"value":"In the Linux kernel, the following vulnerability has been resolved:\n\ns390/bpf: Fix bpf_plt pointer arithmetic\n\nKui-Feng Lee reported a crash on s390x triggered by the\ndummy_st_ops/dummy_init_ptr_arg test [1]:\n\n [<0000000000000002>] 0x2\n [<00000000009d5cde>] bpf_struct_ops_test_run+0x156/0x250\n [<000000000033145a>] __sys_bpf+0xa1a/0xd00\n [<00000000003319dc>] __s390x_sys_bpf+0x44/0x50\n [<0000000000c4382c>] __do_syscall+0x244/0x300\n [<0000000000c59a40>] system_call+0x70/0x98\n\nThis is caused by GCC moving memcpy() after assignments in\nbpf_jit_plt(), resulting in NULL pointers being written instead of\nthe return and the target addresses.\n\nLooking at the GCC internals, the reordering is allowed because the\nalias analysis thinks that the memcpy() destination and the assignments'\nleft-hand-sides are based on different objects: new_plt and\nbpf_plt_ret/bpf_plt_target respectively, and therefore they cannot\nalias.\n\nThis is in turn due to a violation of the C standard:\n\n When two pointers are subtracted, both shall point to elements of the\n same array object, or one past the last element of the array object\n ...\n\nFrom the C's perspective, bpf_plt_ret and bpf_plt are distinct objects\nand cannot be subtracted. In the practical terms, doing so confuses the\nGCC's alias analysis.\n\nThe code was written this way in order to let the C side know a few\noffsets defined in the assembly. While nice, this is by no means\nnecessary. Fix the noncompliance by hardcoding these offsets.\n\n[1] https://lore.kernel.org/bpf/c9923c1d-971d-4022-8dc8-1364e929d34c@gmail.com/"
