hw: usb: hcd-ohci: fix oob access and loop issues
v2 cover hw: usb: hcd-ohci: fix oob access and loop issues
From: Prasad J Pandit <[email protected]>
Hello,
* While servicing transfer descriptors(TD) in ohci_service[_iso]_td routines, it may lead to out-of-bounds access and/or infinite loop issues, as the OHCI controller driver may supply malicious values to derive frame_number, start_addr, end_addr etc. variables.
* This series breaks earlier single patch into two.
One for an out-of-bounds access issue and another to fix infinite loop case.
-> https://lists.nongnu.org/archive/html/qemu-devel/2020-09/msg05145.html
Thank you.
--
Prasad J Pandit (2):
hw: usb: hcd-ohci: check len and frame_number variables
hw: usb: hcd-ohci: check for processed TD before retire
hw/usb/hcd-ohci.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-)
--
2.26.2
PATCH v2 1/2 hw: usb: hcd-ohci: check len and frame_number variables
From: Prasad J Pandit <[email protected]>
While servicing the OHCI transfer descriptors(TD), OHCI host controller derives variables 'start_addr', 'end_addr', 'len' etc. from values supplied by the host controller driver. Host controller driver may supply values such that using above variables leads to out-of-bounds access issues. Add checks to avoid them.
AddressSanitizer: stack-buffer-overflow on address 0x7ffd53af76a0
READ of size 2 at 0x7ffd53af76a0 thread T0
#0 ohci_service_iso_td ../hw/usb/hcd-ohci.c:734
#1 ohci_service_ed_list ../hw/usb/hcd-ohci.c:1180
#2 ohci_process_lists ../hw/usb/hcd-ohci.c:1214
#3 ohci_frame_boundary ../hw/usb/hcd-ohci.c:1257
#4 timerlist_run_timers ../util/qemu-timer.c:572
#5 qemu_clock_run_timers ../util/qemu-timer.c:586
#6 qemu_clock_run_all_timers ../util/qemu-timer.c:672 #7 main_loop_wait ../util/main-loop.c:527
#8 qemu_main_loop ../softmmu/vl.c:1676
#9 main ../softmmu/main.c:50
--- hw/usb/hcd-ohci.c | 24 ++++++++++++++++++++++-- 1 file changed, 22 insertions(+), 2 deletions(-)
Update v2: one patch to fix oob access
-> https://lists.nongnu.org/archive/html/qemu-devel/2020-09/msg05145.html
hw/usb/hcd-ohci.c +22 -2
@@ -731,7 +731,11 @@ static int ohci_service_iso_td(OHCIState *ohci, struct ohci_ed *ed, } start_offset = iso_td.offset[relative_frame_number]; - next_offset = iso_td.offset[relative_frame_number +1 ]; +if (relative_frame_number < frame_count) { + next_offset = iso_td.offset[relative_frame_number +1 ]; + }else { + next_offset = iso_td.be; + } if (!(OHCI_BM(start_offset, TD_PSW_CC) & 0xe) || ((relative_frame_number < frame_count) &&
@@ -764,7 +768,12 @@ static int ohci_service_iso_td(OHCIState *ohci, struct ohci_ed *ed, } }else {/* Last packet in the ISO TD */ - end_addr = iso_td.be; + end_addr = next_offset; + } + +if (start_addr > end_addr) { +trace_usb_ohci_iso_td_bad_cc_overrun (start_addr, end_addr); +return 1 ; }if ((start_addr &OHCI_PAGE_MASK ) != (end_addr &OHCI_PAGE_MASK )) {
@@ -773,6 +782,9 @@ static int ohci_service_iso_td(OHCIState *ohci, struct ohci_ed *ed, }else { len = end_addr - start_addr +1 ; } +if (len >sizeof (ohci->usb_buf)) { + len =sizeof (ohci->usb_buf); + }if (len && dir !=OHCI_TD_DIR_IN ) { if (ohci_copy_iso_td(ohci, start_addr, end_addr, ohci->usb_buf, len,
@@ -975,8 +987,16 @@ static int ohci_service_td(OHCIState *ohci, struct ohci_ed *ed)if ((td.cbp &0xfffff000 ) != (td.be &0xfffff000 )) { len = (td.be &0xfff ) +0x1001 - (td.cbp &0xfff ); }else { +if (td.cbp > td.be) { +trace_usb_ohci_iso_td_bad_cc_overrun (td.cbp, td.be); +ohci_die (ohci); +return 1 ; + } len = (td.be - td.cbp) +1 ; } +if (len >sizeof (ohci->usb_buf)) { + len =sizeof (ohci->usb_buf); + } pktlen = len;if (len && dir !=OHCI_TD_DIR_IN ) {
Signature · 1 line
2.26.2
PATCH v2 2/2 hw: usb: hcd-ohci: check for processed TD before retire
In reply to P J PFrom: Prasad J Pandit <[email protected]>
While servicing OHCI transfer descriptors(TD), ohci_service_iso_td retires a TD if it has passed its time frame. It does not check if the TD was already processed once and holds an error code in TD_CC. It may happen if the TD list has a loop. Add check to avoid an infinite loop condition.
