From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- Documentation/admin-guide/hw_random.rst | 104 ++++++++++++++++++++++++++++++++ 1 file changed, 104 insertions(+) create mode 100644 Documentation/admin-guide/hw_random.rst (limited to 'Documentation/admin-guide/hw_random.rst') diff --git a/Documentation/admin-guide/hw_random.rst b/Documentation/admin-guide/hw_random.rst new file mode 100644 index 000000000..bfc39f1cf --- /dev/null +++ b/Documentation/admin-guide/hw_random.rst @@ -0,0 +1,104 @@ +================================= +Hardware random number generators +================================= + +Introduction +============ + +The hw_random framework is software that makes use of a +special hardware feature on your CPU or motherboard, +a Random Number Generator (RNG). The software has two parts: +a core providing the /dev/hwrng character device and its +sysfs support, plus a hardware-specific driver that plugs +into that core. + +To make the most effective use of these mechanisms, you +should download the support software as well. Download the +latest version of the "rng-tools" package from: + + https://github.com/nhorman/rng-tools + +Those tools use /dev/hwrng to fill the kernel entropy pool, +which is used internally and exported by the /dev/urandom and +/dev/random special files. + +Theory of operation +=================== + +CHARACTER DEVICE. Using the standard open() +and read() system calls, you can read random data from +the hardware RNG device. This data is NOT CHECKED by any +fitness tests, and could potentially be bogus (if the +hardware is faulty or has been tampered with). Data is only +output if the hardware "has-data" flag is set, but nevertheless +a security-conscious person would run fitness tests on the +data before assuming it is truly random. + +The rng-tools package uses such tests in "rngd", and lets you +run them by hand with a "rngtest" utility. + +/dev/hwrng is char device major 10, minor 183. + +CLASS DEVICE. There is a /sys/class/misc/hw_random node with +two unique attributes, "rng_available" and "rng_current". The +"rng_available" attribute lists the hardware-specific drivers +available, while "rng_current" lists the one which is currently +connected to /dev/hwrng. If your system has more than one +RNG available, you may change the one used by writing a name from +the list in "rng_available" into "rng_current". + +========================================================================== + + +Hardware driver for Intel/AMD/VIA Random Number Generators (RNG) + - Copyright 2000,2001 Jeff Garzik + - Copyright 2000,2001 Philipp Rumpf + + +About the Intel RNG hardware, from the firmware hub datasheet +============================================================= + +The Firmware Hub integrates a Random Number Generator (RNG) +using thermal noise generated from inherently random quantum +mechanical properties of silicon. When not generating new random +bits the RNG circuitry will enter a low power state. Intel will +provide a binary software driver to give third party software +access to our RNG for use as a security feature. At this time, +the RNG is only to be used with a system in an OS-present state. + +Intel RNG Driver notes +====================== + +FIXME: support poll(2) + +.. note:: + + request_mem_region was removed, for three reasons: + + 1) Only one RNG is supported by this driver; + 2) The location used by the RNG is a fixed location in + MMIO-addressable memory; + 3) users with properly working BIOS e820 handling will always + have the region in which the RNG is located reserved, so + request_mem_region calls always fail for proper setups. + However, for people who use mem=XX, BIOS e820 information is + **not** in /proc/iomem, and request_mem_region(RNG_ADDR) can + succeed. + +Driver details +============== + +Based on: + Intel 82802AB/82802AC Firmware Hub (FWH) Datasheet + May 1999 Order Number: 290658-002 R + +Intel 82802 Firmware Hub: + Random Number Generator + Programmer's Reference Manual + December 1999 Order Number: 298029-001 R + +Intel 82802 Firmware HUB Random Number Generator Driver + Copyright (c) 2000 Matt Sottek + +Special thanks to Matt Sottek. I did the "guts", he +did the "brains" and all the testing. -- cgit v1.3.1