Describe the bug The instruction hook in QEMU systemmode causes a

Systemmode instruction hook causes a crash about qemu-libafl-bridge HOT 3 CLOSED

aflplusplus commented on July 17, 2024

Systemmode instruction hook causes a crash

from qemu-libafl-bridge.

Comments (3)

andreafioraldi commented on July 17, 2024

Is the current main branch still affected? I merged an update of the helpers calls in the last few days.
Btw, the diff between the commit w/o and w/ the bug is https://github.com/AFLplusplus/qemu-libafl-bridge/compare/f6a2e732e8e225ebb8d1a9399561af7330af31b3..0dc52ed6f3915f727aaec8648706760f278f0571 and the instruction hooks were not touched, so probably a bug introduced with the merge from upstream

from qemu-libafl-bridge.

alwinber commented on July 17, 2024

Yes, i took the current main branch at eae6f04, which has QEMU_REVISION = 3a774d.

from qemu-libafl-bridge.

andreafioraldi commented on July 17, 2024

for future reference: TCGHelperInfo for the libafl_hook strict was partially not initialized and it worked till a specific commit of QEMU because the code was just writing to it, while it checks the values and so it was keeping the dirty memory from malloc. Specifically, the assert was triggered because info->nr_in was trash (85 in my debugger).

from qemu-libafl-bridge.

Recommend Projects