In the Linux kernel, the following vulnerability has been resolved: tracing: Fix race issue between cpu buffer write and swap Warning happened in rbendcommit() at code: if (RBWARNON(cpubuffer, !localread(&cpubuffer->committing))) WARNING: CPU: 0 PID: 139 at kernel/trace/ringbuffer.c:3142 rbcommit+0x402/0x4a0 Call Trace: ringbufferunlockcommit+0x42/0x250 tracebufferunlockcommitregs+0x3b/0x250 traceeventbuffercommit+0xe5/0x440 traceeventbufferreserve+0x11c/0x150 traceeventraweventschedswitch+0x23c/0x2c0 _traceiterschedswitch+0x59/0x80 _schedule+0x72b/0x1580 schedule+0x92/0x120 workerthread+0xa0/0x6f0 It is because the race between writing event into cpu buffer and swapping cpu buffer through file percpu/cpu0/snapshot: Write on CPU 0 Swap buffer by percpu/cpu0/snapshot on CPU 1 -------- -------- tracingsnapshotwrite() [...] ringbufferlockreserve() cpubuffer = buffer->buffers[cpu]; // 1. Suppose find 'cpubuffera'; [...] rbreservenextevent() [...] ringbufferswapcpu() if (localread(&cpubuffera->committing)) goto outdec; if (localread(&cpubufferb->committing)) goto outdec; buffera->buffers[cpu] = cpubufferb; bufferb->buffers[cpu] = cpubuffera; // 2. cpubuffer has swapped here. rbstartcommit(cpubuffer); if (unlikely(READONCE(cpubuffer->buffer) != buffer)) { // 3. This check passed due to 'cpubuffer->buffer' [...] // has not changed here. return NULL; } cpubufferb->buffer = buffera; cpubuffera->buffer = bufferb; [...] // 4. Reserve event from 'cpubuffera'. ringbufferunlockcommit() [...] cpubuffer = buffer->buffers[cpu]; // 5. Now find 'cpubufferb' !!! rbcommit(cpubuffer) rbendcommit() // 6. WARN for the wrong 'committing' state !!! Based on above analysis, we can easily reproduce by following testcase: bash #!/bin/bash dmesg -n 7 sysctl -w kernel.panic_on_warn=1 TR=/sys/kernel/tracing echo 7 > ${TR}/buffer_size_kb echo "sched:sched_switch" > ${TR}/set_event while [ true ]; do echo 1 > ${TR}/per_cpu/cpu0/snapshot done & while [ true ]; do echo 1 > ${TR}/per_cpu/cpu0/snapshot done & while [ true ]; do echo 1 > ${TR}/per_cpu/cpu0/snapshot done &
To fix it, IIUC, we can use smpcallfunctionsingle() to do the swap on the target cpu where the buffer is located, so that above race would be avoided.