link: https://github.com/KSPP/linux/issues/109
Reported-by: kernel test robot <redacted>
Build-tested-by: kernel test robot [off-list ref]
Link: https://lore.kernel.org/lkml/60641d9b.2eNLedOGSdcSoAV2%25lkp@intel.com/
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
---
Changes in v2:
- None.
drivers/net/wireless/wl3501.h | 22 ++++++++++++----------
drivers/net/wireless/wl3501_cs.c | 4 ++--
2 files changed, 14 insertions(+), 12 deletions(-)
On Wed, Mar 31, 2021 at 04:44:29PM -0500, Gustavo A. R. Silva wrote:
Fix the following out-of-bounds warning by enclosing
structure members daddr and saddr into new struct addr:
arch/x86/include/asm/string_32.h:182:25: warning: '__builtin_memcpy' offset [18, 23] from the object at 'sig' is out of the bounds of referenced subobject 'daddr' with type 'u8[6]' {aka 'unsigned char[6]'} at offset 11 [-Warray-bounds]
Refactor the code, accordingly:
$ pahole -C wl3501_md_req drivers/net/wireless/wl3501_cs.o
struct wl3501_md_req {
u16 next_blk; /* 0 2 */
u8 sig_id; /* 2 1 */
u8 routing; /* 3 1 */
u16 data; /* 4 2 */
u16 size; /* 6 2 */
u8 pri; /* 8 1 */
u8 service_class; /* 9 1 */
struct {
u8 daddr[6]; /* 10 6 */
u8 saddr[6]; /* 16 6 */
} addr; /* 10 12 */
/* size: 22, cachelines: 1, members: 8 */
/* last cacheline: 22 bytes */
};
The problem is that the original code is trying to copy data into a
couple of arrays adjacent to each other in a single call to memcpy().
Now that a new struct _addr_ enclosing those two adjacent arrays
is introduced, memcpy() doesn't overrun the length of &sig.daddr[0],
because the address of the new struct object _addr_ is used as
destination, instead.
Also, this helps with the ongoing efforts to enable -Warray-bounds and
avoid confusing the compiler.
Link: https://github.com/KSPP/linux/issues/109
Reported-by: kernel test robot <redacted>
Build-tested-by: kernel test robot [off-list ref]
Link: https://lore.kernel.org/lkml/60641d9b.2eNLedOGSdcSoAV2%25lkp@intel.com/
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Thanks, this makes the code much easier for the compiler to validate
at compile time. These cross-field memcpy()s are weird. I like the
solution here.
Reviewed-by: Kees Cook <redacted>
--
Kees Cook
On Wed, Mar 31, 2021 at 04:45:34PM -0500, Gustavo A. R. Silva wrote:
Fix the following out-of-bounds warning by enclosing
some structure members into new struct req:
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]
Refactor the code, accordingly:
$ 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 {
u16 beacon_period; /* 36 2 */
u16 dtim_period; /* 38 2 */
u16 cap_info; /* 40 2 */
u8 bss_type; /* 42 1 */
u8 bssid[6]; /* 43 6 */
struct iw_mgmt_essid_pset ssid; /* 49 34 */
/* --- cacheline 1 boundary (64 bytes) was 19 bytes ago --- */
struct iw_mgmt_ds_pset ds_pset; /* 83 3 */
struct iw_mgmt_cf_pset cf_pset; /* 86 8 */
struct iw_mgmt_ibss_pset ibss_pset; /* 94 4 */
struct iw_mgmt_data_rset bss_basic_rset; /* 98 10 */
} req; /* 36 72 */
This section is the same as a large portion of struct wl3501_scan_confirm:
struct wl3501_scan_confirm {
u16 next_blk;
u8 sig_id;
u8 reserved;
u16 status;
char timestamp[8];
char localtime[8];
from here
u16 beacon_period;
u16 dtim_period;
u16 cap_info;
u8 bss_type;
u8 bssid[ETH_ALEN];
struct iw_mgmt_essid_pset ssid;
struct iw_mgmt_ds_pset ds_pset;
struct iw_mgmt_cf_pset cf_pset;
struct iw_mgmt_ibss_pset ibss_pset;
struct iw_mgmt_data_rset bss_basic_rset;
through here
u8 rssi;
};
It seems like maybe extracting that and using it in both structures
would make more sense?
quoted hunk
/* size: 108, cachelines: 2, members: 10 */
/* last cacheline: 44 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 _req_ enclosing all those adjacent
members is introduced, memcpy() doesn't overrun the length of
&sig.beacon_period, because the address of the new struct object
_req_ is used as the destination, instead.
Also, this helps with the ongoing efforts to enable -Warray-bounds and
avoid confusing the compiler.
