"value":"In the Linux kernel, the following vulnerability has been resolved:\n\nbpf: Fix a sdiv overflow issue\n\nZac Ecob reported a problem where a bpf program may cause kernel crash due\nto the following error:\n Oops: divide error: 0000 [#1] PREEMPT SMP KASAN PTI\n\nThe failure is due to the below signed divide:\n LLONG_MIN/-1 where LLONG_MIN equals to -9,223,372,036,854,775,808.\nLLONG_MIN/-1 is supposed to give a positive number 9,223,372,036,854,775,808,\nbut it is impossible since for 64-bit system, the maximum positive\nnumber is 9,223,372,036,854,775,807. On x86_64, LLONG_MIN/-1 will\ncause a kernel exception. On arm64, the result for LLONG_MIN/-1 is\nLLONG_MIN.\n\nFurther investigation found all the following sdiv/smod cases may trigger\nan exception when bpf program is running on x86_64 platform:\n - LLONG_MIN/-1 for 64bit operation\n - INT_MIN/-1 for 32bit operation\n - LLONG_MIN%-1 for 64bit operation\n - INT_MIN%-1 for 32bit operation\nwhere -1 can be an immediate or in a register.\n\nOn arm64, there are no exceptions:\n - LLONG_MIN/-1 = LLONG_MIN\n - INT_MIN/-1 = INT_MIN\n - LLONG_MIN%-1 = 0\n - INT_MIN%-1 = 0\nwhere -1 can be an immediate or in a register.\n\nInsn patching is needed to handle the above cases and the patched codes\nproduced results aligned with above arm64 result. The below are pseudo\ncodes to handle sdiv/smod exceptions including both divisor -1 and divisor 0\nand the divisor is stored in a register.\n\nsdiv:\n tmp = rX\n tmp += 1 /* [-1, 0] -> [0, 1]\n if tmp >(unsigned) 1 goto L2\n if tmp == 0 goto L1\n rY = 0\n L1:\n rY = -rY;\n goto L3\n L2:\n rY /= rX\n L3:\n\nsmod:\n tmp = rX\n tmp += 1 /* [-1, 0] -> [0, 1]\n if tmp >(unsigned) 1 goto L1\n if tmp == 1 (is64 ? goto L2 : goto L3)\n rY = 0;\n goto L2\n L1:\n rY %= rX\n L2:\n goto L4 // only when !is64\n L3:\n wY = wY // only when !is64\n L4:\n\n [1] https://lore.kernel.org/bpf/tPJLTEh7S_DxFEqAI2Ji5MBSoZVg7_G-Py2iaZpAaWtM961fFTWtsnlzwvTbzBzaUzwQAoNATXKUlt0LZOFgnDcIyKCswAnAGdUF3LBrhGQ=@protonmail.com/"
