staging: rtl8723bs: properly validate the data in rtw_get_ie_ex()
Greg Kroah-Hartman
maintainerPATCH 1/1 staging: rtl8723bs: properly validate the data in rtw_get_ie_ex()
Just like in commit 154828bf9559 ("staging: rtl8723bs: fix out-of-bounds read in rtw_get_ie() parser"), we don't trust the data in the frame so we should check the length better before acting on it
---
Navaneeth, any chance you can test this or at least verify my logic is correct here? I got a "hit" from a tool that the work you did in your commit also needs to be done here, and I _think_ I got it right but do not have the hardware to test this with at all. Thanks!
drivers/staging/rtl8723bs/core/rtw_ieee80211.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-)
drivers/staging/rtl8723bs/core/rtw_ieee80211.c +10 -5
@@ -186,20 +186,25 @@ u8 *rtw_get_ie_ex(u8 *in_ie, uint in_len, u8 eid, u8 *oui, u8 oui_len, u8 *ie, u cnt =0 ; -while (cnt < in_len) { +while (cnt +2 <= in_len) { +u8 ie_len = in_ie[cnt +1 ]; + +if (cnt +2 + ie_len > in_len) +break ; +if (eid == in_ie[cnt] - && (!oui || !memcmp (&in_ie[cnt+2 ], oui, oui_len))) { + && (!oui || (ie_len >= oui_len && !memcmp (&in_ie[cnt +2 ], oui, oui_len)))) { target_ie = &in_ie[cnt];if (ie) -memcpy (ie, &in_ie[cnt], in_ie[cnt+1 ]+2 ); +memcpy (ie, &in_ie[cnt], ie_len +2 );if (ielen) - *ielen = in_ie[cnt+1 ]+2 ; + *ielen = ie_len +2 ;break ; } - cnt += in_ie[cnt+1 ]+2 ;/* goto next */ + cnt += ie_len +2 ;/* goto next */ }return target_ie;
Signature · 1 line
2.53.0
I don't have the physical RTL8723BS hardware on hand anymore to test a live connection, but I was able to verify your logic thoroughly in user-space.
To be absolutely sure, I extracted both the old and patched functions into a standalone C harness and ran them through Asan and AFL++.
Feeding a crafted 5-byte allocation(with a lying length byte) to the old code predictably triggered a 47-byte heap-buffer-overflow right at the memcpy.
Against your patched code, I ran that same payload along with 20 other edge-case tests (1-byte buffers, empty bodies, OUI boundary mismatches, etc.). It cleanly rejected all of them with zero ASan errors.
I also compiled the patched function as an AFL++ target and let it freely mutate the EID, length, and body bytes. After over 100,000 executions, it reported 0 crashes and 0 hangs.
The logic is definitely solid. Changing the loop guard to "while (cnt + 2 <= in_len)" guarantees we always have at least 2 bytes before we touch the EID and length. Reading ie_len once and explicitly checking if it exceeds in_len completely stops the memcpy from reading past the end of the allocation.
I have the raw ASan logs, AFL stats, and the C test harnesses saved if needed!.
Quoted 52 lines · Greg Kroah-Hartman
Just like in commit
154828bf9559("staging: rtl8723bs: fix out-of-bounds read inrtw_get_ie()parser"), we don't trust the data in the frame so we should check the length better before acting on itCc Navaneeth K <[email protected]>Cc stable <[email protected]>Assisted-by gkh_clanker_2000Signed-off-by Greg Kroah-Hartman <[email protected]>---
Navaneeth, any chance you can test this or at least verify my logic is correct here? I got a "hit" from a tool that the work you did in your commit also needs to be done here, and I _think_ I got it right but do not have the hardware to test this with at all. Thanks!drivers/staging/rtl8723bs/core/rtw_ieee80211.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-)drivers/staging/rtl8723bs/core/rtw_ieee80211.c +10
-5@@ -186,20 +186,25 @@ u8 *rtw_get_ie_ex(u8 *in_ie, uint in_len, u8 eid, u8 *oui, u8 oui_len, u8 *ie, u cnt =0 ; -while (cnt < in_len) { +while (cnt +2 <= in_len) { +u8 ie_len = in_ie[cnt +1 ]; + +if (cnt +2 + ie_len > in_len) +break ; +if (eid == in_ie[cnt] - && (!oui || !memcmp (&in_ie[cnt+2 ], oui, oui_len))) { + && (!oui || (ie_len >= oui_len && !memcmp (&in_ie[cnt +2 ], oui, oui_len)))) { target_ie = &in_ie[cnt];if (ie) -memcpy (ie, &in_ie[cnt], in_ie[cnt+1 ]+2 ); +memcpy (ie, &in_ie[cnt], ie_len +2 );if (ielen) - *ielen = in_ie[cnt+1 ]+2 ; + *ielen = ie_len +2 ;break ; } - cnt += in_ie[cnt+1 ]+2 ;/* goto next */ + cnt += ie_len +2 ;/* goto next */ }return target_ie;
Greg Kroah-Hartman
maintainerQuoted 18 lines · Navaneeth K
I don't have the physical RTL8723BS hardware on hand anymore to test a live connection, but I was able to verify your logic thoroughly in user-space. To be absolutely sure, I extracted both the old and patched functions into a standalone C harness and ran them through Asan and AFL++. Feeding a crafted 5-byte allocation(with a lying length byte) to the old code predictably triggered a 47-byte heap-buffer-overflow right at the memcpy.
Against your patched code, I ran that same payload along with 20 other edge-case tests (1-byte buffers, empty bodies, OUI boundary mismatches, etc.). It cleanly rejected all of them with zero ASan errors.
I also compiled the patched function as an AFL++ target and let it freely mutate the EID, length, and body bytes. After over 100,000 executions, it reported 0 crashes and 0 hangs.
The logic is definitely solid. Changing the loop guard to "while (cnt + 2 <=in_len)" guarantees we always have at least 2 bytes before we touch the EID and length. Readingie_lenonce and explicitly checking if it exceedsin_lencompletely stops the memcpy from reading past the end of the allocation.
Quoted 5 lines
I have the raw ASan logs, AFL stats, and the C test harnesses saved if needed!.
Tested-by Navaneeth K <[email protected]>Reviewed-by Navaneeth K <[email protected]>
Wow, that's way more than I expected, thank you! Raw logs and test harness is not needed, I'll trust you, worst case, we revert it if someone complains it breaks their device :)
thanks,
greg k-h