)]}' { "commit": "aea45057c47d1547f5bcc898d9c2cb83fbc823a9", "tree": "452c0146928acc6193d6657b2f6185b83c81956a", "parents": [ "ff376a11626187fabdfd3864fdb7b15a5237c76a" ], "author": { "name": "Qiao Ma", "email": "mqaio@linux.alibaba.com", "time": "Tue Oct 15 14:01:48 2024 +0800" }, "committer": { "name": "Kevin Berry", "email": "kpberry@google.com", "time": "Sat Nov 09 05:52:34 2024 +0000" }, "message": "uprobe: avoid out-of-bounds memory access of fetching args\n\n[ Upstream commit 373b9338c9722a368925d83bc622c596896b328e ]\n\nUprobe needs to fetch args into a percpu buffer, and then copy to ring\nbuffer to avoid non-atomic context problem.\n\nSometimes user-space strings, arrays can be very large, but the size of\npercpu buffer is only page size. And store_trace_args() won\u0027t check\nwhether these data exceeds a single page or not, caused out-of-bounds\nmemory access.\n\nIt could be reproduced by following steps:\n1. build kernel with CONFIG_KASAN enabled\n2. save follow program as test.c\n\n```\n\\#include \u003cstdio.h\u003e\n\\#include \u003cstdlib.h\u003e\n\\#include \u003cstring.h\u003e\n\n// If string length large than MAX_STRING_SIZE, the fetch_store_strlen()\n// will return 0, cause __get_data_size() return shorter size, and\n// store_trace_args() will not trigger out-of-bounds access.\n// So make string length less than 4096.\n\\#define STRLEN 4093\n\nvoid generate_string(char *str, int n)\n{\n int i;\n for (i \u003d 0; i \u003c n; ++i)\n {\n char c \u003d i % 26 + \u0027a\u0027;\n str[i] \u003d c;\n }\n str[n-1] \u003d \u0027\\0\u0027;\n}\n\nvoid print_string(char *str)\n{\n printf(\"%s\\n\", str);\n}\n\nint main()\n{\n char tmp[STRLEN];\n\n generate_string(tmp, STRLEN);\n print_string(tmp);\n\n return 0;\n}\n```\n3. compile program\n`gcc -o test test.c`\n\n4. get the offset of `print_string()`\n```\nobjdump -t test | grep -w print_string\n0000000000401199 g F .text 000000000000001b print_string\n```\n\n5. configure uprobe with offset 0x1199\n```\noff\u003d0x1199\n\ncd /sys/kernel/debug/tracing/\necho \"p /root/test:${off} arg1\u003d+0(%di):ustring arg2\u003d\\$comm arg3\u003d+0(%di):ustring\"\n \u003e uprobe_events\necho 1 \u003e events/uprobes/enable\necho 1 \u003e tracing_on\n```\n\n6. run `test`, and kasan will report error.\n\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\nBUG: KASAN: use-after-free in strncpy_from_user+0x1d6/0x1f0\nWrite of size 8 at addr ffff88812311c004 by task test/499CPU: 0 UID: 0 PID: 499 Comm: test Not tainted 6.12.0-rc3+ #18\nHardware name: Red Hat KVM, BIOS 1.16.0-4.al8 04/01/2014\nCall Trace:\n \u003cTASK\u003e\n dump_stack_lvl+0x55/0x70\n print_address_description.constprop.0+0x27/0x310\n kasan_report+0x10f/0x120\n ? strncpy_from_user+0x1d6/0x1f0\n strncpy_from_user+0x1d6/0x1f0\n ? rmqueue.constprop.0+0x70d/0x2ad0\n process_fetch_insn+0xb26/0x1470\n ? __pfx_process_fetch_insn+0x10/0x10\n ? _raw_spin_lock+0x85/0xe0\n ? __pfx__raw_spin_lock+0x10/0x10\n ? __pte_offset_map+0x1f/0x2d0\n ? unwind_next_frame+0xc5f/0x1f80\n ? arch_stack_walk+0x68/0xf0\n ? is_bpf_text_address+0x23/0x30\n ? kernel_text_address.part.0+0xbb/0xd0\n ? __kernel_text_address+0x66/0xb0\n ? unwind_get_return_address+0x5e/0xa0\n ? __pfx_stack_trace_consume_entry+0x10/0x10\n ? arch_stack_walk+0xa2/0xf0\n ? _raw_spin_lock_irqsave+0x8b/0xf0\n ? __pfx__raw_spin_lock_irqsave+0x10/0x10\n ? depot_alloc_stack+0x4c/0x1f0\n ? _raw_spin_unlock_irqrestore+0xe/0x30\n ? stack_depot_save_flags+0x35d/0x4f0\n ? kasan_save_stack+0x34/0x50\n ? kasan_save_stack+0x24/0x50\n ? mutex_lock+0x91/0xe0\n ? __pfx_mutex_lock+0x10/0x10\n prepare_uprobe_buffer.part.0+0x2cd/0x500\n uprobe_dispatcher+0x2c3/0x6a0\n ? __pfx_uprobe_dispatcher+0x10/0x10\n ? __kasan_slab_alloc+0x4d/0x90\n handler_chain+0xdd/0x3e0\n handle_swbp+0x26e/0x3d0\n ? __pfx_handle_swbp+0x10/0x10\n ? uprobe_pre_sstep_notifier+0x151/0x1b0\n irqentry_exit_to_user_mode+0xe2/0x1b0\n asm_exc_int3+0x39/0x40\nRIP: 0033:0x401199\nCode: 01 c2 0f b6 45 fb 88 02 83 45 fc 01 8b 45 fc 3b 45 e4 7c b7 8b 45 e4 48 98 48 8d 50 ff 48 8b 45 e8 48 01 d0 ce\nRSP: 002b:00007ffdf00576a8 EFLAGS: 00000206\nRAX: 00007ffdf00576b0 RBX: 0000000000000000 RCX: 0000000000000ff2\nRDX: 0000000000000ffc RSI: 0000000000000ffd RDI: 00007ffdf00576b0\nRBP: 00007ffdf00586b0 R08: 00007feb2f9c0d20 R09: 00007feb2f9c0d20\nR10: 0000000000000001 R11: 0000000000000202 R12: 0000000000401040\nR13: 00007ffdf0058780 R14: 0000000000000000 R15: 0000000000000000\n \u003c/TASK\u003e\n\nThis commit enforces the buffer\u0027s maxlen less than a page-size to avoid\nstore_trace_args() out-of-memory access.\n\nLink: https://lore.kernel.org/all/20241015060148.1108331-1-mqaio@linux.alibaba.com/\n\nBUG\u003db/377471333\nTEST\u003dpresubmit\nRELEASE_NOTE\u003dFixed CVE-2024-50067 in the Linux kernel.\n\ncos-patch: security-high\nFixes: dcad1a204f72 (\"tracing/uprobes: Fetch args before reserving a ring buffer\")\nChange-Id: I5138ebc229ee8c008d3edd7b2aeb3743ad323060\nSigned-off-by: Qiao Ma \u003cmqaio@linux.alibaba.com\u003e\nSigned-off-by: Masami Hiramatsu (Google) \u003cmhiramat@kernel.org\u003e\nSigned-off-by: Sasha Levin \u003csashal@kernel.org\u003e\nReviewed-on: https://cos-review.googlesource.com/c/third_party/kernel/+/85845\nReviewed-by: Kevin Berry \u003ckpberry@google.com\u003e\nReviewed-by: Oleksandr Tymoshenko \u003covt@google.com\u003e\nTested-by: Cusky Presubmit Bot \u003cpresubmit@cos-infra-prod.iam.gserviceaccount.com\u003e\n", "tree_diff": [ { "type": "modify", "old_id": "0d52588329b29004704b03eac1f7c1429b1b9a9e", "old_mode": 33188, "old_path": "kernel/trace/trace_uprobe.c", "new_id": "d72ac9dde5321855a777fd73a1d0cd95c3f6f0e1", "new_mode": 33188, "new_path": "kernel/trace/trace_uprobe.c" } ] }