Link: https://github.com/KSPP/linux/issues/109
Reported-by: kernel test robot <redacted>
Build-tested-by: kernel test robot [off-list ref]
Link: https://lore.kernel.org/lkml/60641d9b.2eNLedOGSdcSoAV2%25lkp@intel.com/
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
---
Changes in v2:
- None.
drivers/net/wireless/wl3501.h | 22 ++++++++++++----------
drivers/net/wireless/wl3501_cs.c | 4 ++--
2 files changed, 14 insertions(+), 12 deletions(-)
If not, then probably something like this should be added to make sure
nothing unexpected happens to change structure sizes:
BUILD_BUG_ON(sizeof(sig.req) != 72);
From: Gustavo A. R. Silva <hidden> Date: 2021-04-13 21:27:52
Hi all,
Friendly ping: could somebody give us some feedback or take
this series, please?
Thanks
--
Gustavo
On 3/31/21 16:43, Gustavo A. R. Silva wrote:
Fix the a couple of out-of-bounds warnings by making the code
a bit more structured.
This helps with the ongoing efforts to enable -Warray-bounds and
avoid confusing the compiler.
Link: https://github.com/KSPP/linux/issues/109
Changes in v2:
- Update changelog text in patch 1/2.
- Replace a couple of magic numbers with new variable sig_addr_len.
Gustavo A. R. Silva (2):
wl3501_cs: Fix out-of-bounds warning in wl3501_send_pkt
wl3501_cs: Fix out-of-bounds warning in wl3501_mgmt_join
drivers/net/wireless/wl3501.h | 28 ++++++++++++++++------------
drivers/net/wireless/wl3501_cs.c | 11 ++++++-----
2 files changed, 22 insertions(+), 17 deletions(-)
From: Gustavo A. R. Silva <hidden> Date: 2021-04-13 21:46:03
Hi all!
On 4/7/21 14:02, Kees Cook wrote:
On Wed, Mar 31, 2021 at 04:45:34PM -0500, Gustavo A. R. Silva wrote:
quoted
Fix the following out-of-bounds warning by enclosing
some structure members into new struct req:
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]
Refactor the code, accordingly:
$ 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 {
u16 beacon_period; /* 36 2 */
u16 dtim_period; /* 38 2 */
u16 cap_info; /* 40 2 */
u8 bss_type; /* 42 1 */
u8 bssid[6]; /* 43 6 */
struct iw_mgmt_essid_pset ssid; /* 49 34 */
/* --- cacheline 1 boundary (64 bytes) was 19 bytes ago --- */
struct iw_mgmt_ds_pset ds_pset; /* 83 3 */
struct iw_mgmt_cf_pset cf_pset; /* 86 8 */
struct iw_mgmt_ibss_pset ibss_pset; /* 94 4 */
struct iw_mgmt_data_rset bss_basic_rset; /* 98 10 */
} req; /* 36 72 */
This section is the same as a large portion of struct wl3501_scan_confirm:
struct wl3501_scan_confirm {
u16 next_blk;
u8 sig_id;
u8 reserved;
u16 status;
char timestamp[8];
char localtime[8];
from here
u16 beacon_period;
u16 dtim_period;
u16 cap_info;
u8 bss_type;
u8 bssid[ETH_ALEN];
struct iw_mgmt_essid_pset ssid;
struct iw_mgmt_ds_pset ds_pset;
struct iw_mgmt_cf_pset cf_pset;
struct iw_mgmt_ibss_pset ibss_pset;
struct iw_mgmt_data_rset bss_basic_rset;
through here
u8 rssi;
};
It seems like maybe extracting that and using it in both structures
would make more sense?
If I do this, I would therefore have to make a bunch of other changes,
accordingly. I'm OK with that but I'd like to have the opinion of the
maintainers on all this. So, I will go and ping them from the cover
letter of this series with the hope that we can get some feedback from
them. :) They have been silent for a couple of weeks now.
quoted
/* size: 108, cachelines: 2, members: 10 */
/* last cacheline: 44 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 _req_ enclosing all those adjacent
members is introduced, memcpy() doesn't overrun the length of
&sig.beacon_period, because the address of the new struct object
_req_ is used as the destination, instead.
Also, this helps with the ongoing efforts to enable -Warray-bounds and
avoid confusing the compiler.
Link: https://github.com/KSPP/linux/issues/109
Reported-by: kernel test robot <redacted>
Build-tested-by: kernel test robot [off-list ref]
Link: https://lore.kernel.org/lkml/60641d9b.2eNLedOGSdcSoAV2%25lkp@intel.com/
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
---
Changes in v2:
- None.
drivers/net/wireless/wl3501.h | 22 ++++++++++++----------
drivers/net/wireless/wl3501_cs.c | 4 ++--
2 files changed, 14 insertions(+), 12 deletions(-)
If not, then probably something like this should be added to make sure
nothing unexpected happens to change structure sizes:
BUILD_BUG_ON(sizeof(sig.req) != 72);
Yep, this is sensible.
Thanks for the feedback!
--
Gustavo