--- hw/usb/hcd-ohci.c | 4 ++++ 1 file changed, 4 insertions(+)
Update v2: one patch for loop issue
-> https://lists.nongnu.org/archive/html/qemu-devel/2020-09/msg05145.html
hw/usb/hcd-ohci.c +4 -0
@@ -691,6 +691,10 @@ static int ohci_service_iso_td(OHCIState *ohci, struct ohci_ed *ed,the next ISO TD of the same ED */ trace_usb_ohci_iso_td_relative_frame_number_big (relative_frame_number, frame_count); +if (OHCI_CC_DATAOVERRUN ==OHCI_BM (iso_td.flags,TD_CC )) { +/* avoid infinite loop */ +return 1 ; + }OHCI_SET_BM (iso_td.flags,TD_CC ,OHCI_CC_DATAOVERRUN ); ed->head &= ~OHCI_DPTR_MASK ; ed->head |= (iso_td.next &OHCI_DPTR_MASK );
Signature · 1 line
2.26.2
Re: [PATCH v2 2/2] hw: usb: hcd-ohci: check for processed TD before retire
Quoted 2 lines · P J P <[email protected]> 于2020年9月16日周三 上午2:25写道:
From: Prasad J Pandit <[email protected]>
Quoted 33 lines
While servicing OHCI transfer descriptors(TD),
ohci_service_iso_tdretires a TD if it has passed its time frame. It does not check if the TD was already processed once and holds an error code inTD_CC. It may happen if the TD list has a loop. Add check to avoid an infinite loop condition.Signed-off-by Prasad J Pandit <[email protected]>--- hw/usb/hcd-ohci.c | 4 ++++ 1 file changed, 4 insertions(+)Update v2: one patch for loop issue
-> https://lists.nongnu.org/archive/html/qemu-devel/2020-09/msg05145.htmlhw/usb/hcd-ohci.c +4
-1@@ -691,6 +691,10 @@ static int ohci_service_iso_td(OHCIState *ohci, struct ohci_ed *ed,the next ISO TD of the same ED */ trace_usb_ohci_iso_td_relative_frame_number_big (relative_frame_number, frame_count); +if (OHCI_CC_DATAOVERRUN ==OHCI_BM (iso_td.flags,TD_CC )) { +/* avoid infinite loop */ +return 1 ; + }OHCI_SET_BM (iso_td.flags,TD_CC ,OHCI_CC_DATAOVERRUN ); ed->head &= ~OHCI_DPTR_MASK ; ed->head |= (iso_td.next &OHCI_DPTR_MASK ); --2.26.2
Quoted 9 lines · P J P
From: Prasad J Pandit <[email protected]>
Hello,
* While servicing transfer descriptors(TD) in ohci_service[_iso]_td routines, it may lead to out-of-bounds access and/or infinite loop issues, as the OHCI controller driver may supply malicious values to derive
frame_number,start_addr,end_addretc. variables.
Quoted 4 lines
* This series breaks earlier single patch into two.
One for an out-of-bounds access issue and another to fix infinite loop case.
-> https://lists.nongnu.org/archive/html/qemu-devel/2020-09/msg05145.html
Added to usb patch queue.
thanks,
Gerd